Sensors 22 05693 v3

Download as pdf or txt
Download as pdf or txt
You are on page 1of 22

sensors

Review
A Survey on Modelling of Automotive Radar Sensors for
Virtual Test and Validation of Automated Driving
Zoltan Ferenc Magosi 1, * , Hexuan Li 1 , Philipp Rosenberger 2 , Li Wan 3 and Arno Eichberger 1

1 Institute of Automotive Engineering, Graz University of Technology, 8010 Graz, Austria;


hexuan.li@tugraz.at (H.L.); arno.eichberger@tugraz.at (A.E.)
2 Institute of Automotive Engineering, Technical University Darmstadt, 64289 Darmstadt, Germany;
philipp.rosenberger@tu-darmstadt.de
3 IPG Automotive GmbH, 76185 Karlsruhe, Germany; li.wan@ipg-automotive.com
* Correspondence: zoltan.magosi@tugraz.at; Tel.: +43-316-873-35275

Abstract: Radar sensors were among the first perceptual sensors used for automated driving. Al-
though several other technologies such as lidar, camera, and ultrasonic sensors are available, radar
sensors have maintained and will continue to maintain their importance due to their reliability in
adverse weather conditions. Virtual methods are being developed for verification and validation
of automated driving functions to reduce the time and cost of testing. Due to the complexity of
modelling high-frequency wave propagation and signal processing and perception algorithms, sensor
models that seek a high degree of accuracy are challenging to simulate. Therefore, a variety of
different modelling approaches have been presented in the last two decades. This paper compre-
hensively summarises the heterogeneous state of the art in radar sensor modelling. Instead of a
technology-oriented classification as introduced in previous review articles, we present a classifi-
cation of how these models can be used in vehicle development by using the V-model originating
from software development. Sensor models are divided into operational, functional, technical, and
Citation: Magosi, Z.F.; Li, H.; individual models. The application and usability of these models along the development process are
Rosenberger, P.; Wan, L.; Eichberger, summarised in a comprehensive tabular overview, which is intended to support future research and
A. A Survey on Modelling of development at the vehicle level and will be continuously updated.
Automotive Radar Sensors for Virtual
Test and Validation of Automated Keywords: radar sensor; machine perception; radar sensor model; automated driving; virtual testing
Driving. Sensors 2022, 22, 5693.
https://doi.org/10.3390/s22155693

Academic Editors: Daniel Rodriguez,


Changzhi Li and Marco Mercuri 1. Introduction

Received: 24 June 2022


It has already been shown that supporting the human driver in complex traffic sit-
Accepted: 26 July 2022
uations and transferring some or all of the driving tasks to automated driving systems
Published: 29 July 2022
increases traffic safety, efficiency of energy use and traffic flow, and travel comfort. Techno-
logical advances in semiconductors and information technology are enabling the develop-
Publisher’s Note: MDPI stays neutral
ment of increasingly sophisticated sensors, decision algorithms, and intervention elements.
with regard to jurisdictional claims in
The technology required for automated driving has now advanced to the point where
published maps and institutional affil-
thousands of kilometres can be travelled accident-free. Given the economic and societal
iations.
benefits of automated driving systems (ADS) and the level of technological development
required for their deployment, rapid market penetration would be warranted. Since the
development of motor vehicles and the market introduction of their functions are subject
Copyright: © 2022 by the authors.
to stringent legislation and validation measures, automotive manufacturers must ensure
Licensee MDPI, Basel, Switzerland. that automated driving functions (ADF) provide more safety than the human driver. In
This article is an open access article the automotive industry, testing and validation methods are used to ensure the desired
distributed under the terms and level of safety for less automated systems through the reliability of software and hardware
conditions of the Creative Commons components. With the introduction of driving automation systems (DAS) and automated
Attribution (CC BY) license (https:// driving systems, referred to as DAS/ADS systems, vehicles are becoming more complex,
creativecommons.org/licenses/by/ so the test and validation tasks go far beyond reliability testing of the vehicle’s hardware
4.0/). and software components.

Sensors 2022, 22, 5693. https://doi.org/10.3390/s22155693 https://www.mdpi.com/journal/sensors


Sensors 2022, 22, 5693 2 of 22

The methodology currently used in the automotive industry follows the V-model
development process. In this approach, system requirements are defined in parallel with
their verification and validation (V&V) activities throughout the development process. The
concept of V&V is to find an appropriate way to determine whether a product, function,
system, or subsystem meets or complies with safety requirements, specifications, and
regulations. A comprehensive V&V methodology is required to ensure that the complex
vehicle system will operate safely in an unsafe traffic environment. As system complexity
increases, traditional V&V methodologies, including on-road and test-bed testing, are no
longer sufficient to meet safety requirements through meaningful coverage of test cases
and scenarios. To close this gap, simulation must go beyond development support and
become a credible test tool for virtual approval and homologation of safety-critical ADS.
DAS/ADS systems combine capabilities that fit into one of the three segments of
the sense-plan-act model adopted from the robotics and automation literature, and this
model provides a high-level, implementation-independent view of a complex automated
vehicle system.
The testing and validation activities of the automotive industry are closely linked to
the regulatory framework. Product validation and quality assessment are continuously
performed according to industry-specific standards. Following the ISO-26262 standard (ISO-
International Organization for Standardization), one of the most important safety standards
in the automotive industry, conformity assessment is carried out through comprehensive
test procedures. The traditional test procedures of model-in-the-loop (MiL), software-in-
the-loop (SiL), processor-in-the-loop (PiL), hardware-in-the-loop (HiL), driver-in-the-loop
(DiL), and vehicle-in-the-loop (ViL), also referred to as X-in-the-loop, have led to detailed
testing and modelling of the integrated vehicle dynamics as well as the entire vehicle
system, providing well-developed models for testing the plan and act capabilities in the
framework. Sensor models have also been developed to improve DAS/ADS capabilities
and are used in state-of-the-art (SOA) environmental simulation software. However, there
is still a significant discrepancy between the results of real sensors and the results of most
sensor models in terms of all possible effects that occur during signal propagation and
processing, such as due to tire spray or multipath propagation. However, it should be
noted at this point that in order to perform simulations efficiently, models with varying
degrees of realistic representation must be used, depending on the stage of development
and the specific scenario.
In our review, we focus on radio detection and ranging (radar) sensor models because
they have special characteristics due to the complexity of radar wave propagation. This
complexity has led to a variety of proposed simulation approaches.
The remainder of the paper is structured as follows: Section 2 summarises challenges
in automotive testing procedures, Section 3 presents the vehicle development process
and the usage of sensor models, Section 4 gives an overview of previous survey in radar
sensor modelling, Section 5 introduces a new classification approach that is more useful for
selection of the appropriate modelling approach, and Section 6 discusses the findings of
the paper.

2. The Challenge in Automotive Testing


In recent years, more and more advanced features have been incorporated into vehi-
cles. The robustness and reliability of these systems is highly dependent on their sensing
capabilities, their processing of complex perception algorithms, and their operation by
electrical and/or electronic (E/E) systems. The testing and validation activities for such
distributed systems are already a complex task and are closely related to various national
and international laws, regulations, and industry standards. Szalay et al. identify in [1]
three different areas of automotive testing activities according to the life cycle manage-
ment process of automotive products. Figure 1 illustrates these different phases of vehicle
development, which will be explained in the next section. We therefore assume that the
Sensors 2022, 22, 5693 3 of 22

modelling approaches for radar sensors must differ in the different product development
phases.

Product Life cycle


Phase I. Phase II. Phase III.
Field monitoring updates
Product Development Product Approval Product Quality Control
ISO 26262 UN-ECE 3rd Party
V&V NCAP Accident databases
ISO/PAS 21448 USA Self-certification
Use Cases Definitions for scenario based testing
NCAP Test procedure used
in Product Development

Figure 1. Automotive product life cycle phases related to V&V of safety.

2.1. Vehicle Development Phase


In the first phase of the product life cycle, manufacturers are responsible for devel-
oping safe, reliable, and well-functioning vehicles. The basic legal framework for the
development of vehicles or safety-critical systems is set out in product liability legislation.
Product liability requires that the product placed on the market must provide a reasonable
expectation of safety and must be developed according to the state of the art. These meth-
ods are specified in national and international standards maintained by the International
Organization for Standardization and/or national standards bodies. The main objective
of standardisation work is to ensure the comparability and consistency of analysis results
performed independently by different companies [2]. In the automotive industry, develop-
ment processes are carried out according to the V-model proposed in ISO-26262 [3]. This
approach defines system requirements in parallel with their verification and validation
throughout the development process, including the software and hardware development
phases with the corresponding test activities, especially the various X-in-the-loop test
solutions. Standardised test and validation methods are only available for DAS functions
with lower levels of automation (SAE level L0-L2).
Requirements are usually set at the vehicle level; for longitudinal control, these are
defined for adaptive cruise control (ACC) in ISO-15622 [4] and for autonomous emergency
braking (AEB) in ISO-22733 [5]. For lateral guidance, there are standards for lane departure
warning (LDW) systems in ISO-17361 [6], for lane change decision aids systems (LCDAS)
in ISO-17387 [7], for partially automated lane change systems (PALS) in ISO-21202 [8], and
for lane keeping assistance systems (LKAS) in ISO-11270 [9].

2.2. Type Approval Phase


In the second phase of the product life cycle, vehicles are brought to market. If
an automobile manufacturer wishes to market its product in Europe, it must meet the
requirements of the United Nations Economic Commission for Europe (UN-ECE) in a
process known as type approval (or homologation). In type approval, vehicles are tested by
an independent third party (e.g., TÜV, Dekra), and approval is granted by the authorities.
In contrast, there is also the process of self-certification (e.g., in the U.S.), in which the
vehicle manufacturer analyses and certifies the safety strategy through a voluntary safety
self-assessment to confirm that the product meets market requirements [1]. The criticality
of these testing activities is much higher than for feature development. The reason is that
if type approval/homologation is successful, the vehicle is deemed safe by the authority
and therefore must meet the authority’s specifications and the public’s expectations [2].
For longitudinal control, type approval is specified in UN-ECE R 131 for the advanced
emergency braking system (AEBS) [10] and for lateral control in UN-ECE R 157 [11] for the
automated lane keeping system (ALKS).
Sensors 2022, 22, 5693 4 of 22

2.3. Consumer Protection Phase


In the third phase of the product life cycle, products are already on the market and
consumer organizations must demonstrate the quality of the product to protect consumers
from the risk of unfair commercial interactions. In the automotive industry, the New Car
Assessment Program (NCAP) is one of the best-known consumer protection organiza-
tions [1]. It was established in the late 1970s by the U.S. National Highway Traffic Safety
Administration (NHTSA) and focuses on testing the major operating and safety systems of
motor vehicles. Later, NCAPs were adopted by the automotive industry in other parts of
the world, such as the European New Car Assessment Program (Euro NCAP), the Japanese
New Car Assessment Program (JNCAP), and the Chinese New Car Assessment Program
(C-NCAP) [12]. Euro NCAP’s test protocols, developed based on real-world accident
scenarios, are not just a test and evaluation procedure for the final product. The automotive
industry also uses them as a tool to directly improve safety by incorporating them into
the vehicle development process. By virtually performing NCAP test evaluation protocols
with the MiL and SiL simulation methods, development engineers can elaborate and verify
the vehicle’s perception strategy in a time- and cost-efficient manner at an early stage of
development, during the concept phase. They also provide a good basis for functional
testing of software and hardware prototypes on HiL test benches in later development
phases [13].

3. Vehicle Development Process


The inherent complexity of modern systems is increasing significantly. Consequently,
the system development process and the verification and validation process are becoming
increasingly complex. In order to cope with this major challenge, suitable development
processes have been introduced. Combining these with appropriate simulation techniques,
different system designs can be evaluated and the number of physical prototypes can be
reduced. The increasing complexity of modern vehicle systems requires a modular system
design, both in terms of the integrated hardware components and the algorithms running
on them. This means that the operation and correct behaviour of the system depend on the
interactions between the modules and the content and quality of their input and output
information. Therefore, it is becoming increasingly important to test and evaluate the entire
system in its intended operating environment. Traditional system design methods address
this complexity problem by providing appropriate process models that contain detailed
specifications of all components as well as the overall system with all relevant interfaces
and relationships. The product development method currently used in the automotive
industry is based on the V-model as proposed in the ISO-26262 standard [3] (see Figure 2).
Product development starts with a hierarchical top-down analysis and design phase,
followed by implementation and a reverse bottom-up integration and test phase. Test
tools corresponding to the development phases, such as MiL, SiL, PiL and HiL simulations,
are also provided. The development of modern systems requires the integration of many
disciplines, leading to a need for standardised interfaces and coordination between the
standard methods of the disciplines involved [14]. To meet this challenge, the methodology
of model-based systems engineering (MBSE) plays an increasingly important role in system
design, that is, the formalised application of modelling to support the entire development
process at all levels of abstraction [15]. The integration of virtual models into the vehicle
development process has led to the systematic use of simulation techniques that enable
virtually based V&V to prove the correctness of the system already in the requirements
analysis phase. One of the key points of this methodology is the V&V concept, as different
models of varying complexity can be used for the V&V activities in the different phases of
the product development cycle, depending on their complexity. Once the complete system
is designed, specified, and verified, it can be implemented, integrated, and validated. The
validation processes are designed to ensure that all requirements arising from the ’safety
by engineering’ strategy along the left-hand side of the V-model in a decreasing direction
are met, that known scenarios are covered, and that the system behaves as specified. In the
Sensors 2022, 22, 5693 5 of 22

validation processes, which represent the right side of the V-model, the verified system is
tested using well-defined test methods in an ascending direction to confirm that the system
meets all safety design requirements and behaves as intended and specified.

Requirements Verification Use cases definition


Open Road Type Approval NCAP
Full-Vehicle Full-Vehicle Function and Performance test
Verification
Operational Concept
Technology specific HW and
Full-System Verification
Customer Functions/Items SW knowledge and
experience gained over the
OEM OEM years available.
Sub-System
Verification Performance
Distributed Functions
Validation

Perception Technology
Knowledge Knowledge GAP
Physical Sensor
Overlap Sense
Behaviour
Plan
Act
Supplier

Technology specific HW and


MiL: Model-in-the-Loop SW knowledge and
SiL: Software-in-the-Loop experience gained over the
PiL: Processor-in-the-Loop years available.
HiL: Hardware-in-the-Loop Supplier
DiL: Driver-in-the-Loop Performance
PG: Proving Ground Validation

Figure 2. Vehicle development process with respect to driving automation.

In order to validate the safety of the intended functions and systems, the vehicles are
tested in all phases of development using a variety of predefined realistic test scenarios in
virtual and real traffic. A number of validation tools are already defined in the standards
mentioned above, such as SiL and HiL. At the level of the complete vehicle, real test drives
are performed on the proving ground, but also on public roads, to ensure that the systems
function properly in road traffic. Combining simulations in the above context with tests
at the vehicle level is intended to establish statistical confidence in operational safety. As
specified in the ISO/PAS-21448 Safety of the Intended Functionality [16] and ISO-26262
Functional Safety standards, once the ADS is on the market, the safety of the system will be
continuously monitored through field operational test (FOT) by collecting and analysing
anonymous data from the field during on-road testing.
Figure 2 also illustrates that there is a knowledge gap in V&V of automotive radar
sensor based systems in the phase where there is a hand-over from the system supplier to
the vehicle’s original equipment manufacturer (OEM). It also includes a proposal of which
overall modelling approaches are appropriate in the different phases. These approaches
will be explained in Section 5.

4. State-of-the-Art Radar Sensor Model Classifications


In recent years, many different approaches to modelling radar sensor systems have
been developed. An early example of radar sensor modelling can be found in [17], from
1990. There are a number of different approaches to generating synthetic radar data. The
authors in [18] distinguish between ideal models, probabilistic models, and physics-based
models. They also distinguish three fidelity levels for sensor models with increasing
complexity: low-fidelity models, medium-fidelity models, and high-fidelity models [19].
Ideal sensor models simply generate ground truth (GT) data for objects in the sensor’s
detection range. This allows sensor-type-specific object list information to be generated for
all objects in the sensor’s field of view (FOV). In contrast to ideal sensor models, physics-
based sensor models attempt to model the physical sensing process of the real sensor as
accurately as possible. They are computationally intensive and require more computational
power, often at the expense of real-time capability. In addition to the required expertise
in sensor technology, a detailed description of the environmental conditions (material
Sensors 2022, 22, 5693 6 of 22

properties, weather conditions, etc.) is required to accurately model the physics of the
sensor. The simulated output data from these models can be a raw analog signal comparable
to that of the real sensor. As suggested in [20,21], the use of probabilistic sensor models
can provide a reasonable trade-off between complexity and computational efficiency. The
simplified parameter set and reduced model complexity allow simulation tests to be
performed faster than real time. The output of these models can be set up to provide object
lists or even raw data. Even though the output data is less realistic compared to the real
sensor, phenomenological models can be used at most levels of the development process to
test and validate the safety of the ADS.
Rosenberger et al. in [22] distinguish between ground truth models that neglect any
operation on the GT object list except for the transformation from world to sensor origin
coordinates, idealised models that additionally cut out the FOV of the sensor from the
object list, and phenomenological models that consist of stochastic and physical model
parts. In radar modelling, this could mean that a ray tracing approach to physical modelling
of signal propagation is accompanied by a stochastic Gaussian noise model. While pure
stochastic modelling is often applied in sensor models for high-level object list output, pure
physical modelling is almost completely avoided in scenario-based simulation due to the
computational time required for such finite element method (FEM) simulation or the like.
In [23], the author categorises sensor models into three groups, defined as ideal,
physical, and functional. Ideal sensor models directly read any measurable objects or GT
information provided in the virtual environment without including any real sensor-related
uncertainty. In contrast, physical models or white-box models implement the real physical
sensor properties, but at the expense of real-time simulation performance. The functional
model ignores the sensor hardware architecture and signal processing process. Such a
black-box model focuses only on the detection result of the measured object. This type of
sensor model can be implemented by combining certain geometric and selected physical
properties. This improves the real-time performance of the sensor simulation without
completely ignoring the detection limits and characteristics of the real sensor.
Similarly, in [24], sensor models are classified as function-based, physics-based, or a
combination of both. For simplicity and real-time capability, the function-based modelling
approach includes geometric models and models dealing with scattering centres. The
physics-based approach has been divided into two classes, one dealing with modelling the
electronic components of the radar sensor, including the propagation channel modelled
by ray-tracing techniques. The other approach considers radar and clutter echoes as well
as noise. Another approach is presented that distinguishes between a radar model and
a radar system model, where the radar system model includes the environment and the
target vehicle model in addition to the radar model.
Holder et al., in [25], distinguish three groups of sensor models and define them
as follows. The ideal sensor models generally generate a list of perfectly sensed objects
from the simulation environment, that is, they do not model sensing errors. These are
followed by phenomenological sensor models that already take into account additional
sensor properties such as the FOV of the sensor, limited resolution, and measurement
uncertainties [26,27]. Finally, physical sensor models aim to reproduce the raw sensor data
by modelling the physical phenomena specific to the sensor being modelled. The authors
in [28–30] provide an assessment of radar sensor models in the literature, based on some
predefined modelling criteria, with the goal of helping the modeller estimate the effort
required to create such models. Based on their defined criteria, radar sensor models can be
classified into three categories. In the physical sensor model, called the white-box model, all
physical aspects of the radar are considered and calculated based on a detailed description
of the environment [24,31,32]. The scattering centre sensor models exploit the property,
known from radar cross-section (RCS) studies, that electromagnetic scattering from an
electrically large target can be approximated by a sparse set of points at a fixed position
on the target, called scattering centres (SC) [33–35]. Data-driven or black-box models do
Sensors 2022, 22, 5693 7 of 22

not require information about the behaviour of the radar and can only learn its operation
based on recorded data from real experiments [29,36,37].
Schlager et al. [19] determine the accuracy of a sensor model based on its inputs,
outputs, and modelling principle. Low-fidelity sensor models are based on geometric
aspects such as the sensor-specific FOV and object occlusion. The input data format is
object lists with ground truth information, and the output data format is also object lists
but with filtered GT information [38]. Medium-fidelity sensor models take into account
some physical aspects of the real sensor and material properties of the objects, as well as
the sensor’s field of view and detection probability. The input for medium-fidelity sensor
models are object lists corresponding to ground truth. The output data formats are object
lists or raw data processed, according to the modelled perceptual effects [39]. High-fidelity
sensor models are the most accurate representations of real-world sensors. They incorporate
rendering methods such as rasterisation or ray tracing. They combine environmental
parameters and material properties as well as physical effects such as diffraction and
interference. High-fidelity sensor models use the entire 3D virtual environment, a mesh
grid describing objects and their surfaces, as input and produce sensor-specific raw data
as output [40].
Cao et al. [41] introduced the white-box, grey-box, and black-box classifications
for sensor models. Black-box models can be used for system function verification but
not for validation. White-box modelling is the simulation of electromagnetic (EM) wave
propagation by solving Maxwell’s equations, simulating semiconductor components and
the propagation channel. Grey-box modelling is an effective combination of the above
models in terms of complexity and real-time capability.
Ngo et al. [42] distinguish time-domain electromagnetic simulation techniques, ray
tracing, data-driven, and idealised modelling approaches. They describe a method for
evaluating a radar sensor model by comparing the results of clustering algorithms with
real and synthetic radar data and provide a sensitivity analysis for the various parameters
of their radar sensor model.
In a later publication, Ngo et al. [43] describe common sensor model categorisation
into physics-based, probabilistic, and phenomenological. They also describe the expansion
stages of their radar models, which are later used to demonstrate their multilayer model
validation study: ideal radar model (IRM), data-driven model (DDM), and ray-tracing-
based model (RTM).
Holder in [44] classifies sensor models according to the information flow in the simu-
lation, the model input (I) and output (O), and the error modelling in between. Six levels
are distinguished along this path. The model input can be vectorised information such
as an object list (O) or rendering (R). The output is distinguished between object list (O),
detections (D), or raw data (R). With these abbreviations, a naming scheme is provided
from a combination of them: for example, object in object out (OIOO) or rendering in
detection out (RIDO), and so on.
A new area of research in the modelling community combines the virtual world with
real-world hardware components to drive a dedicated HiL testbed for radar sensor stimu-
lation, because they provide the ability to thoroughly test radar sensors under laboratory
conditions [45]. With advances in analog and/or digital millimetre-wave signal processing
technology, powerful real-time radar target stimulators have emerged that can accurately
stimulate the radar signature of real targets represented by point scatterers. The general
operating principle is that the receiver mixes the received radar signal into the baseband,
digitises the baseband signal, modifies the waveform on one or more FPGAs, and converts
it into an analog signal. Possible signal modifications include amplitude and phase changes,
time delays, and frequency shifts. Finally, the mixed radar echo is transmitted back over
the air at the transmitter end. In [46], an HiL approach is used that emulates a virtual
radar environment corresponding to a defined test scenario. The relevant test scenarios are
parameterised and then mapped to the antennas of the target stimulator. A radar HiL test
approach based on OTA stimulation can be found in [47,48]. In [47], a simulation platform
Sensors 2022, 22, 5693 8 of 22

is developed using multi-body simulation (MBS) software to combine traffic and scenario
simulations. The radar echoes are mixed with Gaussian noise to improve the realism
of the test. Based on the flexibility and scalability of available high-frequency hardware
components combined with MBS, the authors presented a whole-vehicle-level DAS/ADS
testbed based on OTA radar target stimulation in [49–51]. In [52], a dynamic OTA radar
stimulator is demonstrated. The system provides the illuminated antennas with three
degrees of freedom of motion to simulate more complex scenarios with different angular
positions of the target vehicle. The authors in [53] calculate the radar characteristics using
theoretical formulations and implement them in FPGA hardware components to produce
even more realistic radar echoes. Another ViL test method based on the OTA radar target
stimulation approach is presented in [54]. The ambient perception simulation is based on
statistical distributions and the radar signature of the target is estimated as a function of
the target vehicle dynamics. In addition to the aforementioned development and research
systems, radar target stimulators are also used in series production as end-of-line (EOL)
test equipment.
The open integration and test platform CarMaker of IPG Automotive GmbH classifies
sensor models into three groups: Ideal, High Fidelity (HiFi) and Raw Signal Interface
(RSI) [55]. Ideal sensor models represent a generic interface. Object information can be
accessed within the defined sensor range, and the model is technology-independent. High-
fidelity models correspond to phenomenological sensor models and provide a higher level
of detail than ideal models. They use ideal environmental information and overlay it with
technology-dependent effects known from theory and measurements. Physical models
with an RSI account for actual signal propagation. This includes the main physical effects
involving the interaction of the signal with objects in the simulation and the transmission
media along the propagation path.
Furthermore, most simulation tools provide different interfaces for sensor modelling
with varying complexity [55]. We used the definition given in [56], which describes the
complexity of the model as follows:
The ground truth model contains all information about all objects within the search
radius. The geometric models take into account some basic information from the sensor’s
data sheet and map the detection area and field of view. The stochastic models assume some
uncertainty in the detection probability and provide parallel measurements with artificially
generated noise. The most complex models for perceptual sensors are physics-based
models that use the physical properties of the object, wave propagation, and reflections.
From the point of view of computational requirements and reduced parameter space,
phenomenological models are a good alternative since mathematical methods can be used
to simulate a sensor-specific phenomenon. This classification, presented in Figure 3, serves
as the basis for a classification more focused on vehicle development applications, as
described below.

Target
Search Radius Search Radius

EGO EGO

Ground-Truth Geometric Stochastic Phenomenological Physics-based


Model Model Model Model Model

Operational Models Functional Models/Technical Models Individual Models

Figure 3. Sensor modelling approaches .


Sensors 2022, 22, 5693 9 of 22

5. Classification from the System Integrator’s Perspective


Modeling the performance of perceptual sensors at different levels of abstraction in
the development process is critical because it provides a preliminary estimate of sensing
capabilities, enabling the development and verification of different sensing strategies that
are essential for automated driving functions [57]. The radar sensor models available in the
literature have been studied extensively by researchers and, as described earlier, have been
classified into many different categories.
X-in-the-loop test methods for verifying and validating the safety of ADSs range from
field tests to simulation of all subcomponents of the complete vehicle. Due to the high
complexity of an ADS, test activities are shifted to the virtual test domain by replacing one
or more physical components with an appropriate simulation model [20,58]. The scenario-
based method can be applied to the X-in-the-loop testing methodology throughout the
development process described in Section 3 because the scenario description is based
on different levels of abstraction or detail. Our motivation to introduce a new naming
convention for the state-of-the-art radar sensor models is based on the scenario abstraction
levels [59] and the X-in-the-loop test methods [60], which are jointly mapped to the phases
of the development process represented in the V-model [61].
We have also defined two prerequisites: First, that the automotive manufacturer
involves an DAS/ADS supplier with specialised knowledge and unique experience in the
field of environmental perception sensing at the subsystem or at least component level.
Second, radar sensor models with a more detailed output data level than object lists, such as
target/cluster lists or raw signals, cannot be integrated into the vehicle development process
by automotive manufacturers due to a lack of technological knowledge and hardware
resources appropriate to high-frequency (77–79 GHz) automotive radar technology. It is
assumed that these models are more likely to be used on the supplier side.
In increasing order of complexity, we introduce operational models , functional models,
technical models, and individual models. The application of the different models in the vehicle
development process is illustrated in Figure 2 and described below.
Following the V-model, the sequence of development phases is: operational > functional
> individual > technical > functional. For OEMs, however, only the operational, functional,
technical, and functional radar sensor models can be used. After applying the technical models
in HiL testing at the component and sub-system level, functional models can be reused for
DiL and ViL testing, for example, for testing the human–machine interface (HMI).

5.1. Operational Model


5.1.1. Definition
The term operational means “in or ready for use”; ground truth and geometric models
are considered. In the concept phase, simplified sensor models can be used to specify the
perception concept of the automated driving system. For example, it must be determined
which areas of the vehicle environment are to be perceived and at what distance objects
must be detected. For this purpose, typical sensor properties such as sensor FOV, detection
range, and so on can be modelled easily and quickly even without specific knowledge of
perception sensor technology.

5.1.2. Application
The automotive industry has a wide range of simulation tools to support the devel-
opment process and to accelerate V&V test activities. In the concept phase, simplified
sensor models can be used to specify the perception concept of the automated driving
system. For example, it must be determined which areas of the vehicle environment are
to be perceived and at what distance objects must be detected. In the further course of
development, sensor models can support the selection of the sensor technology to be used
for the automated driving system (radar, lidar, camera, etc.). For this purpose, typical
sensor characteristics such as sensor FOV, detection range, and so on can be modelled. With
the ability to generate the GT for each simulation step, all of these modern simulation tools
Sensors 2022, 22, 5693 10 of 22

have easy-to-parameterise, built-in, and technology-independent generic sensor models.


These models are often referred to as idealised or geometric models and provide object
lists as output data. As examples, we mention some products that are widely used in
the automotive industry: in [62], TwT GmbH, TASS-PreScan, dSpace-ASM; in [57], TESIS
Dyna4-Driver Assistance, MathWorks-ADAS Toolbox; in [63], CARLA, AirSim, DeepDrive,
Udacity, Constellation, Helios, GLIDAR, RADSim, SIMSonic; and in [64], CarMaker from
IPG Automotive GmbH, VIRES-VTD, CARLA, and AirSim can provide GT information.
In vehicle development, operational models are used in the descending branch of the
V-model, in the early stage where the operational concept of the vehicle is tested with
abstract scenarios against the requirements (see also Figure 2).

5.2. Functional Models


5.2.1. Definition
The term functional means "of or having a special activity, purpose, or task”. Stochastic,
phenomenological, and data-driven models are being considered with increasing attention.
Given a particular sensor technology, such as the radar sensor, a functional representation of
the acquisition process can be modelled by simulating the antenna properties with a simple
cone and a complex target with a cuboid. Discretised scatterers can then be generated from
the object, which can additionally be overlaid with noise to provide a more realistic object
list. To accommodate different design considerations, the complexity can vary over a wide
range. These models may require moderate sensor-specific knowledge to realise realistic
behaviour with a reduced parameter space.

5.2.2. Application
The output of a functional model usually does not deal with the internal processes or
algorithms of the real sensor, but focuses on reproducing the effects that distinguish the
sensor output from the reference data. Unlike operational models, functional models already
contain more information and details about the real sensor properties. The authors in [27,36]
illustrate this point that a non-parametric modelling approach is able to model sensor range,
occlusion, latency, ghost objects, and object loss without explicit programming, and can
be used efficiently in real-time simulation. The same concept is developed in [26,65],
where the geometric information of the target is transformed into the sensor model, and
then the signal noise and statistically based signal loss are superimposed on the original
signal. The method described above has provided good estimation and modelling of
relative distance, velocity, and other sensor-specific information. The data-driven approach
requires a large number of experiments to obtain a statistical distribution that can be applied
by the model. However, in the real world, there are often crucial parameters that affect
the detection results, and a given statistical distribution may not do justice to the sensor’s
detection performance. Therefore, a data-driven approach based on machine learning (ML)
is introduced. In the work of [64], different ML methods are investigated and used to
build RCS models, demonstrating that better prediction accuracy can be achieved with
ML models. In addition to data-driven methods, geometric-based approaches are also
commonly used for radar feature modelling. The geometric approach focuses more on the
specific details of the target and models according to the statistics of the reflection points at
different locations on the surface of the target to create the sensor-specific object list.
In vehicle development, functional models are used in the descending branch of the
V-model after the use of operational models, but can also be reused in the ascending branch
after the use of technical models. For more detailed logical scenarios, functional models can
be used at the subsystem level in the design phase to produce sensor-technology-specific
outputs that are used as inputs to a sensor fusion algorithm. In addition, the models can
be used to verify that real sensors meet the requirements of the system specifications. In
this regard, see also Figure 2. The figure shows that functional models can be used not
only for verification purposes in the design phase, but also in the ascending branch of the
V-model in the integration phase for testing. Since functional models are expected to perform
Sensors 2022, 22, 5693 11 of 22

their task in real time or even faster than real time, they can replace the real sensors for
functional testing of the vehicle’s HMI in DiL testing or for testing some vehicle functions
at the system level in ViL testing methods.

5.3. Technical Models


5.3.1. Definition
“Involving or concerned with applied and industrial sciences”, models for over-the-air
(OTA) target stimulator testbeds are considered. This group of radar sensor models includes
models developed for a specific well-defined high-frequency radar target stimulator. These
models have much lower performance compared to their simulation-only counterparts and
are developed according to the available hardware components. These models typically
provide only radial range, angular position, radial range rate, or velocity, as well as RCS
information for radar signature generation in the form of a point target.

5.3.2. Application
Technical models provide object lists with reduced content to represent a radar signature,
usually in the form of a point target. The model output is the input data for a particular
radar target stimulator high-frequency HiL testbed. A typical radar signature consists
of: Doppler shift f d due to relative velocity, range in the form of propagation delay ∆t,
spatial direction (azimuth Φ, elevation Θ), and RCS σ describing the effective area of the
identified objects. Once the concept of the perceptual sensor system is fully defined, the
integration phase begins (right side of the V-model). For the integration of initial hardware
prototypes with mostly limited functionality, technical sensor models provide input signals to
validate the intended functionality on HiL, DiL, or even ViL test benches. Due to the lack of
detailed technological knowledge of the subcomponents and the complex high frequency
technology, the focus is on quantitative rather than qualitative or performance analysis.
Although OTA radar sensor simulation is widely used in co-simulation for HiL and
ViL, the huge investment in hardware equipment is still a challenge. In addition, due to
the high computational requirements of the real-time system, the effects of environmental
conditions on the radar echo are often ignored or reduced to a probability distribution.
Furthermore, the number of objects to be simulated is also limited.
In vehicle development, technical models are used in the ascending branch of the V-
model, after the usage of individual models. The test cases are defined in concrete scenarios
with well-defined requirements. Technical models support X-in-the-Loop on vehicle level,
see also Figure 2.

5.4. Individual Model


5.4.1. Definition
The term individual means "single; separate, for a particular use”; physically based
models are considered here. Parametrisation is only possible with the expertise of the
system supplier. To verify the detection performance of an individual sensor or a sensor
cluster under non-optimal detection conditions, sensor-typical phenomena such as range
reduction, multipath wave propagation, reflections, attenuation, or the unexpected rapid
changes of the radar cross-section have to be modelled.

5.4.2. Application
Under the individual model, physical models are considered, provided the sensor
supplier has all the technology and hardware-specific parameters to perform a reliable
performance evaluation. The application is in the lower part of the V-model, at the com-
ponent or subsystem level, where performance verification can only be performed by the
supplier. Models that provide raw data, reflection points and target lists belong to this
class of models because the data processing algorithms for clustering and tracking are not
known to the vehicle integrator. Individual models can be used effectively by system suppli-
ers who have the technological knowledge to perform simulations of everything down to
Sensors 2022, 22, 5693 12 of 22

semiconductor components. Examples include ray-tracer-related models, any time-domain


electromagnetic wave simulation, finite-difference time-domain (FDTD) method, finite
element method (FEM), raw input-raw output (RIRO) method, raw input-object-output
(RIOO) method, and so on [44].
Academic and industrial researchers are making great efforts to develop an efficient,
easy-to-use sensor model that behaves like the real sensor. However, without the specific
infrastructure, such as an echo chamber the size of a vehicle [66], or knowledge of charac-
teristic parameters, such as the antenna radiation pattern [32], the modelling can only be
based on compromises and must focus on a very specific domain or use case [67]. These
compromises include simplifications and assumptions, for example:
• replacing of the real material description model by a probabilistic material model
in [68];
• assuming the radiation pattern of the antenna is known in [32];
• replacing complex objects by multiple scattering centres in [69];
• or treating all metallic surfaces as perfect conductors (PEC) [54], while considering all
other materials as absorbing in [40].
The authors show in [70] that a virtual representation of a real perceptual sensor in the
form of a physics-based model is possible if the right supplier information and resources
are available. In this paper, some basic properties of radar detection are investigated by
measurement and calibrated simulation, including RCS estimation using ray tracer method.
The main contribution of this work is that the calibration parameters for simulation can
be derived from real measurements if the key parameters of the real sensor hardware
are known.
In vehicle development, individual models are used in the ascending branch of the
V-model, after the usage of functional models. The test cases are defined in concrete scenarios
with well-defined requirements. Individual models support X-in-the-loop at the component
and subsystem leve (see also Figure 2).

5.5. Classification Overview


Summarizing the results of our review in Section 4, we list existing approaches and
review papers and the related classification approach in Table 1. Approaches can be found
in the rows, and previous mentions in survey papers in columns.
We also organised the radar sensor modelling approaches found in the literature in
our new classification scheme including operational, functional, technical, and individual
models. For better readability, the whole table including the classification for each review
paper is available at https://doi.org/10.3217/kgg17-wq710 (accessed on 24 June 2022 ).
Here, the existing classifications from previous review papers are also mentioned.
The last row summarises how often the approach is mentioned in an existing review
paper, which is an indicator of the recognition of the approach. A number zero means that
previous reviews did not classify the approach.

Table 1. Classification of radar sensor model approaches and relation to previous reviews.

[23] [25] [71] [30] [19] [41] [43] [42] [44] [67] [18] Sum:
[72] x x 2
[38] x x 2
OPERATIONAL

[73] x x x 3
[18] x 1
[74] x x 2
[75] x 1
[28] x 1
[76] 0
[77] x 1
Sensors 2022, 22, 5693 13 of 22

Table 1. Cont.

[23] [25] [71] [30] [19] [41] [43] [42] [44] [67] [18] Sum:
[27] x x x x 4
[36] x 1
[26] x x x x 4
[41]
x x x x x 5
[39]
[65] 0
[78] x x x 3
[79] 0
[33] x x 2
[64] 0
[80] 0
[81] x 1
[82] 0
[83] x 1
[29] x x x x x x 6
[84] 0
[24] x x 2
FUNCTIONAL

[71] x 1
[85] 0
[86] 0
[87] 0
[37] x x 2
[30] 0
[88] x 1
[89] x 1
[90] x 1
[91] x 1
[57] x 1
[92] x 1
[93] x 1
[94] x 1
[95] x 1
[96] 0
[97] 0
[98] x 1
[99] x x 2
[45] 0
[46] 0
[47] 0
[48] 0
[49] 0
TECHNICAL

[50] x 1
[100] 0
[51] 0
[52] 0
[53] x 1
[101] x 1
[102] x 1
[103] 0
[54] x x 2
Sensors 2022, 22, 5693 14 of 22

Table 1. Cont.

[23] [25] [71] [30] [19] [41] [43] [42] [44] [67] [18] Sum:
[69] x 1
[32] x x x x x 5
[68] 0
[67] x 1
[70] 0
[104] x 1
[105] 0
[35] x x x x 4
[106] x x 2
[40] x x 2
[107] 0
[108] 0
[109] x 1
[110] x x 2
[111] x 1
[112] x 1
[31] x 1
[34] x 1
[113] x 1
[114] x x 2
[115] x 1
FUNCTIONAL

[116] x 1
[117] x 1
[118] x x 2
[119] x 1
[120] x 1
[121] x 1
[122] x 1
[123] x 1
[124] x x 2
[125] x 1
[126] x 1
[127] x 1
[128] x 1
[129] x 1
[130] x 1
[55] x 1
[131] x 1
[132] x 1
[133] 0
[134] 0
[135] 0
[136] x 1

6. Discussion
As complexity increases, the verification and validation (V&V) of automated driv-
ing systems (ADS) becomes exponentially more inefficient in terms of time-to-market
and cost when the focus is on on-road testing. A variety of X-in-the-loop methods have
been introduced to support efficient V&V of the safe performance of ADS. However, the
quality of the predictions depends mainly on the ability of the V&V method to replicate
the performance of real machine perception. Automotive radar sensors are superior in
certain characteristics, such as performance in adverse weather conditions and precision in
measuring the relative speed and distance of objects, especially moving objects. However,
radar sensors are difficult to model due to the complex physical relationships between mul-
tipath propagation and electromagnetic wave reflection. In previous research, numerous
attempts have been made to develop models using very different modelling techniques.
In previous literature reviews, different classifications were established to structure the
Sensors 2022, 22, 5693 15 of 22

large amount of available research. This paper summarises the methods described in the
literature, but also introduces a new perspective. Since X-in-the-loop-based approaches
ultimately support the goal of whole-vehicle-level development, we have classified the
available approaches in a perspective of how they are used in the development process and
have introduced—in increasing order of complexity and use along the V-model integration
approach— operational models, functional models, technical models, and individual models.
We summarised the different approaches and classifications and provided a com-
prehensive table that combines previous classifications with our new approach. Thus,
the reader is able to quickly get an overview and select a suitable modelling method for
further use. Finally, we provide a link to a dynamic spreadsheet that is publicly available
at: https://doi.org/10.3217/kgg17-wq710. (accessed on 24 June 2022) This spreadsheet is
being continuously enhanced as more progress in radar sensor modelling is achieved and
will include comments of researchers and readers in that tabular overview in future.

Author Contributions: Conceptualization , Z.F.M., H.L., and A.E.; methodology, Z.F.M., H.L., and
A.E.; investigation, Z.F.M., H.L., and A.E.; resources, Z.F.M., H.L., and A.E.; writing—original draft
preparation, Z.F.M., H.L., A.E., P.R., and L.W.; writing—review and editing, Z.F.M., H.L., and A.E.;
visualization, Z.F.M.; supervision, A.E.; project administration, A.E. All authors have read and agreed
to the published version of the manuscript.
Funding: Open Access Funding by the Graz University of Technology. This activity is part of the
research project InVADE (FFG nr. 889349) and has received funding from the program Mobility
of the Future, operated by the Austrian research funding agency FFG. Mobility of the Future is
a mission-oriented research and development program to help Austria create a transport system
designed to meet future mobility and social challenges.
Informed Consent Statement: Not applicable.
Data Availability Statement: The full classification data, presented in a comprehensive tabular
format can be found at: the link has been changed to the same link as in the discussion section
https://doi.org/10.3217/kgg17-wq710 (accessed on 24 June 2022). In addition, the literature cited in
this table is provided in END-NOTE format.
Conflicts of Interest: The authors declare no conflict of interest.

Abbreviations
The following abbreviations are used in this manuscript:

ACC Adaptive Cruise Control


ADF Automated Driving Function
ADS Automated Driving Systems
AEB Autonomous Emergency Braking
AEBS Advanced Emergency Braking System
ALKS Automated Lane Keeping Systems
C-NCAP Chinese New Car Assessment Program
DAS Driving Automation System
DDM Data-Driven Model
DiL Driver-In-the-Loop
E/E Electrical and/or Electronic
EM Electro Magnetic
EOL End-Of-Line
Euro NCAP European New Car Assessment Program
FDTD Finite Difference Time Domain
FEM Finite Element Method
FOT Field Operational Test
Sensors 2022, 22, 5693 16 of 22

FOV Field of View


FPGA Field Programmable Gate Array
GT Ground Truth
HiFi High Fidelity
HiL Hardware-In-the-Loop
HMI Human-Machine Interface
IRM Ideal Radar Model
ISO International Organization for Standardization
JNCAP Japanese New Car Assessment Program
LCDAS Lane Change Decision Aid System
LDW Lane Departure Warning
LIDAR Light Detection and Ranging
LKAS Lane Keeping Assistance System
MBS Multi-Body Simulation
MBSE Model-Based Systems Engineering
MiL Model-In-the-Loop
ML Machine Learning
NCAP New Car Assessment Program
NHTSA National Highway Traffic Safety Administration
OEM Original Equipment Manufacturer
OIOO Object In Object Out
OTA Over-The-Air
PALS Partially Automated Lane Change System
PEC Perfect Conductors
PiL Processor-In-the-Loop
RADAR Radio Detection and Ranging
RCS Radar Cross Section
RIDO Rendering In Detection Out
RIOO Raw Input Object Output
RIRO Raw Input Raw output
RSI Raw Signal Interface
RTM Ray-Tracing based Model
SAE Society of Automotive Engineers
SC Scattering Centre
SiL Software-In-the-Loop
SOA State-Of-the-Art
SOTIF Safety Of The Intended Function
UN-ECE United Nations Economic Commission for Europe
V&V Validation and Verification
ViL Vehicle-In-the-Loop

References
1. Szalay, Z. Next Generation X-in-the-Loop Validation Methodology for Automated Vehicle Systems. IEEE Access 2021, 9,
35616–35632. [CrossRef]
2. Doms, T.; Rauch, B.; Schrammel, B.; Schwald, C.; Spahovic, E.; Schwarzl, C. Highly Automated Driving-The New Challenges for
Functional Safety and Cyber Security: White Paper;TÜV Austria Holding AG and Virtual Vehicle: Vienna, Austria, 2018.
3. ISO 26262; Part 1-12: Road Vehicles—Functional Safety. International Standard of Organisation: Geneva, Switzerland, 2011 .
4. ISO 15622:2018; Intelligent Transport Systems—Adaptive Cruise Control Systems—Performance Requirements and Test Proce-
dures. International Standard of Organisation: Geneva, Switzerland . 2019.
5. ISO 22733-1:2021; Road Vehicles—Test Method to Evaluate the Performance of Autonomous Emergency Braking Systems—Part
1: Car-to-Car. International Standard of Organisation: Geneva, Switzerland , 2021.
6. ISO 17361:2017; Intelligent Transport Systems—Lane Departure Warning Systems—Performance Requirements and Test
Procedures. International Standard of Organisation: Geneva, Switzerland , 2017.
7. ISO 17387:2008; Intelligent Transport Systems—Lane Change Decision Aid Systems (LCDAS)—Performance Requirements and
Test Procedures. International Standard of Organisation: Geneva, Switzerland , 2018.
8. ISO 21202:2020; Intelligent Transport Systems—Partially Automated Lane Change Systems (PALS)—Functional/Operational
Requirements and Test Procedures. International Standard of Organisation: Geneva, Switzerland , 2020.
Sensors 2022, 22, 5693 17 of 22

9. ISO 11270:2014; Intelligent Transport Systems—Lane Keeping Assistance Systems (LKAS)—Performance Requirements and Test
Procedures. International Standard of Organisation: Geneva, Switzerland , 2012.
10. United Regulation No 131 of the Economic Commission for Europe of the United Nations (UN/ECE)-Uniform Provisions Concerning the
Approval of Motor Vehicles with Regard to the Advanced Emergency Braking Systems (AEBS); Nations Economic Commission for
Europe: Geneva, Switzerland, 2014; pp. 47–62.
11. UN Regulation No. 157—Automated Lane Keeping Systems (ALKS); Nations Economic Commission for Europe: Geneva, Switzerland,
2021; pp. 75–137.
12. Hutchinson, A.C. (Ed.) Cryptocurrencies and the Regulatory Challenge; Routledge: London, UK, 2021. [CrossRef]
13. Wishart, J.; Como, S.; Forgione, U.; Weast, J.; Weston, L.; Smart, A.; Nicols, G.; S, R. Literature Review of Verification and
Validation Activities of Automated Driving Systems. SAE Int. J. Connect. Autom. Veh. 2020, 3, 267–323. [CrossRef]
14. Dahmen, U.; Roßmann, J. Simulation-based Verification with Experimentable Digital Twins in Virtual Testbeds. In Tagungsband
des 3. Kongresses Montage Handhabung Industrieroboter; Schüppstuhl, T., Tracht, K., Franke, J., Eds.; Springer: Berlin/Heidelberg,
Germany, 2018; pp. 139–147.
15. Schmidt, S.; Schlager, B.; Muckenhuber, S.; Stark, R. Configurable Sensor Model Architecture for the Development of Automated
Driving Systems. Sensors 2021, 21, 4687. [CrossRef]
16. ISO/PAS 21448:2019; Road Vehicles—Safety of the Intended Functionality. International Organization for Standardization:
Geneva, Switzerland, 2019.
17. Lipa, B.J.; Barrick, D.E. FMCW Signal Processing. 1990. Available online: http://www.codar.com/images/about/1990LipaBarr_
FMCW.pdf (accessed on 24 June 2022).
18. Muckenhuber, S.; Holzer, H.; Rubsam, J.; Stettinger, G. Object-based sensor model for virtual testing of ADAS/AD functions. In
Proceedings of the 8th IEEE International Conference on Connected Vehicles and Expo (ICCVE), Graz, Austria, 4–8 November
2019; IEEE: Piscataway, NJ, USA, 2019. [CrossRef]
19. Schlager, B.; Muckenhuber, S.; Schmidt, S.; Holzer, H.; Rott, R.; Maier, F.M.; Saad, K.; Kirchengast, M.; Stettinger, G.; Watzenig, D.;
et al. State-of-the-Art Sensor Models for Virtual Testing of Advanced Driver Assistance Systems/Autonomous Driving Functions.
SAE Int. J. Connect. Autom. Veh. 2020, 3, 233–261. [CrossRef]
20. Safety First for Automated Driving. Available online: https://wiki.unece.org/download/attachments/87622238/FRAV-01-11.
pdf?api=v2 (accessed on 24 June 2022).
21. Schnelle, S.C.; Salaani, M.K.; Rao, S.J.; Barickman, F.S.; Elsasser, D. Review of Simulation Frameworks and Standards Related to Driving
Scenarios; National Highway Traffic Safety Administration: Washington, DC, USA, 2019.
22. Rosenberger, P.; Wendler, J.T.; Holder, M.; Linnhoff, C.; Berghöfer, M.; Winner, H.; Maurer, M. Towards a generally accepted
validation methodology for sensor models-challenges, metrics, and first results. In Proceedings of the 2019 Graz Symposium
Virtual Vehicle, Graz, Austria, 7–8 May 2019; pp. 1–13.
23. Jun, Z.; Kai, Y.; Xuecai, D.; Zhangu, W.; Huainan, Z.; Chunguang, D. New modeling method of millimeter-wave radar considering
target radar echo intensity. Proc. Inst. Mech. Eng. Part D J. Automob. Eng. 2021, 235, 2857–2870. [CrossRef]
24. Guo, J.; Deng, W.; Zhang, S.; Qi, S.; Li, X.; Wang, C.; Wang, J. A Novel Method of Radar Modeling for Vehicle Intelligence. SAE
Int. J. Passeng.-Cars-Electron. Electr. Syst. 2017, 10, 50–56. [CrossRef]
25. Holder, M.; Rosenberger, P.; Bert, F.; Winner, H. Data-driven Derivation of Requirements for a Lidar Sensor Model. In Graz
Symposium Virtual Vehicle 2018; Universitäts-und Landesbibliothek, Darmstadt, Germany, 2018; pp. 1–10.
26. Bernsteiner, S.; Magosi, Z.; Lindvai-Soos, D.; Eichberger, A. Radar Sensor Model for the Virtual Development Process. Atzelektronik
Worldw. 2015, 10, 46–52. [CrossRef]
27. Hirsenkorn, N.; Hanke, T.; Rauch, A.; Dehlink, B.; Rasshofer, R.; Biebl, E. A non-parametric approach for modeling sensor
behavior. In Proceedings of the 2015 16th International Radar Symposium (IRS), Dresden, Germany, 24–26 June 2015; pp. 131–136.
[CrossRef]
28. Gschwandtner, M.; Kwitt, R.; Uhl, A.; Pree, W. BlenSor: Blender Sensor Simulation Toolbox. In Advances in Visual Computing,
Proceedings of the 7th International Symposium, ISVC 2011, Las Vegas, NV, USA, 26–28 September 2011; Bebis, G., Boyle, R., Parvin, B.,
Koracin, D., Wang, S., Kyungnam, K., Benes, B., Moreland, K., Borst, C., DiVerdi, S., et al., Eds.; Springer: Berlin/Heidelberg,
Germany, 2011; pp. 199–208. [CrossRef]
29. Wheeler, T.A.; Holder, M.; Winner, H.; Kochenderfer, M.J. Deep stochastic radar models. In Proceedings of the 2017 IEEE
Intelligent Vehicles Symposium (IV), Los Angeles, CA, USA, 11–14. June 2017; pp. 47–53. [CrossRef]
30. Jasinski, M. A Generic Validation Scheme for Real-Time Capable Automotive Radar Sensor Models integrated into an Autonomous
Driving Simulator. In Proceedings of the 2019 24th International Conference on Methods and Models in Automation and Robotics
(MMAR), Miedzyzdroje, Poland, 26–29 August 2019; IEEE: Piscataway, NJ, USA, 2019.
31. Ling, H.; Chou, R.C.; Lee, S.W. Shooting and bouncing rays: calculating the RCS of an arbitrarily shaped cavity. IEEE Trans.
Antennas Propag. 1989, 37, 194–205. [CrossRef]
32. Hirsenkorn, N.; Subkowski, P.; Hanke, T.; Schaermann, A.; Rauch, A.; Rasshofer, R.; Biebl, E. A ray launching approach for
modeling an FMCW radar system. In Proceedings of the 18th International Radar Symposium IRS 2017, Prague, Czech Republic,
28–30 June 2017; IEEE: Piscataway, NJ, USA, 2017.
Sensors 2022, 22, 5693 18 of 22

33. Martowicz, A.; Gallina, A.; Karpiel, G. Uncertainty propagation for vehicle detections in experimentally validated radar model
for automotive application. In Proceedings of the 2019 24th International Conference on Methods and Models in Automation and
Robotics (MMAR), Mi˛edzyzdroje, Poland, 26–29 August 2019; IEEE: Piscataway, NJ, USA, 2019.
34. Bhalla, R.; Ling, H. Three-dimensional scattering center extraction using the shooting and bouncing ray technique. IEEE Trans.
Antennas Propag. 1996, 44, 1445–1453. [CrossRef]
35. Schuler, K.; Becker, D.; Wiesbeck, W. Extraction of Virtual Scattering Centers of Vehicles by Ray-Tracing Simulations. IEEE Trans.
Antennas Propag. 2008, 56, 3543–3551. [CrossRef]
36. Hirsenkorn, N.; Hanke, T.; Rauch, A.; Dehlink, B.; Rasshofer, R.; Biebl, E. Virtual sensor models for real-time applications. Adv.
Radio Sci. 2016, 14, 31–37. [CrossRef]
37. Scheel, A.; Dietmayer, K. Tracking Multiple Vehicles Using a Variational Radar Model. IEEE Trans. Intell. Transp. Syst. 2019,
20, 3721–3736. [CrossRef]
38. Hanke, T.; Hirsenkorn, N.; Dehlink, B.; Rauch, A.; Rasshofer, R.; Biebl, E. Generic architecture for simulation of ADAS sensors.
In Proceedings of the Generic Architecture for Simulation of ADAS Sensors, Dresden, Germany, 24–26 June 2015; pp. 125–130.
[CrossRef]
39. Cao, P. Modeling Active Perception Sensors for Real-Time Virtual Validation of Automated Driving Systems. Ph.D. Thesis,
Technische Universität Darmstadt, Darmstadt, Germany, 2017.
40. Holder, M.; Linnhoff, C.; Rosenberger, P.; Winner, H. The Fourier Tracing Approach for Modeling Automotive Radar Sensors. In
Proceedings of the 20th International Radar Symposium (IRS), Ulm, Germany, 26–28 June 2019; IEEE: Piscataway, NJ, USA, 2019.
41. Cao, P.; Wachenfeld, W.; Winner, H. Perception sensor modeling for virtual validation of automated driving. Inf. Technol. 2015,
57, 243–251. [CrossRef]
42. Ngo, A.; Bauer, M.P.; Resch, M. A Sensitivity Analysis Approach for Evaluating a Radar Simulation for Virtual Testing of
Autonomous Driving Functions. In Proceedings of the 2020 5th Asia-Pacific Conference on Intelligent Robot Systems (ACIRS),
Singapore, 17–19 July 2020; IEEE: Piscataway, NJ, USA, 2020. [CrossRef]
43. Ngo, A.; Bauer, M.P.; Resch, M. A Multi-Layered Approach for Measuring the Simulation-to-Reality Gap of Radar Perception
for Autonomous Driving. In Proceedings of the 2021 IEEE International Intelligent Transportation Systems Conference (ITSC),
Indianapolis, IN, USA, 19–22 September 2021; pp. 4008–4014.
44. Holder, M.F. Synthetic Generation of Radar Sensor Data for Virtual Validation of Autonomous Driving. Ph.D. Thesis, TU
Darmstadt, Darmstadt, Germany, 2021. [CrossRef]
45. Dallmann, T.; Mende, J.K.; Wald, S. A Radar Target Simulator for Complex Traffic Scenarios ATRIUM. In Proceedings of the 2018
IEEE MTT-S International Conference on Microwaves for Intelligent Mobility (ICMIM), Munich, Germany, 16–17 April 2018;
IEEE: Piscataway, NJ, USA, 2018; pp. 1–4. [CrossRef]
46. Gowdu, S.B.J.; Asghar, M.E.; Stephan, R.; Hein, M.A.; Nagel, J.; Baumgartner, F. System architecture for installed-performance
testing of automotive radars over-the-air. In Proceedings of the 2018 IEEE MTT-S International Conference on Microwaves for
Intelligent Mobility (ICMIM), Munich, Germany, 15–17 April 2018; IEEE: Piscataway, NJ, USA, 2018; pp. 1–4. [CrossRef]
47. Zhu, B.; Sun, Y.; Zhao, J.; Zhang, S.; Zhang, P.; Song, D. Millimeter-Wave Radar in-the-Loop Testing for Intelligent Vehicles. IEEE
Trans. Intell. Transp. Syst. 2021, 1–11. [CrossRef]
48. Sobotka, J.; Novak, J. Digital Vehicle Radar Sensor Target Simulation. In Proceedings of the 2020 IEEE International Instrumenta-
tion and Measurement Technology Conference (I2MTC), Dubrovnik, Croatia, 25–29 May 2020; IEEE: Piscataway, NJ, USA, 2020.
[CrossRef]
49. Diewald, A.; Kurz, C.; Kannan, P.V.; Gießler, M.; Pauli, M.; Göttel, B.; Kayser, T.; Gauterin, F.; Zwick, T. Radar Target Simulation
for Vehicle-in-the-Loop Testing. Vehicles 2021, 3, 257–271. [CrossRef]
50. Gadringer, M.E.; Schreiber, H.; Gruber, A.; Vorderderfler, M.; Amschl, D.; Bösch, W.; Metzner, S.; Pflügl, H.; Paulweber, M. Virtual
reality for automotive radars. Elektrotechnik Und Informationstechnik 2018, 135, 335–343. [CrossRef]
51. Cheng, J.J.; Xu, Z.G.; Gao, Y.; Wang, Z.; Wang, W.W.; Zhao, X.M. An Indoor Rapid Testing Platform for Autonomous Vehicles
Using Vehicle-in-the-Loop Simulation. In Proceedings of the CICTP 2020, Xi’an, China, 16–19 December 2020; Wei, H., Wang, H.,
Zhang, L., An, Y., Zhao, X., Eds.; American Society of Civil Engineers: Reston, VA, USA, 2020; pp. 411–422. [CrossRef]
52. Asghar, M.E.; Buddappagari, S.; Baumgartner, F.; Graf, S.; Kreutz, F.; Loffler, A.; Nagel, J.; Reichmann, T.; Stephan, R.; Hein,
M.A. Radar Target Simulator and Antenna Positioner for Real-Time Over-the-air Stimulation of Automotive Radar Systems. In
Proceedings of the 2020 17th European Radar Conference (EuRAD), Jaarbeurs Utrecht, Netherlands, 16–18 September 2020; IEEE:
Piscataway, NJ, USA,, Jaarbeurs Utrecht, Netherlands, 16-18 Sep. 2020; pp. 95–98. [CrossRef]
53. Lutz, S.; Erhart, C.; Walte, T.; Weigel, R. Target simulator concept for chirp modulated 77 GHz automotive radar sensors. In
Proceedings of the 2014 11th European Radar Conference, Rome, Italy, 8–10 October 2014; pp. 65–68. [CrossRef]
54. Maier, F.M.; Makkapati, V.P.; Horn, M. Environment perception simulation for radar stimulation in automated driving function
testing. Elektrotechnik Und Informationstechnik 2018, 135, 309–315. [CrossRef]
55. Herrmann, M.; Schön, H. Efficient Sensor Development Using Raw Signal Interfaces. In Fahrerassistenzsysteme 2018; Bertram, T.,
Ed.; Springer: Wiesbaden, Germany, 2019; pp. 30–39. [CrossRef]
56. Schoener, H.P. Automotive Needs and Expectations towards Next Generation Driving Simulation. In Proceedings of the 2018
Driving Simulator Conference, Antibes, France, 5–7 September 2018; pp. 11–16.
Sensors 2022, 22, 5693 19 of 22

57. Hartstern, M.; Rack, V.; Kaboli, M.; Stork, W. Simulation-based Evaluation of Automotive Sensor Setups for Environmental
Perception in Early Development Stages. In Proceedings of the 2020 IEEE Intelligent Vehicles Symposium (IV), Las Vegas, NV,
USA, 19 October–13 November 2020; IEEE: Piscataway, NJ, USA, 2020; pp. 858–864. [CrossRef]
58. Gregory, Y. (GRVA) New Assessment/Test Method for Automated Driving (NATM)-Master Document. 2021. Available online:
https://unece.org/sites/default/files/2021-04/ECE-TRANS-WP29-2021-61e.pdf (accessed on 24 June 2022).
59. Menzel, T.; Bagschik, G.; Maurer, M. Scenarios for Development, Test and Validation of Automated Vehicles. In Proceedings of
the 2018 IEEE Intelligent Vehicles Symposium (IV), Changshu, China, 26–30 June 2018 ; IEEE: Piscataway, NJ, USA; pp. 1821–1827.
60. Bringmann, E.; Kr, A. Model-Based Testing of Automotive Systems. In Proceedings of the 2008 International Conference
on Software Testing, Verification, and Validation, Lillehammer, Norway, 9–11 April 2008; IEEE: Piscataway, NJ, USA, 2008;
pp. 485–493. [CrossRef]
61. Schutt, B.; Steimle, M.; Kramer, B.; Behnecke, D.; Sax, E. A Taxonomy for Quality in Simulation-Based Development and Testing
of Automated Driving Systems. IEEE Access 2022, 10, 18631–18644. [CrossRef]
62. Linnhoff, C.; Rosenberger, P.; Holder, M.F.; Cianciaruso, N.; Winner, H. Highly Parameterizable and Generic Perception Sensor
Model Architecture. In Automatisiertes Fahren 2020; Bertram, T., Ed.; Springer: Wiesbaden, Germany, 2021; pp. 195–206. [CrossRef]
63. Rosique, F.; Navarro, P.J.; Fernández, C.; Padilla, A. A Systematic Review of Perception System and Simulators for Autonomous
Vehicles Research. Sensors 2019, 19, 648. [CrossRef]
64. Muckenhuber, S.; Museljic, E.; Stettinger, G. Performance evaluation of a state-of-the-art automotive radar and corresponding
modeling approaches based on a large labeled dataset. J. Intell. Transp. Syst. 2021, 1–20. [CrossRef]
65. Choi, W.Y.; Yang, J.H.; Chung, C.C. Data-Driven Object Vehicle Estimation by Radar Accuracy Modeling with Weighted
Interpolation. Sensors 2021, 21, 2317. [CrossRef] [PubMed]
66. Bel Kamel, E.; Peden, A.; Pajusco, P. RCS modeling and measurements for automotive radar applications in the W band. In
Proceedings of the 2017 11th European Conference on Antennas and Propagation (EUCAP), Paris, France, 19–24 March 2017;
IEEE: Piscataway, NJ, USA, 2017, pp. 2445–2449. [CrossRef]
67. Thieling, J.; Frese, S.; RoBmann, J. Scalable and Physical Radar Sensor Simulation for Interacting Digital Twins. IEEE Sens. J. 2021,
21, 3184–3192. [CrossRef]
68. Schuesler, C.; Hoffmann, M.; Braunig, J.; Ullmann, I.; Ebelt, R.; Vossiek, M. A Realistic Radar Ray Tracing Simulator for Large
MIMO-Arrays in Automotive Environments. IEEE J. Microwaves 2021, 1, 962–974. [CrossRef]
69. Dudek, M.; Wahl, R.; Kissinger, D.; Weigel, R.; Fischer, G. Millimeter wave FMCW radar system simulations including a 3D ray
tracing channel simulator. In Proceedings of the 2010 Asia-Pacific Microwave Conference, Yokohama, Japan, 7–10 December
2010; pp. 1665–1668.
70. Zhu, L.; He, D.; Ai, B.; Zhong, Z.; Zhu, F.; Wang, Z. Measurement and Ray-Tracing Simulation for Millimeter-Wave Automotive
Radar. In Proceedings of the 2021 IEEE 4th International Conference on Electronic Information and Communication Technology
(ICEICT), Xi’an, China, 18–20 August 2021; IEEE: Piscataway, NJ, USA, 2021; pp. 582–587. [CrossRef]
71. Li, X.; Deng, W.; Zhang, S.; Li, Y.; Song, S.; Wang, S.; Wang, G. Research on Millimeter Wave Radar Simulation Model for
Intelligent Vehicle. Int. J. Automot. Technol. 2020, 21, 275–284. [CrossRef]
72. Roth, E.; Dirndorfer, T.J.; Knoll, A.; v. Neumann-Cosel, K.; Ganslmeier, T.; Kern, A.; Fischer, M.O. Analysis and validation of
perception sensor models in an integrated vehicle and environment simulation. In Proceedings of the 22nd Conference on
Enhanced Saferty of Vehicles (ESV 2011), Washington, DC, USA, 13–16 June 2011.
73. Stolz, M.; Nestlinger, G. Fast generic sensor models for testing highly automated vehicles in simulation. Elektrotechnik Und
Informationstechnik 2018, 135, 365–369. [CrossRef]
74. Elgharbawy, M.; Schwarzhaupt, A.; Scheike, G.; Frey, M.; Gauterin, F. A generic architecture of ADAS sensor fault injection
for virtual tests. In Proceedings of the 2016 IEEE/ACS 13th International Conference of Computer Systems and Applications
(AICCSA), Agadir, Morocco, 29 November–2 December 2016; IEEE: Piscataway, NJ, USA, 2016; pp. 1–7. [CrossRef]
75. Velazquez, J.M.R.; Mailly, F.; Nouet, P. A generic model for sensor simulation at system level. In Proceedings of the 2018
Symposium on Design, Test, Integration & Packaging of MEMS and MOEMS (DTIP), Rome, Italy, 22–25 May 2018; IEEE:
Piscataway, NJ, USA, 2018; pp. 1–4. [CrossRef]
76. Hanke, T. Virtual Sensorics: Simulated Environmental Perception for Automated Driving Systems. Ph.D. Thesis, TU Darmstadt,
Darmstadt, Germany, 2022.
77. Deng, W.; Dai, J.; Zhao, Q.; Litkouhi, B.; Moshchuk, N.; Nisonger, R. Modeling of range sensing and object detection, for vehicle
active safety. In Proceedings of the 2009 12th International IEEE Conference on Intelligent Transportation Systems, St. Louis, MO,
USA, 4–7 October 2009; IEEE: Piscataway, NJ, USA, 2009; pp. 1–6. [CrossRef]
78. Holder, M.; Rosenberger, P.; Winner, H.; Dhondt, T.; Makkapati, V.P.; Maier, M.; Schreiber, H.; Magosi, Z.; Slavik, Z.; Bringmann,
O.; et al. Measurements revealing Challenges in Radar Sensor Modeling for Virtual Validation of Autonomous Driving.
In Proceedings of the 2018 IEEE Intelligent Transportation Systems Conference, Maui, HI, USA, 4–7 November 2018; IEEE:
Piscataway, NJ, USA, 2018; pp. 2616–2622. [CrossRef]
79. Slavik, Z.; Mishra, K.V. Phenomenological Modeling of Millimeter-Wave Automotive Radar. In Proceedings of the 2019 URSI
Asia-Pacific Radio Science Conference (AP-RASC), New Delhi, India, 9–15 March 2019; IEEE: Piscataway, NJ, USA, 2019; pp. 1–4.
[CrossRef]
Sensors 2022, 22, 5693 20 of 22

80. Arnelid, H.; Zec, E.L.; Mohammadiha, N. In Proceedings of the Recurrent Conditional Generative Adversarial Networks for
Autonomous Driving Sensor Modelling, Auckland, New Zealand, 27–30 October 2019; IEEE: Piscataway, NJ, USA, 2019.
81. Zec, E.L.; Mohammadiha, N.; Schliep, A. In Proceedings of the Statistical Sensor Modelling for Autonomous Driving Using
Autoregressive Input-Output HMMs, Maui, Hawaii, 4–7 November 2018; IEEE: Piscataway, NJ, USA, 2018.
82. Song, Y.; Wang, Y.; Li, Y. Radar data simulation using deep generative networks. J. Eng. 2019, 2019, 6699–6702. [CrossRef]
83. Suhre, A.; Malik, W. (Eds.) Simulating Object Lists Using Neural Networks in Automotive Radar . In Proceedings of the 2018
19th International Conference on Thermal, Mechanical and Multi-Physics Simulation and Experiments in Microelectronics and
Microsystems (EuroSimE), Toulouse, France, 15–18 April 2018. [CrossRef]
84. Cao, X.; Lan, J.; Li, X.R.; Liu, Y. Automotive Radar-Based Vehicle Tracking Using Data-Region Association. IEEE Trans. Intell.
Transp. Syst. 2021, 23, 1–14. [CrossRef]
85. Mishra, K.V.; Shankar M.R., B.; Ottersten, B. Stochastic-Geometry-Based Interference Modeling in Automotive Radars Using
Matérn Hard-Core Process. In Proceedings of the 2020 IEEE Radar Conference (RadarConf20), lorence, Italy, 21–25 September
2020; IEEE: Piscataway, NJ, USA, 2020; pp. 1–5. [CrossRef]
86. Choi, W.Y.; Kang, C.M.; Lee, S.H.; Chung, C.C. Radar Accuracy Modeling and Its Application to Object Vehicle Tracking. Int. J.
Control. Autom. Syst. 2020, 18, 3146–3158. [CrossRef]
87. Belyaev, A.A.; Frolov, I.O.; Suanov, T.A.; Trots, D.O. In Proceedings of the Object Detection in an Urban Environment Using
77GHz Radar: Divnomorskoe, Krasnodar Region, Russia, 24–28 June 2019; IEEE: Piscataway, NJ, USA, 2019.
88. Buddendick, H.; Eibert, T.F. Radio channel simulations using multiple scattering center models. In Proceedings of the Radio
Channel Simulations Using Multiple Scattering Center Models, North Charleston, SC, USA, 1–5 June 2009; Buddendick, H.,
Eibert, T.F.., Eds.; IEEE: Piscataway, NJ, USA, 2009; pp. 1–4. [CrossRef]
89. Schubert, R.; Mattern, N.; Bours, R. Simulation of Sensor Models for the Evaluation of Advanced Driver Assistance Systems.
Atzelektronik Worldw. 2014, 9, 26–29. [CrossRef]
90. Sligar, A.P. Machine Learning-Based Radar Perception for Autonomous Vehicles Using Full Physics Simulation. IEEE Access
2020, 8, 51470–51476. [CrossRef]
91. Buhren, M.; Yang, B. Initialization Procedure for Radar Target Tracking without Object Movement Constraints. In Proceeding s of
the 2007 7th International Conference on ITS Telecommunications, Sophia Antipolis, France, 6–8 June 2007; IEEE: Piscataway, NJ,
USA, 2007, pp. 1–6.
92. Ponn, T.; Lanz, T.; Diermeyer, F. Automatic Generation of Road Geometries to Create Challenging Scenarios for Automated
Vehicles Based on the Sensor Setup. In Proceedings of the 2020 IEEE Intelligent Vehicles Symposium (IV), Las Vegas, NV, USA,
23–26 June 2020; IEEE: Piscataway, NJ, USA, 2020; pp. 694–700. [CrossRef]
93. Haider, A.; Kim, J.; Sachse, M.; Zeh, T.; Schneider, S.A.; Thumann, M.; Eryildirim, A.; Ebeling, T. Automotive Radar Sensor
Behavioral Models for Closed Loop Simulations. In Proceedings of the 5th International Symposium on Future Active Safety
Technology toward Zero Accidents (FAST-zero’19), Blacksburg, VA, USA, 9–11 September 2019.
94. Dietmayer, K. Predicting of Machine Perception for Automated Driving. In Autonomous Driving; Maurer, M., Gerdes, J.C., Lenz,
B., Winner, H., Eds.; Springer: Berlin/Heidelberg, Germany, 2016; pp. 407–424. [CrossRef]
95. Doerr, D. Using Virtualization to Accelerate the Development of ADAS & Automated Driving Functions: IPG Presentation; IPG
Automotive GmbH: Karlsruhe, Germany, 2017.
96. Gubelli, D.; Krasnov, O.; Yarovyi, O. Ray-tracing simulator for radar signals propagation in radar networks. Eur. Radar Conf.
2013, 2013, 73–76.
97. Schneider, R. Modellierung der Wellenausbreitung für ein Bildgebendes Kfz-Radar. Ph.D. Thesis, Universität Karlsruhe,
Karlsruhe, Germany, 1998.
98. Stellet, J.E.; Schumacher, J.; Lange, O.; Branz, W.; Niewels, F.; Zöllner, J.M. Statistical Modelling of Object Detection in Stereo
Vision-Based Driver Assistance. In Intelligent Autonomous Systems 13; Menegatti, E., Michael, N., Berns, K., Yamaguchi, H., Eds.;
Advances in Intelligent Systems and Computing; Springer International Publishing: Cham, Switzerland, 2016; Volume 302,
pp. 749–761. [CrossRef]
99. Deng, W.; Zeng, S.; Zhao, Q.; Dai, J. Modeling and simulation of sensor-guided autonomous driving. Int. J. Veh. Des. 2011,
56, 341. [CrossRef]
100. Gadringer, M.E.; Maier, F.M.; Schreiber, H.; Makkapati, V.P.; Gruber, A.; Vorderderfler, M.; Amschl, D.; Metzner, S.; Pflügl, H.;
Bösch, W.; et al. Radar target stimulation for automotive applications. IET Radar Sonar Navig. 2018, 12, 1096–1103. [CrossRef]
101. van Driesten, C.; Schaller, T. Overall Approach to Standardize AD Sensor Interfaces: Simulation and Real Vehicle. In Fahrerassis-
tenzsysteme 2018; Bertram, T., Ed.; Springer: Wiesbaden, Germany, 2019; pp. 47–55. [CrossRef]
102. Feilhauer, M.; Häring, J. A real-time capable multi-sensor model to validate ADAS in a virtual environment. In Fahrerassistenzsys-
teme 2017; Isermann, R., Ed.; Springer: Wiesbaden, Germany, 2017; pp. 227–256. [CrossRef]
103. Maier, F.M. Radar Perception Simulation for Automated Driving Tests. Phd thesis, Graz University of Technology, Graz, 2022.
104. Buhren, M.; Yang, B. Extension of Automotive Radar Target List Simulation to consider further Physical Aspects. In Proceedings
of the 2007 7th International Conference on ITS Telecommunications, Sophia Antipolis, France, 6–8 June 2007; pp. 1–6. [CrossRef]
105. Prinz, A.; Peters, L.T.; Schwendner, J.; Ayeb, M.; Brabetz, L. Automotive Radar Signal and Interference Simulation for Testing
Autonomous Driving. In Intelligent Transport Systems, from Research and Development to the Market Uptake; Martins, A.L., Ferreira,
J.C., Kocian, A., Costa, V., Eds.; Springer: Cham, Switzerland, 2021; Volume 364, pp. 223–240. [CrossRef]
Sensors 2022, 22, 5693 21 of 22

106. Maier, M.; Makkapati, V.P.; Horn, M. Adapting Phong into a Simulation for Stimulation of Automotive Radar Sensors. In
Proceeding s of the 2018 IEEE MTT-S International Conference on Microwaves for Intelligent Mobility (ICMIM), Munich, Germany,
15–17 April 2018; IEEE: Piscataway, NJ, USA, 2018; pp. 1–4. [CrossRef]
107. Bogdan, T.; Dorina, I. Simulation of Automotive MIMO Radar. In Proceedings of the 2020 International Symposium on Electronics
and Telecommunications (ISETC), Timisoara, Romania, 5–6 November 2020; IEEE: Piscataway, NJ, USA, 2020; pp. 1–4. [CrossRef]
108. Blasch, E.; Hensel, M. Fusion of Distributions for Radar Clutter Modeling; Defense Technical Information Center: Fort Belvoir, VA,
USA, 2004.
109. Etinger, A.; Litvak, B.; Pinhasi, Y. Multi Ray Model for Near-Ground Millimeter Wave Radar. Sensors 2017, 17, 1983. [CrossRef]
[PubMed]
110. Hammarstrand, L.; Lundgren, M.; Svensson, L. Adaptive Radar Sensor Model for Tracking Structured Extended Objects. IEEE
Trans. Aerosp. Electron. Syst. 2012, 48, 1975–1995. [CrossRef]
111. Hammarstrand, L.; Svensson, L.; Sandblom, F.; Sorstedt, J. Extended Object Tracking using a Radar Resolution Model. IEEE
Trans. Aerosp. Electron. Syst. 2012, 48, 2371–2386. [CrossRef]
112. Hanke, T.; Schaermann, A.; Matthias, G.; Konstantin, W.; Hirsenkorn, N.; Rauch, A.; Stefan-Alexander Schneider. Generation and
validation of virtual point cloud data for automated driving systems. In Proceedings of the 20th International Conference on
Intelligent Transportation Systems: Mielparque Yokohama in Yokohama, Kanagawa, Japan, 16–19 October 2017; IEEE: Piscataway,
NJ, USA, 2017.
113. Peinecke, N.; Döhler, H.U.; Korn, B.R. Simulation of imaging radar using graphics hardware acceleration. In Proceedings of the
Enhanced and Synthetic Vision 2008, Orlando, FL, USA, 20 May 2008; Güell, J.J.; Uijt de Haag, M., Eds.; SPIE: Bellingham, WA,
USA , 2008; p. 69570L. [CrossRef]
114. Buhren, M.; Yang, B. Simulation of Automotive Radar Target Lists using a Novel Approach of Object Representation. In
Proceedings of the 2006 IEEE Intelligent Vehicles Symposium, Meguro-Ku, Japan, 13–15 June 2006; IEEE: Piscataway, NJ, USA,
2006; pp. 314–319. [CrossRef]
115. Bühren, M.; Yang, B. Simulation of automotive radar target lists considering clutter and limited resolution. In Proceedings of the
International Radar Symposium, Cologne, Germany, 5–7 September 2007; pp. 195–200.
116. Danielsson, L. Tracking and Radar Sensor Modelling for Automotive Safety Systems 2010; Doktorsavhandlingar vid Chalmers
Tekniska Högskola;Chalmers Univercity of Technology: Göteborg, Germany, 2010; Volume 3064.
117. Schuler, K. Intelligente Antennensysteme für Kraftfahrzeug-Nahbereichs-Radar-sensorik. Ph.D. Thesis, Universität Karlsruhe,
Zugl.: Karlsruhe, Germany, 2007.
118. Hermann Buddendick.; Thomas Eibert.; Jurgen Hasch. In Proceedings of the Bistatic Scattering Center Models_for_the_simulation_
of_wave_propagation_in_automotive_radar_systems: (GeMiC 2010), Berlin, Germany, 15–17 March 2010; IEEE: Piscataway, NJ,
USA, 2010.
119. Harrington, R.F. Field Computation by Moment Methods; Wiley-IEEE Press: Hoboken, NJ, USA, 1993.
120. Weinmann, F. Ray Tracing With PO/PTD for RCS Modeling of Large Complex Objects. IEEE Trans. Antennas Propag. 2006,
54, 1797–1806. [CrossRef]
121. Clemens.; M..; Weiland, T. Discrete Electromagnetism with finite integration technique. Prog. Electromagn. Res. 2001, 2001, 65–87.
[CrossRef]
122. Machida, T.; Owaki, T. In Proceedings of the Rapid and Precise Millimeter-wave Radar Simulation for ADAS Virtual Assessment:
Auckland, New Zealand, 27–30 October 2019; IEEE: Auckland, New Zealand, 2019.
123. Yee, K. Numerical solution of initial boundary value problems involving maxwell’s equations in isotropic media. IEEE Trans.
Antennas Propag. 1966, 14, 302–307. [CrossRef]
124. Chipengo, U.; Krenz, P.M.; Carpenter, S. From Antenna Design to High Fidelity, Full Physics Automotive Radar Sensor Corner
Case Simulation. Model. Simul. Eng. 2018, 2018, 4239725. [CrossRef]
125. Owaki, T.; Machida, T. In Proceedings of the Hybrid Physics-Based and Data-Driven Approach to Estimate the Radar Cross-
Section of Vehicles: Auckland, New Zealand, 27–30 October 2019; IEEE: Piscataway, NJ, USA, 2019.
126. Schäferle, S. How Do I Choose the Correct Sensor Model for My Specific Application? CarMaker Tips & Tricks No. 6-004: Choosing the
Correct Sensor Model for Your Application; IPG Automotive GmbH, Karlsruhe, Germany, 2019.
127. Eder, T.; Hachicha, R.; Sellami, H.; van Driesten, C.; Biebl, E. Data Driven Radar Detection Models: A Comparison of Artificial
Neural Networks and Non Parametric Density Estimators on Synthetically Generated Radar Data. In Proceedings of the 2019
Kleinheubach Conference, Miltenberg, Germany, 23–25 September 2019.
128. Koch, W. Verfeinerte Sensormodelle für Tracking-Anwendungen (Advanced Sensor Models for Tracking Applications). Tech.
Mess. 2007, 74, 112–120. [CrossRef]
129. Berthold, P.; Michaelis, M.; Luettel, T.; Meissner, D.; Wuensche, H.J.J. In Proceedings of the An Abstracted Radar Measurement
Model for Extended Object Tracking, Maui, HI, USA, 4–7 November 2018; IEEE: Piscataway, NJ, USA, 2018.
130. Felbecker, R.; Raschkowski, L.; Keusgen, W.; Peter, M. Electromagnetic wave propagation in the millimeter wave band using the
NVIDIA OptiX GPU ray tracing engine. In Proceeding s of the 2012 6th European Conference on Antennas and Propagation
(EUCAP), Prague, Czech Republic, 26–30 March 2012; IEEE: Piscataway, NJ, USA, 2012; pp. 488–492. [CrossRef]
131. Yun, Z.; Iskander, M.F. Ray Tracing for Radio Propagation Modeling: Principles and Applications. IEEE Access 2015, 3, 1089–1100.
[CrossRef]
Sensors 2022, 22, 5693 22 of 22

132. Topak, A.E.; Hash, J.; Zwick, T. A system simulation of a 77 GHz phased array radar sensor. In Proceedings of the 12th
International Radar Symposium (IRS) 2011, Leipzig, Germany, 7–9 September 2011; pp. 175–180.
133. Eder, T. Simulation of Automotive Radar Point Clouds in Standardized Frameworks; Cuvillier Verlag: Göttingen, Germany, 2021.
134. Hirsenkorn, N.Q. Modellbildung und Simulation der Fahrzeugumfeldsensorik. Ph.D. Thesis, Technische Universität München,
München, Germany, 2018.
135. Degen, R.; Ott, H.; Overath, F.; Schyr, C.; Leijon, M.; Ruschitzka, M. Methodical Approach to the Development of a Radar
Sensor Model for the Detection of Urban Traffic Participants Using a Virtual Reality Engine. J. Transp. Technol. 2021, 11, 179–195.
[CrossRef]
136. Kedzia, J.C.; de Souza, P.; Gruyer, D. Advanced RADAR sensors modeling for driving assistance systems testing. In Proceedings
of the 2016 10th European Conference on Antennas and Propagation (EuCAP); IEEE: Piscataway, NJ, USA, 2016; pp. 1–2.
[CrossRef]

You might also like