A Software-in-the-Loop Simulation of an Intelligent

MicroSatellite within a Virtual Environment

Kaveh Hassani and Won-Sook Lee

School of Electrical Engineering and Computer Science
University of Ottawa
Ottawa, Canada
[email protected] and [email protected]

Abstract— Rapid growth in space missions necessitates the automatically detect and react to events such as natural
onboard intelligence, which creates autonomous space systems by phenomenon. High crew costs, mission inflexibility, lack of
providing high level decision making, robust execution of onboard problem solving ability, and limitations in mission
decisions, and automatic fault repairing. Mostly, autonomous definition are other drawbacks of conventional space systems.
space systems are implemented as hybrid architectures with a
few conceptual layers. Validating the stability and evaluating the Autonomous space technology emerged to alleviate the
performance of an autonomous architecture is critical for space flaws of classic space systems. Polle has classified the
missions. Software−in−the−loop simulation is a suitable approach motivations for autonomy in space technology to: cost
for addressing this demand. However, the data acquired from reduction, improvement of the quality and quantity of mission
simulation is represented as alphanumeric values or diagrams, products, availability, performance improvement, and new
which needs to be interpreted. In this paper, we propose an mission feasibility [2]. Dehgan, Cheheltani, and Kiapasha have
intelligent architecture to provide onboard autonomy for an mentioned the advantages of autonomous space systems as
observation micro-satellite. The architecture integrates the low omitting time consuming activities by human, reliability
level physical actions with conceptual decision making ability in a enhancement by continuous system health monitoring,
hierarchical manner. To evaluate the proposed architecture, we protecting system against its internal issues, and high ratio of
have implemented a distributed software−in−the−loop simulation useful data [3]. Onboard autonomy refers to the ability of
to simulate the space, satellite, ground stations, and intelligent
performing a set of onboard activities to reach the mission
onboard software. Moreover, for the first time, we have used
goals in optimal configuration. In this paradigm, instead of low
virtual reality to visualize the satellite’s autonomous behavior in
the orbit. It lets the users have a high level feedback from
level instructions, high level missions are transmitted to the
integrated simulation. Scenario-based evaluations have shown satellite. The intelligent onboard software plans, schedules, and
the stability and efficiency of the proposed architecture. executes the activities to achieve the mission goals with
optimal resource allocation. Also, it monitors and reacts to the
Keywords—Hybrid intelligent system; onboard autonomy; internal and external events. In case of none-fatal failures, it
software-in-the-loop simulation; satellite visualization detects, isolates, and repairs the error. In summary, onboard
autonomy pursues two main objectives: achieving high level
I. INTRODUCTION goals and reacting in real-time [1].
Since the beginning of space age in 1957, space technology Validating the stability and evaluating the performance of
has been progressing rapidly. Nowadays, satellite constellation an autonomous architecture is critical for space missions.
systems such as GPS are serving individuals in daily life; Software−in−the−loop (SIL) simulation is a suitable approach
interplanetary missions such as curiosity mars rover are for addressing this demand. In SIL simulation, part of the
transmitting rich information about other planets; and model exists in standard simulation tool such as Simulink and
observation satellites are providing high resolution images for the rest is compiled code [4]. However, the data acquired from
scientists. Space is an extreme environment which demands simulation are alphanumeric values or diagrams, and need to be
precise, efficient and robust technology [1]. Early space interpreted. A suitable approach for evaluating autonomous
missions were proceeding under the ground team’s direct space systems is integrating virtual reality with SIL simulation
supervision. A large crew was responsible for preparing low to visualize the satellite behavior under the control of onboard
level instructions to satisfy a few mission goals. These intelligent software.
instructions were uplinked to the satellite when it was passing In this paper, we propose an intelligent architecture to
over the ground station. However, due to the limited bandwidth provide onboard autonomy for an observation micro-satellite.
and the low frequency of over passes, the satellite was mostly The architecture integrates the low level physical actions with
idle. Moreover, the satellite was not able to handle the conceptual decision making ability in a hierarchical manner.
unpredicted situations and had to be recovered manually by To evaluate the proposed architecture, we have implemented a
ground team. Another disadvantage was lack of ability to distributed SIL to simulate the space, satellite, ground stations,
and intelligent onboard software. Moreover, for the first time, have proposed an integrated system for design and simulation
we have used virtual reality to visualize the satellite’s of a small satellite [20]. Miao, Chen, and Sun have proposed a
autonomous behavior in orbit. It lets the users have a high level distributed platform for satellite simulation [21]. A few
feedback from integrated simulation. research works have employed virtual environments for
satellite simulation. A satellite simulator based on virtual
The paper is organized as follows: in section 2 an overview reality is proposed in [22]. It visualizes the satellite orbiting the
of related works is presented; while Section 3 presents the earth in 3D for training purposes. Hao, Pei, Yongkang, and
flight scenario for the observation microsatellite. In section 4, Chao have integrated the 3D visualization and SIL for satellite
the proposed intelligent architecture is described. In section 5, simulation [23]. Another research work has been carried out to
the integrated SIL simulation and visualization of the system is visualize the earth observing satellite within a virtual
discussed. Finally, section 6 concludes the paper. environment to facilitate the data interpretations [24]. In this
paper, for the first time, we integrate virtual reality with SIL
II. RELATED WORK simulation to visualize the satellite behavior under the control
Autonomous architectures are originated from robotics. of onboard intelligent software.
Early intelligent architectures were mostly based on sense-
decide-act cycle [5] in which data is acquired from sensors and III. FLIGHT SCENARIO
integrated with pre-defined goals to generate and execute a The hypothetical mission is a remote sensing mission in
plan. Shakey was the first robot exploited this paradigm [6]. which an observation microsatellite provides high resolution
Although this paradigm is able to provide real-time responses, images from natural phenomena such as volcanoes for
it cannot provide deliberative functionality. Stroupe et al. scientific analyses. The main components of the mission
classified the autonomous architectures to three types: include an observation microsatellite with onboard autonomy;
behavioral, hierarchical, and hybrid [1]. Behavioral two unidirectional receiving ground stations; a Tracking,
architectures are bottom-up paradigms that model the system telemetry and command (TTC) station; and a mission control
components as behaviors. This type is similar to sense-decide- center (MCC). The mission components are shown in Fig. 1.
act-based architectures and inherits the same problem. Brook’s
subsumption architecture is an example of behavioral
architectures [7]. Hierarchical architectures are top-down A. Microsatellite specification
approaches that decompose high level goals to low level sub- The hypothetical satellite is a small scale observation
goals. Although they can handle the complex tasks, they are satellite circling in low earth orbit (LEO). It is assumed that the
not efficient for tight reactive real-time tasks. RCS is an microsatellite is already in the orbit and hence launching,
example of hierarchical architectures [8]. Hybrid architectures releasing, and de−tumbling phases are ignored. The
are layered architectures that combine deliberative features of microsatellite includes four main subsystems. The payload
hierarchical architectures with reactive real-time characteristics subsystem is a panchromatic camera fixed to the satellite body.
of behavioral architectures. LAAS [9] and Claraty [10] are It is employed to take high resolution images from points of
examples of hybrid architectures. Remote agent (RA) is a interest on the earth. The second subsystem is attitude
hybrid architecture tested on deep space 1 [11]. It is designed determination and control system (ADCS). The satellite has
to provide reliable autonomy for extended periods [12]. 3+1 DOF (i.e. yaw, pitch, roll, and longitude), and cannot
Intelligent distributed execution architecture (IDEA) is an change orbit (altitude) or latitude. The ADCS is equipped with
agent oriented architecture that supports distributed autonomy a GPS receiver, a biaxial solar tracker and an inertial
[13]. It views each layer as an agent. IDEA has been evaluated measurement unit (IMU). Furthermore, it includes three
on K9 rover [12]. Mission data system (MDS) is a hybrid reaction wheels each fixed along a rotational axis for closed-
software framework which emphasizes the state estimation and loop attitude control. The third subsystem is power provider
state control [14]. TITAN is another hybrid architecture that which puts the satellite in an attitude with maximal solar
emphasizes model-based programming paradigm [14]. The energy absorption. The communication subsystem uplinks and
modern space systems including satellite systems such as EO-1 downlinks data.
[15] and Techsat-21 [16], and interplanetary missions such as
DS1 [11] are exploiting hybrid architectures. Hybrid
architectures are most efficient when coupled with hardware
architectures such as GUARDS that address parallel processing
and decomposition among layers [17].
Although many research works have been conducted on
onboard autonomy, only a few have employed proper
visualization methods. Some research works have exploited
hardware simulations. Aghili, Namvar and Vukovich have
mounted a satellite on a hydraulic manipulator to simulate a
satellite in orbit [18]. Bodin, Nylund, and Battelino have
implemented a hardware-in-the-loop (HIL) simulation to
evaluate the onboard software in a simulated real-time
environment [19]. Other research works have employed
software modeling. Zhaowei, Guodong, Xiaohui, and Xibin Fig. 1. The scenario of remote sensing mission.
B. Ground Stations and physical functionalities. Each action is defined by its pre-
Four ground stations are considered: two unidirectional conditions, sub-actions, services, possible termination
receiving ground stations, a TTC, and a MCC. The stations are conditions, resource production and consumption, logs,
connected via a virtual private network (VPN). The estimated execution time, and effects. Pre-conditions of an
unidirectional ground stations receive the data and send it to action must be satisfied before it can begin. When an action is
MCC. TTC is a bidirectional station that monitors the activated, it changes some states of the system, known as
performance of the satellite, tracks it in the orbit, and sends the action effects. Sub-actions are functions executed in serial or
logs to MCC. Also, it transmits prioritized missions from MCC parallel to map the action to low level instructions. As an
to the satellite. MCC prioritizes the users’ requests regarding example, attitude control is an action represented as follows:
parameters such as scientific importance. It generates high Action: Attitude_Control
level missions and transmits them to TTC. Also, it Estimated Time: 70s
synchronizes the ground stations and the satellite. The format Resources: Battery:0.05watt×T, CPU
Sub_Actions: Parallel(Sample(IMU,Solar)), Turn_On(Wheel), Control
of missions generated at MCC is depicted in (1). (Wheel), Turn_Off(Wheel)
Pre-conditions: IMU_STAT:idle, Solar_STAT:idle, Wheel_STAT:idle,
Battery > α
 IN#LO#LA#NO#FR#AT#PR  Effects: Direction_STAT: Ready

IN is the mission index; LO and LA are the longitude and

IV. PROPOSED ARCHITECTURE FOR ONBOARD AUTONOMY
latitude of the point of interest respectively. NO is the number
of images to be taken from that point, FR is the imaging The proposed intelligent architecture, shown in Fig. 2,
frequency, AT refers to the attitude mode (i.e. exact attitude consists of three layers including decision layer, execution
pointing or Nadir attitude pointing), and PR indicates the layer and functional layer.
mission priority. Using high frequency, the taken images will
have small angular differences. This mode is useful for A. Decision Layer
producing 3D images from the point of interest. On the other Decision layer plans and schedules the actions to reach the
hand, having low frequency is proper to observe the dynamics mission goals subject to resource constraints. It consists of two
of the point of interest. The payload is able to take images in main components including mission manager (MM) and
both exact attitude and Nadir attitude pointing modes. In deliberative planer/scheduler (PS). MM prioritizes the missions
former mode, the given longitude and latitude must appear in using a weighted combination of priorities determined by
the image center whereas in the later mode, appearance of the ground team, resource constraints and spatial priorities. It uses
given coordinates inside the image suffices. a priority queue to sort and send the missions to PS. It includes
two databases: action database and plan database. The action
C. Scenarios database keeps pre-defined actions and their attributes. Plan
When the microsatellite passes over the TTC, the missions database contains pre-planned mission prepared by experts to
are uplinked to it. The onboard intelligent software plans and boost up the planning process for routine missions. Before
schedules some activities to complete uploaded missions based commencing the planning process, the mission is checked
on priorities and resource constraints. As an example, when the against this database. If any matches are found, the plans are
satellite wants to take an image from a specific point, it first fetched and directly scheduled.
checks the power level. Then, it checks the availability of PS exploits action database to search for the sequence of
memory and camera. If these pre-conditions are satisfied, it actions that reach the mission goals with minimum costs. It
determines the current attitude. Then, it changes the attitude uses constraint-based iterative backtracking algorithm. PS
using ADCS to aim to the point of interest. Finally, it takes the determines the order of actions regarding the pre-conditions,
image. However, the scenarios are usually more complex. As effects, constraints and resource demands of actions. It
another example, assume that the mission is to take three incrementally adds actions to action sequence and in case that
images from TTC and when the microsatellite is passing over the pre-conditions are not satisfied, it backtracks and adds
TTC, the sun is in the view. In this scenario, satellite can another action. Also, it uses data provided by state
employ all of the four subsystems. However, there is a identification and estimation unit to determine the beginning
contradiction between power providing and image taking time of the root action as lumped in (2).
subsystems. They both need to employ ADCS. The former
subsystem wants to aim to the sun, and the later one wants to n

aim to the earth. Simultaneously, the satellite wants to Ts  Tc  Te   Ti (2)

i 1
communicate with TTC. Onboard intelligent software
determines the sequence of activities to optimize the mission Ts is start time, Tc is current time, Te is the time estimated to
achievements with minimum cost. view the target, n is the number of actions to be executed, and
The smallest entity used by onboard autonomy in decision Ti is an approximate execution time for each action.
making process is called action. An action represents an PS is a deliberative module. It needs high processing time
identical abstract activity to be performed by microsatellite. It in scale of minutes, and once it has finished the process, the
encapsulates the low level details of an activity and provides a outputs suffice for a long temporal horizon in scale of hours.
common knowledge representation between decision making
 SIE keeps the resource information (e.g. battery, memory,
bandwidth etc.), mission information (e.g. ground station
coordinates), reference timer, and updates them with high
frequency. Furthermore, it assimilates data acquired from
sensors using Kalman filter [26] and converts them to symbolic
values (e.g. HIGH, OFF etc.) using fuzzifiers [27]. Moreover,
it employs prediction functions to estimate the future states.
These functions are dynamic orbital equations that get current
state as input and calculate the next states. As an example,
passing over function gets the current orbital coordinates,
current reference time, and the target coordinates as input and
computes the reference time in which satellite will pass over
target coordinate. Using prediction functions, decision layer
generates accurate plans for wide temporal horizons.

C. Functional Layer
Fig. 2. Proposed autonomus architecture for onboard autonomy. Functional layer is an interface between software and
hardware. It consists of modules which connect each sub-
action to a particular hardware in a bijective manner. These
B. Execution Layer modules can be categorized to four classes including: sensing
This layer is a bridge between decision layer and flight modules, actuation modules, communication modules, and
software. It exploits the services in functional layer provided onboard services. Sensing modules are used to manage sensors.
by flight software to execute the actions planed and scheduled A sensing module activates a sensor, samples its data and
by decision layer. Furthermore, it provides reactive responses deactivates it. The data from sensors are propagated to SIE via
in critical situations. Execution layer consists of reactive smart executive. Actuation modules are employed to control
planner, smart executive and state identification and estimation actuators. An actuation module consists of activation,
unit (SIE). deactivation and control sub-modules. Control sub-module can
be a closed-loop controller (e.g. a digital PID controller for
Reactive planner responds to unpredicted situations. As controlling a reaction wheel) or an open-loop controller which
mentioned before, PS plans for wide temporal horizons. just sends specific signals (e.g. a shutter opening signal to
However, accumulation of sensor errors, estimation errors, and payload). It is noteworthy that modules have local
unpredicted events can invalidate the plans. In this case, intercommunications to support real-time functionality. Also, it
reactive planner responds to the situation using a heuristic is noteworthy that those sensors that do not affect the decision
algorithm. The heuristic algorithm uses the plan sequence and making or execution process are hidden from upper layers.
current state to perform an action to handle the situation in Shaft encoders used in reaction wheels are examples of these
minimum cost. The cost is a linear combination of the mission hidden sensors [28]. Communication modules include uplink
failure cost and resource costs. In case of unrecoverable failure, and downlink modules which are employed to transmit data
reactive planner suspends the onboard autonomy and delivers between satellite and ground stations. Onboard services are
the satellite control to the ground crew. utility software such as image processing and data compression
The smart executive receives scheduled actions from libraries. These services are invoked by actions to perform
decision layer. Prior to determined execution time, it checks the some data processing tasks.
pre-conditions of the action. If the pre-conditions are satisfied,
it assigns a thread to the corresponding action and starts it. The V. SIMULATION AND EXPERIMENTAL RESULTS
threads are synchronized by semaphores [25]. Assigning
In order to evaluate the stability and the performance of the
threads to each action provides parallel execution. Moreover,
proposed architecture, we have implemented an integrated SIL
the smart executive monitors the execution and detects
simulation and 3D visualization as illustrated in Fig. 3. SIL
inconsistencies. It monitors the estimated resource demands
simulation provides accurate data from satellite’s behavior in
and the actual resource consumptions. If the difference exceeds
response to various missions and situations. However, this data
the threshold, it terminates the action and informs the PS. If the
is represented as alphanumeric values and 2D diagrams, and
failure is fatal (e.g. power level is very low), it activates
needs to be interpreted to provide high level knowledge. To do
reactive planner to transform the satellite to a safe mode. Also,
so, we employed 3D visualization to show the real-time
when the plans are executed successfully, it informs the PS and
feedback from satellite’s behavior. It is noteworthy that we, for
demands new plans to execute. When an action thread starts, it
the first time, have integrated SIL simulations of flight
invokes the sub-actions and executes them in serial or parallel.
software, onboard autonomy, and 3D visualization.
Each sub-action directly activates a hardware-oriented
operation (i.e. sensing or actuating) or a service-oriented The simulation consists of onboard autonomy simulator,
operation (i.e. calling image processing or data compression satellite simulator, environment simulator, ground station
services) in functional layer. simulator and the communicator. The simulators perform a
distributed SIL simulation. Each simulator runs on an identical
computer with a 2.5 GHz processor and 1GB main memory.
The computers communicate with each other within a wireless
local area network (LAN) using TCP/IP protocol.
Onboard autonomy simulator implements the decision layer
and execution layer. This simulator is implemented using
visual C#.Net 2008. Each layer is implemented as an identical
process that communicates with its pair via a local socket. In
other words, decision and execution layers are executed in
parallel on a same computer. The communication protocol is
TCP/IP.
The satellite simulator implements the functional layer and
the satellite model. It simulates the virtual sensors such as GPS,
sun tracker, star tracker, and IMU with high precision using Fig. 3. Integrated software-in-the-loop simulation and visualization.
STK 8 (satellite tool kit). Also, a linear model of actuators (e.g.
reaction wheels) and the corresponding digital PID controller Using this interface, two users view the real-time
for real-time attitude control is implemented using visual simulation simultaneously by wearing two HMDs. Two Video
C++.Net 2008. The functional layer and the satellite model 3D Pro i-glasses are employed as HMDs. Furthermore, one of
communicate via a local port. The environment simulator the users is considered as the leader user who can interact with
employs dynamic orbital model to simulate LEO in which the environment by changing the point of view using an
satellite is circling. It forces the satellite to follow the Extreme 3D Pro joystick. When the leader user changes the
Newtonian mechanical rules. This simulator employs the point of view, the follower user feels the changes concurrently.
encapsulated STK models to simulate the space environment. By employing this interface, the users monitor the satellite’s
Communicator is a middleware that manages the message behavior in real-time. Users perceive behaviors such as power
passing between simulators. The messages contain data and providing (visualizing attitude changes toward the sun),
header. Communicator uses headers to extract the destination imaging (visualizing attitude changes toward the target point
simulator. It automatically converts the data embedded in and highlighting target point on earth), communicating
message to the knowledge representation understandable by the (visualizing transmitted electromagnetic signals between
target simulator. Finally, it composes a new message satellite and ground station, and highlighting target point on
containing converted data as message body and the id of sender earth) by viewing realistic physical behaviors of satellite in
simulator as its header, and transmits it to the target simulator. real-time in the space environment. Using this technic, users
The possible communication errors are automatically handled can primitively evaluate the stability and performance of the
via TCP/IP exceptions. Furthermore, the communicator satellite and then confirm their observations using data logs
synchronizes the simulators by providing a common time from first interface.
reference. The simulators periodically synchronize their local To validate the stability and evaluate the performance of
time with the common time reference by sending a time request the proposed autonomous architecture, various missions are
to the communicator. Communicator is implemented using planned by onboard intelligent software. The experts are asked
visual C#.Net 2008. to plan the same missions too. Then, the planned missions
The ground station simulator provides two types of user prepared by both software and experts are executed on
interface. The first interface simulates MCC. It let users to simulation framework and the results are compared as shown
communicate with satellite by uplinking missions, receiving in Table I. Comparing the average number of actions planned
data from satellite, and monitoring the satellite’s subsystems. by experts and intelligent software per mission shows that the
This interface is implemented using visual C#.Net 2008 and intelligent software outperforms the experts by 30%. Also,
employs windows graphical user interface (GUI) components. comparison between average failures per mission shows that
Although this interface displays data as alphanumeric values the intelligent software is 32% more reliable than experts.
and visualizes them as curve plots in real-time, due to the high Moreover, considering the computed standard deviations, it can
volume and update rate of data, it is not possible to interpret the be inferred that the intelligent software performs better in terms
data online. This interface is useful for monitoring the of stability. It is noteworthy that by increasing the number of
elaborated behavior of subsystems such as ADCS. Moreover, missions, efficiency and stability of experts’ plans decrease
this interface creates rich logs from data flow and lets the crew rapidly, whereas these criteria decrease slightly for onboard
analyze the preferred aspects of the satellite offline by using software. A comparison between the characteristics of four
these logs. hybrid architectures and our architecture is shown in Table II.
The second interface complements the first interface. It
TABLE I. PERFORMANCE OF INTELLIGET SOFTWARE AND EXPERTS
visualizes the satellite’s behavior in a 3D virtual environment
and provides an interactive 3D view of satellite orbiting the Per Intelligent software Experts
earth. This interface gives a high level interpretation of system mission Mean SD Mean SD
which is easily understandable by users.
Actions 4.37 1.45 6.23 3.84

Failures 0.08 0.03 0.25 0.17

 TABLE II. COMPARISON BETWEEN THE ARCHITECTURES [8] J. Albus and W. Rippey, “RCS: a reference model architecture for
intelligent control”, in Proc. From Perception to Action, 1994, pp.218-
Architecture Methodology #Abstraction Mission type 229.
IDEA Muti-agent Two levels Constellation [9] R. Alami, R. Chautila, S. Fleury, M. Ghallab, and F. Ingrand, “An
architecture for autonomy” Int. J. of Robot. Research, vol. 17, no. 4,
RA Multi-layered Three levels Interplanetary pp.315-337, Apr. 1998.
[10] I. Nesnas, R. Simmons, D. Gaines, C. Kunz, A. Calderon, T. Estlin, R.
MDS State-based One level Flight-Test
Madison, J. Guineau, M. McHenry, I. Shu, and D. Apfelbaum,
TITAN Model-based One level Flight-Test “CLARAty: challenges and steps toward reusable robotic software”, Int.
J. of Adv. Robot. Sys., vol. 3, no. 1, pp.23-30, 2006.
Proposed Multi-layered Three levels LEO [11] N. Muscettola, P. Nayak, B. Pell, and B. Williams, “Remote agent: to
boldly go where no AI system has gone before”, Artificial Intell., vol.
VI. CONCLUSION AND FUTURE WORKS 103, pp.5-48, 1998.
[12] M.S. Boddy, S.A. Harp, and K.S. Nelson, “CLOCKWORK:
In this paper, we proposed an intelligent architecture to requirements definition and technology evaluation for robust, compiled
provide onboard autonomy for an observation micro-satellite. autonomous spacecraft executives”, NASA, CA, Tech. Rep. Grant
The architecture integrates the low level physical actions with NAG-2-1624, 2004.
conceptual decision making ability in a hierarchical manner. [13] N. Muscettola, G.A. Dorais, C. Fry, R. Levinson, and Christian Plaunt,
To evaluate the proposed architecture, we implemented a “IDEA: planning at the core of autonomous reactive agents”, presented
at the Proc. of the 3rd Int. NASA Workshop on Planning and Scheduling
distributed SIL simulation to simulate the space, satellite, for Space, Houston, TX, 2002.
ground stations, and intelligent onboard software. Moreover,
[14] G. Horvath, M. Ingham, S. Chung, O. Martin, and B. Williams,
for the first time, we used virtual reality to visualize the “Practical application of model-based programming and state-based
satellite’s autonomous behavior in the orbit. It lets the users architecture to space missions”, in Pro. of the 2nd IEEE Int. Con. on
have a high level feedback from integrated simulation. Space Mission Challenges for Information Technology, 2006, pp.80-88.
Scenario-based evaluations showed that in average, the [15] B. Cichy, S. Chien, S. Schaffer, D. Tran, G. Rabideau, and R. Sherwood,
proposed architecture generates 30% less actions in “Validating the autonomous EO-1 agent”, presented at the Int. workshop
comparison with experts per mission. Also, results show that on Planning and Scheduling for Space, Darmstadt, Germany, 2004.
the average failures per mission caused by onboard software [16] R. Sherwood, S. Chien, R. Castano, and G. Rabideau, “Autonomous
planning and scheduling on the TechSat 21 mission”, in Proc. of the
are 32% less than experts. Finally, comparison between 15th Australian Joint Con. on Artificial Intelligence, 2002, pp.213-224.
standard deviations shows that the onboard software shows [17] D. Powell, J. Arlat, L.B. Dukic, A. Bondavalli, P. Coppola, A. Fantechi,
more steady responds in comparison with experts’ plans. Two E. Jenn, C. Rabejac, and A. Wellings, “GUARDS: a generic upgradable
developments are considered as future woks. First, the architecture for real-time dependable systems”, IEEE Trans. Parallel
simulation will be integrated with Google earth to show Distrib. Syst., vol. 10, no. 6, pp.580-599, Jun 1999.
realistic images taken from the point of interest. Second, the [18] F. Aghili, M. Namvar, and G. Vukovich, “Satellite simulator with a
architecture will be modified to handle the distributed missions hydraulic manipulator”, in Proc. of the 2006 IEEE Int. Con. on Robotics
and Automation, Orlando, pp. 3886-3892.
such as cooperating constellation of satellites and rendezvous
[19] P. Bodin, M. Nylund, and M. Battelino, “SATSIM—A real-time multi-
planning. satellite simulator for test and validation in formation flying projects”,
Acta Astronautica, vol. 74, pp.29-39, 2012.
ACKNOWLEDGMENT [20] S. Zhaowei, X. Guodong, L. Xiaohui, and C. Xibin, "The integrated
system for design, analysis, system simulation and evaluation of the
Authors would like to thank the guidance, navigation and small satellite", Adv. in Eng. Softw., vol. 31, pp.437-443, 2000.
control group members, especially Dr. Dehgan, Mr. Karimian, [21] Y. Miao, C. Chen, and Z. Sun, “A satellite system distributed simulation
Mr. Cheheltani, Mr. Gheibi, and Ms. Hemmatabadi. design and synchronous control”, in Proc. of the 2009 IEEE Int. Con.
On Mechatronics and Automation, pp.3889-3893.
REFERENCES [22] M. Mirshams, A. Nahvi, M. Khosrojerdi, H. Tae, and Ali Vahid, “A 6-
DoF satellite virtual simulator design and development”, Appl. Mech.
[1] A. Stroupe, S. Singh, R. Simmons, T. Smith, P. Tompkins, V.Verma, R. and Mater., vol. 186, pp.70-74, 2012.
Vitti-Lyons, and M.D. Wagner, “Technology for autonomous space [23] H. Hao, C. Pei, S. Yong, and H. Chao, “Real-time three dimensional
systems”, Robot. Inst., Carnegie Mellon Univ., Pittsburgh, PA, Tech. simulation platform for satellite mission analysis”, in 6th IEEE Con. on
Rep. CMU-RI-TR-00-02, 2000. Industrial Electronics and Applications, 2011, pp.2593-2598.
[2] B. Polle, “Autonomy requirement and technologies for future [24] R. Witt, M. Fritz, T. Kuwahara, A. Brandt, C. Laurel, H.-P. Roser, and J.
constellation” Astrium, Houston, TX, Tech. Rep. EAA.NT.BP. Eickhoff, “Real-time 3D Visualization in Satellite Development”, in 4th
3682899.02, 2002. Int. Con. on Astrodynamics Tools and Techniques, Madrid, Spane, 2010.
[3] S.M. Dehghan, S.H. Cheheltani, and A. Kiapasha, “Design and [25] A.S. Tanenbaum, Modern Operating Systems. 3rd ed. Upper Saddle
implementation of intelligent decision making system to generate flight River, NJ: Prentice Hall, 2008.
scenario automatically”, in 5th Int. Con. on Recent Advances in Space
Technologies, 2011, pp.178-183. [26] G. Evensen, Data Assimilation: The Ensemble Kalman Filter. 2nd ed.
Berlin, Germany: Springer, 2009.
[4] B.S. El-Haik and A. Shaout, Software Design for Six Sigma: A Roadmap
for Excellence. NJ:Wiley, 2010. [27] A. Kandel and G. Langholz, Fuzzy Control Systems. Boca Raton,
FL:CRC Press, 1994.
[5] R. Siegwart and I.R. Nourbakhsh, Introduction to Autonomous Mobile
Robots. Cambridge, MA: MIT Press, 2004. [28] M.C. Chou, C.M. Liaw, S.B. Chein, F.H. Shieh, J.R. Tsai, and H.C.
Chang, “Robust Current and Torque Controls for PMSM Driven
[6] R.E. Fikes, P. E. Hart. and N.J. Nilsson, “Learning and executing Satellite Reaction Wheel”, IEEE Trans. Aerosp. Electron. Syst., vol. 47,
generalized robot plans”, Artificial Intell., vol. 3, pp.251-288, 1972. no.1, pp. 58-74, Jan. 2012.
[7] R. Brooks, “A robust layered control system for a mobile robot” IEEE
Trans. Robot. Autom., vol. RA-2, pp.14-23, Mar. 1986.