Literature Survey

LITERATURE SURVEY
The survey is organized into sections covering prior work on var-ious

aspects presented in the subsequent chapters. Since a core component
of the material presented in this book is description of an FPGA-
based mobile robot developed for experiments, review of prior work
on sensors and processors for mobile robots is first taken up. This is
followed by a review of the literature on robotic explo-ration,
landmark determination and navigation. Finally, we touch upon
applications where FPGAs have been hitherto employed.
Sensors and Processors for Mobile Robots
Autonomous ground vehicles and mobile robots have been devel-

oped with dierent configurations and sensing capabilities since
approximately the 1970s. One of the earliest wheeled robots was the
SHAKEY [10]. The SHAKEY was based on a combination of on-
board computers and a radio link to larger computers else-where.
SHAKEYs primary sensor was a scanning camera. Around the early
1980s, the CMU Rover represented the state of the art in robotic
mobility. The CMU Rover also had multiple general-purpose
processors on-board and had provisions for connecting to a large
remote computer.
While the robots described had sensing and planning capa-bilities,
others which lacked elaborate planning capabilities were also
simultaneously developed. Example are the Robart-I and Robart-II
mobile robots [11]. Robart-I had a range of sensors, in-cluding
ultrasonic sensors, infrared proximity detectors and tactile
feelers. Robart-II employed a microwave motion detector and a

smoke detector besides infrared sensors. A good account of these
robots is provided by [12].
Mobile robotic platforms with other types of processors are rel-
atively recent. One example is the work reported in [13]. The authors
in [13] report development of a mobile robot with two processors on-
board: a TMS320F2812 DSP for implementing the control system for
the robot and a general-purpose processor (as part of a laptop
computer) for handling algorithms pertaining to navigation. Earlier
work [14] has focussed on use of a digital signal processor for
acquiring data from multiple ultrasonic sensors and processing. No
experiment with an actual robot is reported in [14].
Many different sensors and instruments have been used on mo-bile
robots. These include ultrasonic sensors, infrared sensors, vi-sion
sensors, tactile sensors, and encoders.
Ultrasonic sensors [15], [16] have been one of the simplest and
widely used sensors for measuring distance to the nearest obstacle
from a mobile robot. Design and development of circuits for dis-
tance measurement between a transmitter and receiver ultrasonic
transducer have been extensively researched [17]. Approaches to
correct various types of errors during distance measurement have
also been developed [18]. Use of multiple receivers (along with one
transmitter) has also been studied [19]. The exact point attaining
minimum distance cannot be easily obtained using an ultrasonic
sensor. This, however, is not a severe limitation for many applica-
tions in robotics.
Ultrasonic sensors have a range of approximately 3 metres and

constitute the primary sensor on the mobile robot developed as a part
of this book.
While sensors based on the ultrasonic, pulse-echo technique de-
scribed in the previous paragraph have been the mainstay for col-
lision avoidance systems in mobile robots, other alternatives that are
not based on sound have also been examined by researchers [20]. In
particular, where it is dicult to provide a medium for propagation of
sound waves, light-based sensors have been ex-plored. Some of these
sensors are based on time-of-flight and tri-angulation techniques. In
the mobile robot prototype developed
Robotic Operation in Known and Unknown
Environments 15
in this research, three infrared sensors are used for detection of

objects within a very small distance (less than 20 cm) from the robot
and to cut o power (via a relay) to the motors driving the robot
wheels. It is worth noting that light-based sensors are typ-ically used
to determine the presence of a target (or an object) rather than to
measure distance to them.
Studies on tactile sensors and whiskers have also been made by
researchers in the context of path planning for mobile robots. One
example is the work described in [21]. Work on vision sensors for
mobile robots also exists [22]. The authors in [22] discuss various
aspects of mobile robot navigation using vision sensors. The work
described in this research does not explicitly use tactile or vision
sensors.
2.2 Robotic Operation in Known and Unknown

Environments
A brief review of the literature on robotic operation in environ-ments

where the geometry of the objects is known beforehand (or a map is
available) is presented. We then move on to study prior work on
exploration and navigation in unknown environments.
2.2.1 Environment with Prior Knowledge of Object Geometries

and Locations
It is worth noting that prior exploration (combined perhaps with other

methods such as probing) would yield information on loca-tion and
shape of objects (that function typically as obstacles for the robot
during the course of navigation) in the environment. An important
task then is to find a collision-free path for the robot from some
initial position to a predefined destination. Several algo-rithms have
been proposed for navigation in an environment when the geometry
of the objects is completely known. These are based on notions such
as the visibility graph [2] and Voronoi diagrams [1]. Much work has
been done in the context of the piano movers problem [1], [23]. The
domain of computational geometry [3] has dealt with computational
aspects of geometric problems dealing with polygons representing
objects in the environment. While the solutions proposed for robotic
navigation for dierent models of
objects in the environment are interesting from a theoretical point of
view, it is often dicult to obtain accurate apriori information on
shapes and positions of the objects. The research reported in this
book assumes that the geometry of the objects is not available.
2.2.2 Unknown Environments
Since the core of this book is on various tasks in unknown envi-

ronments, we discuss work in dierent directions for this case.
Exploration and Mapping
One of the early eorts on exploration in unknown environments is

[24]. The authors in [24] develop a method that is particularly
appropriate for exploration and mapping in large-scale spatial en-
vironments. The authors examine the problem of errors in calcu-
lation of position for a mobile robot based on its encoders. The
authors present a topology-based approach that relies on sensor
signatures rather than on coordinates of a reference point on the
moving robot with respect to a global coordinate frame. Several
aspects pertaining to the diculty in building a map using topo-
logical information are brought out. The authors present simula-tion
results for map building.
Another piece of work on exploration is reported in [25]. The
authors in [25] consider the notion of an agent that locates itself on a
graph by matching nodes and the adjacency relationships between
them. The approach assumes that the agent can label each node by
depositing a marker at the nodes. The work reported is of an
algorithmic nature. Since the focus of the work reported in [25] is not
on implementation aspects, it is not clear how easily the ideas can be
realized in hardware.
Exploration of environment for building specific types of geo-
metric structures has also been performed [26], [27]. In [26], the
authors present an incremental approach to construct a Voronoi
diagram-like structure. In particular, the authors consider the problem
of learned navigation of a circular robot through a two-dimensional
terrain whose model is not a priori known. The au-thors present
algorithmic details for (a) the visit problem where the robot is
required to visit a sequence of destination points and
2.2 Robotic Operation in Known and Unknown
Environments 17
(b) the terrain model acquisition problem where the robot is re-quired
to acquire the complete model of the terrain. Experimental results are
not available.
Construction of hierarchical generalized Voronoi graph for an
unknown environment in two and higher dimensions is presented in
[27], [28]. The authors in [27] present a scheme based on predic-tion
and correction for incremental construction of the hierarchical
generalized Voronoi graph. Planar and three-dimensional simula-
tions are presented. Some experiments on a mobile robot equipped
with a general-purpose processor and sonar are briefly described.
Construction of generalized local Voronoi diagram using laser
range scanner data has been studied in [29]. The scheme developed in
[29] is based on clusterization of scan points based on some property.
In particular, clusterizations involving interdistance of successive
scan points and distance to the nearest neighbor have been studied. It
is worth noting that the above scheme is for a static environment and
simulations are only presented.
Construction of other geometric structures based on sensor data
has also been researched. In [30], the authors study the construc-tion
of a Local Tangent Graph (LTG) using range sensor data. The local
tangent graph is then used for navigation. In particular, the local
tangent graph helps to choose the locally optimal direction for a robot
as it moves towards the destination point. Also, the robot uses the
LTG for following the boundary of obstacles. The authors present
simulation results for their scheme.
An approach to exploration based on visiting landmarks is pre-
sented in [31]. In [31], the work is based on a robot that maintains a
list of list of all unvisited landmarks in the environment.
Considerable work has also been done on planning robot mo-tion
strategies for ecient model construction [32]. Work in this direction
has been on finding a function that reflects intuitively how the robot
should explore the space so that we have a com-promise between
possible elimination of unexplored space and dis-tance travelled.
While most of the work in the area of robotic exploration has
concentrated on static environments, there has been some work on
semi-dynamic and dynamic environments. Mapping semi-dynamic
environments where objects (such as chairs, tables) move
18 2 Literature Survey
periodically from one location to another has been studied in [33]. Some work on detecting
changes in a dynamic environment for updating an existing map supported by computer
simulations is described in [34]. Experiments with a mobile robot have been re-ported by some
authors [35], [33].
More recently, an approach for planning exploration strate-gies for simultaneous localization
and mapping in a static en-vironment has been proposed [36]. The authors in [36] give a method
to select the next robot position for exploration based on a novel utility function. The utility
function defined in [36] combines geometric information with intensive usage of results ob-
tained from perceptual algorithms. The outcome of the exploration is a multi-representational
map made up of polygons, landmarks and a roadmap. Experiments with a real robot and
simulations are presented in [36] but the focus is on using a general-purpose processor. Also, this
work does not deal with dynamic environ-ments.

Literature Survey

Uploaded by

Copyright:

Available Formats

Literature Survey

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Literature Survey

Uploaded by

Copyright:

Available Formats

LITERATURE SURVEY

The survey is organized into sections covering prior work on var-ious

Sensors and Processors for Mobile Robots

Autonomous ground vehicles and mobile robots have been devel-

feelers. Robart-II employed a microwave motion detector and a

Ultrasonic sensors have a range of approximately 3 metres and

in this research, three infrared sensors are used for detection of

2.2 Robotic Operation in Known and Unknown

A brief review of the literature on robotic operation in environ-ments

2.2.1 Environment with Prior Knowledge of Object Geometries

It is worth noting that prior exploration (combined perhaps with other

2.2.2 Unknown Environments

Since the core of this book is on various tasks in unknown envi-

Exploration and Mapping

One of the early eorts on exploration in unknown environments is

You might also like