See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/221710040

Mobile Robotics, Moving Intelligence

Chapter · December 2006

DOI: 10.5772/4716 · Source: InTech

CITATIONS READS
7 204

4 authors, including:

Souma Alhaj Ali Xiaoqun Liao

Hamdan Bin Mohammed e-University University of Cincinnati
23 PUBLICATIONS   101 CITATIONS    17 PUBLICATIONS   75 CITATIONS   

SEE PROFILE SEE PROFILE

Ernest L. Hall
University of Cincinnati
525 PUBLICATIONS   3,792 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Inventions View project

Computed tomography CT Scans View project

All content following this page was uploaded by Ernest L. Hall on 26 May 2014.

The user has requested enhancement of the downloaded file.

 5

Mobile Robotics, Moving Intelligence

Souma Alhaj Ali*; Masoud Ghaffari, Xiaoqun Liao and Ernest Hall
*The Hashemite University, Jordan
University of Cincinnati, U.S.A.

1. Introduction
After proving to be an efficient tool for improving quality, productivity, and
competitiveness of manufacturing organizations, robots now expand to service
organizations, offices, and even homes. Global competition and the tendency to
reduce production cost and increase efficiency creates new applications for robots
that stationary robots can't perform. These new applications require the robots to
move and perform certain activities at the same time. The availability and low cost of
faster processors, better programming, and the use of new hardware allow robot
designers to build more accurate, faster, and even safer robots. For example, Egemin
Automation has been selling mobile robots for a number of years, available as a
specialized unit that delivers mail within a large building to warehouse automation
systems. Currently, mobile robots are expanding outside the confines of buildings
and into rugged terrain, as well as familiar environments like schools and city streets
(Wong, 2005).
The problem addressed in this paper is to describe the theory and architecture of robust
learning for mobile robots and to illustrate the theory for designing intelligent mobile robots
for a wide variety of applications anywhere in the world including complex and uncertain
Open Access Database www.i-techonline.com

environments. The proposed architecture for machine learning is also based on the
perceptual creative controller for an intelligent robot that uses a multi- modal adaptive critic
for performing learning in an unsupervised situation but can also be trained for tasks in
another mode and then is permitted to operate autonomously. The robust nature is derived
from the automatic changing of modes based on internal measurements of error at
appropriate locations in the controller.

2. Types of Mobile Robots

Many different types of mobile robots had been developed depending on the kind of
application, velocity, and the type of environment whether its water, space, terrain with fixed
or moving obstacles. Four major categories had been identified (Dudek & Jenkin, 2000):
Terrestrial or ground-contact robots: The most common ones are the wheeled
robots; others are the tracked vehicles and Limbed vehicles.
Aquatic robots: Those operate in water surface or underwater. Most use water jets
or propellers.
Airborne robots: Flying robots like Robotic helicopters, fixed-wing aircraft,
robotically controlled parachutes, and dirigibles.

Source: Mobile Robots, Moving Intelligence, ISBN: 3-86611-284-X, Edited by Jonas Buchli, pp. 576, ARS/plV, Germany, December 2006
96 Mobile Robotics, moving intelligence

Space robots: Those are designed to operate in the microgravity of outer space and
are typically envisioned for space station maintenance. Space robots either move by
climbing or are independently propelled.

2.1 Terrestrial or Ground-contact Robots

There are three main types of ground-contact robots: wheeled robots, tracked vehicles, and
limbed vehicles. Wheeled robots exploit friction or ground contact to enable the robot to
move. Different kinds of wheeled robots exist: the differential drive robot, synchronous
drive robot, steered wheels robots and Ackerman steering (car drive) robots, the tricycle,
bogey, and bicycle drive robots, and robots with complex or compound or omnidirectional
wheels. Army Research Labs with NIST support demonstrated defence robot vehicles, called
Experimental Unmanned Vehicles (XUV) that developed at a cost of approximately $50
million over four years. It used technology from leading robotics laboratories in the US and
Germany. XUV performed autonomous scout missions in difficult off-road terrain and run
with general goal points and mission profiles given by Army scouts. XUV navigates through
woods and fields to find and report enemy targets. In a demonstration, the XUVs drove
autonomously over difficult terrain including dirt roads, trails, tall grass, weeds, brush, and
woods. Using on board sensors, the XUVs were able to detect and avoid both positive and
negative obstacles. The Demo III XUVs have repeatedly navigated kilometres of difficult off-
road terrain with only high-level mission commands provided by an operator from a remote
location. Figure 1 shows the XUV in action at Ft. Indiantown Gap (Alhaj Ali, 2003,
Greenhouse & Norris, 2002). Another example is Bearcat III which was developed in the
University of Cincinnati Robotics Center and is shown in Fig. 2. It has two driven wheels
and a caster wheel. The driven wheels are of a fixed-wheel type.

Fig. 1. Experimental Unmanned Vehicle in action at Ft. Indiantown Gap. Photo courtesy of
the Army Research Labs (Greenhouse & Norris, 2002).
Mobile Robotics, Moving Intelligence 97

Fig. 2. University of Cincinnati Bearcat III.

For more information on the construction, kinematic, and dynamic models of wheeled
mobile robots, refer to Alhaj Ali (2003), de Wit, et al. (1996), and Dudek & Jenkin, (2000).
Tracked vehicles are robust to any terrain environment, their construction are similar to the
differential drive robot but the two differential wheels are extended into treads which provide a
large contact area and enable the robot to navigate through a wide range of terrain.
Limbed vehicles are suitable in rough terrains such as those found in forests, near natural or
man-made disasters, or in planetary exploration, where ground contact support is not
available for the entire path of motion. Limbed vehicles are characterized by the design and
the number of legs, the minimum number of legs needed for a robot to move is one, to be
supported a robot need at least three legs, and four legs are needed for a statically stable
robot, six, eight, and twelve legs robots exists. For example of a limbed robot, ASIMO which
is a humanoid two legs walking robot developed by Honda and features the ability to
pursue key tasks in a real-life environment such as an office, ASIMO has an advanced level
of physical capabilities. The new ASIMO model is shown in Fig. 3 (Honda, 2006).

Fig. 3. The new ASIMO model (Honda, 2006).

2.2 Aquatic Robots

Aquatic vehicles support propulsion by utilizing the surrounding water. There are two
common structures (Dudek & Jenkin, 2000): torpedo-like structures (Feruson & Pope, 1995,
Kloske et al., 1993) where a single propeller provides forward, and reverse thrust while the
navigation direction is controlled by the control surfaces, the buoyancy of the vessel controls
the depth. The disadvantage of this type is poor manoeuvrability.
Twin-burger (Fuji et al., 1939) and URV vehicles (Choi, 1993, Choi et al., 1995) robots which
uses a collection of thrusters that are distributed over the vessel, more manoeuvrable are
attained by controlling sets of the thrusters to change the vehicle orientation and position
98 Mobile Robotics, moving intelligence

independently, however, the URV comes with the expense of operational speed. An
example of an aquatic robot is shown in Fig. 4.a.

2.3 Flying Robots

Fixed-wing autonomous vehicles: This utilizes control systems very similar to the ones
found in commercial autopilots. Ground station can provide remote commands if needed,
and with the help of the Global Positioning System (GPS) the location of the vehicle can be
determined (Dudek & Jenkin, 2000).
Automated helicopters (Baker et al., 1992, Lewis et al., 1993): These use onboard
computation and sensing and ground control, their control is very difficult compared to the
fixed-wing autonomous vehicles.
Buoyant (aerobots, aerovehicles, or blimps) vehicles: These vehicles can float and are
characterized by having high energy efficiency ration, long-range travel and duty cycle, vertical
mobility, and they usually has no disastrous results in case of failure (Dudek & Jenkin, 2000).
Unpowered autonomous flying vehicles: These vehicles reach their desired destination by
utilizing gravity, GPS, and other sensors (Dudek & Jenkin, 2000). An example of a flying
robot is shown in Fig. 4.b.

a. Aquatic Vehicle b. Flying Vehicle

Fig. 4. Aquatic and Flying vehicle (SSC San Diego, 2002).

2.4 Space Robots

These are needed for applications related to space stations like construction, repair, and
maintenance. Free-flying systems have been proposed where the spacecraft is equipped
with thrusters with one or more manipulators, the thrusters are utilized to modify the robot
trajectory. For more information on the construction and kinematic models of mobile robots
please refer to (Dudek & Jenkin, 2000).

3. Navigation
Navigation is the major challenge in the autonomous mobile robots; a navigation system is
the method for guiding a vehicle. Several capabilities are needed for autonomous navigation
(Alhaj Ali, 2003):
• The ability to execute elementary goal achieving actions such as going to a given
location or following a leader;
• The ability to react to unexpected events in real time such as avoiding a suddenly
appearing obstacle;
Mobile Robotics, Moving Intelligence 99

• The ability to formulate a map of the environment;

• The ability to learn which might include noting the location of an obstacle and of a
three-dimensional nature of the terrain and adapt the drive torque to the inclination of
hills (Golnazarian & Hall, 2000).
During the last fifteen years, a great deal of research has been done on the interpretation
of motion fields as the information they contain about the 3-D world. In general, the
problem is compounded by the fact that the information that can be derived from the
sequence of images is not the exact projection of the 3D-motion field, but rather
information about the movement of light patterns and optical flow (Golnazarian & Hall,
2000, Alhaj Ali, 2003).

3.1 Systems and Methods for Mobile Robot Navigation

Odometry and other dead-reckoning methods: These methods use encoders to measure
wheel rotation and/or steering orientation (Alhaj Ali, 2003).
Vision based navigation: Computer vision and image sequence techniques were proposed
for obstacle detection and avoidance for autonomous land vehicles that can navigate in an
outdoor road environment. The object shape boundary is first extracted from the image,
after the translation from the vehicle location in the current cycle to that in the next cycle,
the position of the object shape in the image of the next cycle is predicted, then it is matched
with the extracted shape of the object in the image of the next cycle to decide whether the
object is an obstacle (Alhaj Ali, 2003, Chen & Tsai, 2000).
Sensor based navigation: Sensor based navigation systems that rely on sonar or laser
scanners that provide one dimensional distance profiles have been used for collision and
obstacle avoidance. A general adaptable control structure is also required. The mobile robot
must make decisions on its navigation tactics; decide which information to use to modify its
position, which path to follow around obstacles, when stopping is the safest alternative, and
which direction to proceed when no path is given. In addition, sensors information can be
used for constructing maps of the environment for short term reactive planning and long-
term environmental learning (Alhaj Ali, 2003).
Inertial navigation: This method uses gyroscopes and sometimes accelerometers to measure
the rate of rotation and acceleration.
Active beacon navigation systems: This method computes the absolute position of the robot
from measuring the direction of incidence of three or more actively transmitted beacons.
The transmitters, usually using light or radio frequencies must be located at known sites in
the environment (Janet, 1997, Premvuti & Wang, 1996, Alhaj Ali, 2003).
Landmark navigation: In this method distinctive artificial landmarks are placed at known
locations in the environment to be detected even under adverse environmental conditions
(Alhaj Ali, 2003).
Map-based positioning: In this method information acquired from the robot's onboard
sensors is compared to a map or world model of the environment. The vehicle's absolute
location can be estimated if features from the sensor-based map and the world model map
match (Alhaj Ali, 2003).
Biological navigation: biologically-inspired approaches were utilized in the development of
intelligent adaptive systems; biomimetic systems provide a real world test of biological
navigation behaviours besides making new navigation mechanisms available for indoor
robots (Alhaj Ali, 2003).
100 Mobile Robotics, moving intelligence

Global positioning system (GPS): This system provides specially coded satellite signals
that can be processed in a GPS receiver, enabling it to compute position, velocity, and time
(Alhaj Ali, 2003).

4. Mobile Robots and Artificial Intelligence

While robotics research has mainly been concerned with vision (eyes) and tactile, some
problems regarding adapting, reasoning, and responding to changed environment have
been solved with the help of artificial intelligence using heuristic methods such as ANN.
Neural computers have been suggested to provide a higher level of intelligence that allows
the robot to plan its action in a normal environment as well as to perform non-programmed
tasks (Golnazarian & Hall, 2000, Alhaj Ali, 2003).
A well established field in the discipline of control systems is the intelligent control, which
represents a generalization of the concept of control, to include autonomous anthropomorphic
interactions of a machine with the environment (Alhaj Ali, 2003, Meystel & Albus, 2002). Meystel
and Albus (2002) defined intelligence as “the ability of a system to act appropriately in an
uncertain environment, where an appropriate action is that which increases the probability of
success, and success is the achievement of the behavioural sub goals that support the system’s
ultimate goal”. The intelligent systems act so as to maximize this probability. Both goals and
success criteria are generated in the environment external to the intelligent system. At a
minimum, the intelligent system had to be able to sense the environment which can be achieved
by the use of sensors, then perceive and interpret the situation in order to make decisions by the
use of analyzers, and finally implements the proper control actions by using actuators or drives.
Higher levels of intelligence require the abilities to: recognize objects and events store and use
knowledge about the world, learn, and to reason about and plan for the future. Advanced forms
of intelligence have the ability to perceive and analyze, to plan and scheme, to choose wisely, and
plan successfully in a complex, competitive, and hostile world (Alhaj Ali, 2003).
Intelligent behaviour is crucial to mobile robots; it could be supported by connecting perception
to action (Kortenkamp et al., 1998). In the following a brief review for the literature in the use of
artificial intelligence in mobile robots will be presented (Alhaj Ali, 2003).

4.1 Use of Artificial Neural Networks ( ANN)

ANN has been applied to mobile robot navigation. It had been considered for applications that
focus on recognition and classification of path features during navigation. Kurd and Oguchi
(1997) propose the use of neural network controller that was trained using supervised learning as
an indirect-controller to obtain the best control parameters for the main controller in use with
respect to the position of the AGV. A method that uses incremental learning and classification
based on a self-organizing ANN is described by Vercelli and Morasso (1998). Xue and Cheung
(1996) proposed a neural network control scheme for controlling active suspension. The
presented controller used a multi-layer neural network and a prediction-correction method for
adjusting learning parameters. Dracopoulos (1998) present the application of multi-layer
perceptrons to the robot path planning problem and in particular to the task of maze navigation.
Zhu, et al. (1998) present results of integrating omni-directional view image analysis and a set of
adaptive networks to understand the outdoor road scene by a mobile robot (Alhaj Ali, 2003).
To navigate and recognize where it is, a mobile robot must be able to identify its current location.
The more the robot knows about its environment, the more efficiently it can operate (Cicirelli,
1998). Grudic and Lawrence (1998) used a nonparametric learning algorithm to build a robust
Mobile Robotics, Moving Intelligence 101

mapping between an image obtained from a mobile robot’s on-board camera, and the robot’s
current position. It used the learning data obtained from these raw pixel values to automatically
choose a structure for the mapping without human intervention, or any prior assumptions about
what type of image features should be used (Alhaj Ali, 2003).

4.2 Use of Fuzzy Logic

Fuzzy logic and fuzzy languages have also been used in navigation algorithms for mobile
robots as described in (Wijesoma et al., 1999, Mora and Sanchez, 1998). Lin and Wang (1997)
propose a fuzzy logic approach to guide an AGV from a starting point toward the target
without colliding with any static obstacle as well as moving obstacles; they also study other
issues as sensor modelling and trap recovery. Kim and Hyung (1998) used fuzzy multi-
attribute decision-making in deciding which via-point the robot should proceed to at each
step. The via-point is a local target point for the robot’s movement at each decision step. A
set of candidate via-points is constructed at various headings and velocities. Watanabe, et al.
(1998) described a method using a fuzzy logic model for the control of a time varying
rotational angle in which multiple linear models are obtained by utilizing the original non-
linear model at some representative angles (Alhaj Ali, 2003).

4.3 Use of Neural Integrated Fuzzy Controller

A neural integrated fuzzy controller (NiF-T) that integrates the fuzzy logic representation of
human knowledge with the learning capability of neural networks has been developed for
nonlinear dynamic control problems (Alhaj Ali, 2003). Daxwanger and Schmidt (1998)
presented their neuro-fuzzy approach for visual guidance of a mobile robot vehicle.

5. Navigation in Unstructured Environments

Some research has been conducted regarding robotics in unstructured environment. Alhaj
Ali (2003) presented the development of an autonomous navigation and obstacle avoidance
system for a wheeled mobile robot operating in unstructured outdoor environments. The
algorithm presented produces the robot’s path positioned within the road boundaries and
avoids any fixed obstacles along the path. The navigation algorithm was developed from a
feedforward multilayer neural network. The network used a quasi-Newton
backpropagation algorithm for training. Computed-torque, digital, and adaptive controllers
were developed to select suitable control torques for the motors, which cause the robot to
follow the desired path from the navigation algorithm.
Martínez and Torras (2001) presented visual procedures especially tailored to the constraints
and requirements of a legged robot that works with an un-calibrated camera, with pan and
zoom, freely moving towards a stationary target in an unstructured environment that may
contain independently-moving objects.
Kurazume and Hirose (2000) proposed a new method called Cooperative Positioning
System (CPS). The main concept of CPS is to divide the robots into two groups, A and B,
where group A remains stationary and acts as a landmark while group B moves. Then
group B stops and acts as a landmark for group A. This process is repeated until the target
position is reached. Their application was a floor-cleaning robot system. Torras (1995)
reviewed neural learning techniques for making robots well adapted to their
surroundings.
102 Mobile Robotics, moving intelligence

Yahja et al. (2000) propose an on-line path planner for outdoor mobile robots using a
framed-quadtrees data structure and an optimal algorithm to incrementally re-plan optimal
paths. They showed that the use of framed-quadtrees leads to paths that are shorter and
more direct compared to the other representations; however, their results indicate that
starting with partial information is better than starting with no information at all. Baratoff et
al. (2000) designed a space-variant image transformation, called the polar sector map, which
is ideally suited to the navigational tasks.
Yu et al. (2001) presented a hybrid evolutionary motion planning simulation system for
mobile robots operating in unstructured environments, based on a new obstacle
representation method named cross-line, a follow boundary repair approach, and a hybrid
evolutionary motion planning algorithm. Yan et al. (1997) presents an attempt to devise and
develop a domain-independent reasoning system scheme for handling dynamic threats, and
uses the scheme for automated route planning of defence vehicles in an unstructured
environment (Alhaj Ali, 2003).
Computer vision and image sequence techniques were proposed for obstacle detection and
avoidance for autonomous land vehicles that can navigate in an outdoor road environment.
Jarvis (1996) report some preliminary work regarding an autonomous outdoor robotic
vehicle navigation using flux gate compass, DGPS and range sensing, and distance
transform based path planning. An adaptive navigation method suited for the complex
natural environments had been proposed based on a multi-purpose perception system that
manages different terrain representations, the method focuses on the functions that deal
with the navigation planning and the robot self-localization which have been integrated
within the robot control system (Devy, 1995).
Krishna and Kalra (2001) proposed incorporating cognition and remembrance capabilities in
a sensor-based real-time navigation algorithm, they stated that these features enhance the
robots performance by providing a memory-based reasoning whereby the robot’s
forthcoming decisions are also affected by its previous experiences during the navigation,
which is apart from the current range inputs, the suggested robot navigates in a concave
maze-like unstructured altered environment which has been modelled by classifying
temporal sequences of special sensory patterns.
Marco et al. (1996) developed a hybrid controller for semi-autonomous and autonomous
underwater vehicles in which the missions imply multiple task robot behaviour. They proposed
the use of Prolog as a computer language for the specification of the discrete event system (DES)
aspects of the mission control, and made the connections between a Prolog specification and the
more common Petri Net graphical representation of a DES (Alhaj Ali, 2003).

6. Controllers for Autonomous Mobile Robots

Robots and robots manipulators have complex nonlinear dynamics that make their accurate
and robust control difficult. On the other hand, they fall in the class of Lagrangian
dynamical systems, so that they have several extremely nice physical properties that make
their control straight forwarded (Lewis et al., 1999). Different controllers had been
developed for the motion of robot manipulators, however, not until recently where there has
been an interest in moving the robot itself, not only its manipulators (Alhaj Ali, 2003).
Shim and Sung (2003) proposed a WMR asymptotic control with driftless constraints based
on empirical practice using the WMR kinematic equations. They showed that with the
appropriate selection of the control parameters, the numerical performance of the
Mobile Robotics, Moving Intelligence 103

asymptotic control could be effective. The trajectory control of a wheeled inverse pendulum
type robot had been discussed by Yun-Su and Yuta (1998), their control algorithm consists
of balance and velocity control, steering control, and straight line tracking control for
navigation in a real indoor environments (Alhaj Ali, 2003).
Rajagopalan and Barakat (1997) developed a computed torque control scheme for Cartesian
velocity control of WMRs. Their control structure can be used to control any mobile robot if
its inverse dynamic model exists. A discontinuous stabilizing controller for WMRs with
nonholonomic constraints where the state of the robot asymptotically converges to the
target configuration with a smooth trajectory was presented by Zhang and Hirschorn (1997).
A path tracking problem was formulated by Koh and Cho (1999) for a mobile robot to
follow a virtual target vehicle that is moved exactly along the path with specified velocity.
The driving velocity control law was designed based on bang-bang control considering the
acceleration bounds of driving wheels and the robot dynamic constraints in order to avoid
wheel slippage or mechanical damage during navigation. Zhang et al. (2003) employed a
dynamic modelling to design a tracking controller for a differentially steered mobile robot
that is subject to wheel slip and external loads (Alhaj Ali, 2003).
A sliding mode control was used to develop a trajectory tracking control in the presence of
bounded uncertainties (Corradini & Orlando, 2001). A solution for the trajectory tracking problem
for a WMR in the presence of disturbances that violate the nonholonomic constraint is proposed
later by the same authors based on discrete-time sliding mode control (Corradini et al., 2002).
An electromagnetic approach for path guidance of a mobile-robot-based automatic transport
service system with a PD control algorithm was investigated by Wu et al. (2001). Jiang, et al.
(2001) developed a model-based control design strategy that deals with global stabilization
and global tracking control for the kinematic model with a nonholonomic WMR in the
presence of input saturations. An adaptive robust controller was proposed for the global
tracking problem for the dynamic of the non-holonomic systems with unknown dynamics
(Dong, 1999). However, real time adaptive controls are not common in practical applications
due partly to stability problems (Werbos, 1999, Alhaj Ali, 2003).
A fuzzy logic controller had been tried for WMRs navigation. Montaner and Ramirez-
Serrano (1998) developed a fuzzy logic controller that can deal with the sensors inputs
uncertainty and ambiguity for direction and velocity manoeuvres. A locomotion control
structure was developed based on the integration of an adaptive fuzzy-net torque
controller with a kinematic controller to deal with unstructured unmodeled robot
dynamics for a non-holonomic mobile robot cart (Topalov, 1998). Toda et al. (1999)
employed a sonar-based mapping of crop rows and fuzzy logic control-based steering
for the navigation of a WMR in an agricultural environment. They constructed a crop
row map from the sonar readings and transferred it to the fuzzy logic control system,
which steers the robot along the crop row. A local guidance control method for wheeled
mobile robots using fuzzy logic for guidance, obstacle avoidance and docking was
proposed by Vázquez and Garcia (1994), the method provide a smooth but not
necessary optimal solution (Alhaj Ali, 2003).

7. Creative Control for Intelligent Mobile Robots

Since mobile robots must be able to select among many concurrent tasks, such as moving while
sensing and reacting and learning, a new control structure is needed. The creative controller
concept is excellent for complex environments since it permits a divide and conquer approach,
104 Mobile Robotics, moving intelligence

with one or more tasks at a time in a multi-threaded, distributed computing environment.

Creative learning architectures integrate a Task Control Center (TCC) and a dynamic database
(DD) and adaptive critic learning algorithms to permit these solutions. Determining the task to be
performed and the data base to be updated are the two key elements of the design. These new
decision processes encompass both decision and estimation theory and can be modeled by
neural networks and implemented with multi-threaded computers.
The control architectures for neural network control of vehicles in which the kinematic and
dynamic models are known but one or more parameters must be estimated is a simple task
that has been demonstrated. The mathematical models for the kinematics and dynamics
were developed and the main emphasis was to explore the use of neural network control
and demonstrate the advantages of these learning methods. The results indicate the method
of solution is appropriate and that it has potential application to a large number of currently
unsolved problems in complex environments. The adaptive critic neural network control is
an important starting point for future learning theories that are applicable to robust control
and learning situations.
To obtain broadly applicable results, a generalization of adaptive critic learning called
Creative Control (CC) for intelligent robots in complex, unstructured environments has been
used. The creative control learning architecture integrates a Task Control Center (TCC) and
a Dynamic Knowledge Database (DKD) with adaptive critic learning algorithms. Recently
learning theories such as the adaptive critic have been proposed in which a critic provides a
grade to the controller of an action module such as a robot. The creative control process is
used that is “beyond the adaptive critic”. A mathematical model of the creative control
process is presented that illustrates the use for mobile robots.

7.1 Dynamic Programming

The intelligent robot is defined as a decision maker for a dynamic system that may make
decisions in discrete stages or over a continuous or discrete time horizon. The outcome of
each decision may not be fully predictable but may be anticipated or estimated to some
extent before the next decision is made. Furthermore, an objective or cost function can be
defined for the decision. There may also be natural constraints. Generally, the goal is to
minimize this cost function over some decision space subject to the constraints. With this
definition, the intelligent robot can be considered as a set of problems in dynamic
programming and optimal control (Liao, 2002).
Dynamic programming (DP) is the only approach for sequential optimization applicable to
general nonlinear, stochastic environments. However, DP needs efficient approximate
methods to overcome its dimensionality problems. It is only with the presence of artificial
neural network (ANN) and the invention of back propagation that such a powerful and
universal approximate method has become a reality.
The essence of dynamic programming is Bellman's Principle of Optimality. (Bellman 1957)
“An optimal policy has the property that whatever the initial state and initial
decision are, the remaining decisions must constitute an optimal policy with
regard to the state resulting from the first decision”.
The original Bellman equation of dynamic programming for adaptive critic algorithms may
be written as shown in Eq (1):
J ( R (t )) = max (U ( R (t ), u (t ))+ < J ( R (t + 1)) >) /(1 + r ) − U 0 (1)
u (t )
Mobile Robotics, Moving Intelligence 105

where R(t) is the model of reality or state form, U( R(t),u(t)) is the utility function or local
cost, u(t) is the action vector, J(R(t)) is the criteria or cost-to-go function at time t, r and U0
are constants that are used only in infinite-time-horizon problems and then only sometimes,
and where the angle brackets refer to expected value.
The user provides a utility function, U, and a stochastic model of the plant, R, to be
controlled. The expert system then tries to solve the Bellman equation for the chosen model
and utility function to achieve the optimum value of J by picking the action vector u(t). If an
optimum J cannot be determined, an approximate or estimate value of the J function is used
to obtain an approximate optimal solution. Regarding the finite horizon problems, which we
normally try to cope with, one can use Eq ( 2):

J ( R (t )) = max (U ( R (t ), u (t ))+ < J ( R(t + 1)) >) /(1 + r ) (2)

u (t )

Dynamic programming gives the exact solution to the problem of how to maximize a
utility function U(R(t), u(t)) over the future times, t, in a nonlinear stochastic
environment. Dynamic programming converts a difficult long-term problem in
optimization over time <U(R(t))>, the expected value of U(R(t)) over all the future
times, into a much more straightforward problem in simple, short-term function
maximization after we know the function J. Thus, all of the approximate dynamic
programming methods discussed here are forced to use some kind of general-purpose
nonlinear approximation to the J function, the value function in the Bellman equation,
or something closely related to J.
In most forms of adaptive critic design, we approximate J by using a neural network.
Therefore, we approximate J(R) by some function Jˆ ( R, W ) , where W is a set of weights
or parameters, and Ĵ is called a Critic network. If the weights W are adapted or
iteratively solved for, in real time learning or offline iteration, we call the Critic an
Adaptive Critic.
An adaptive critic design (ACD) is any system which includes an adapted critic component;
a critic, in turn, is a neural net or other nonlinear function approximation which is trained to
converge to the function J(X). In adaptive critic learning or designs, the critic network learns
to approximate the cost-to-go or strategic utility function J and uses the output of an action
network as one of it's inputs, directly or indirectly. When the critic network learns, back
propagation of error signals is possible along its input feedback to the action network. To
the back propagation algorithm, this input feedback looks like another synaptic connection
that needs weights adjustment. Thus, no desired control action information or trajectory is
needed as supervised learning.

7. 2. Adaptive Critic and Creative Control

Most advanced methods in neurocontrol are based on adaptive critic learning
techniques consisting of an action network, adaptive critic network, and model or
identification network as show in Figure 5. These methods are able to control processes
in such a way, which is approximately optimal with respect to any given criteria taking
into consideration of particular nonlinear environment. For instance, when searching
for an optimal trajectory to the target position, the distance of the robot from this target
position can be used as a criteria function. The algorithm will compute the proper
106 Mobile Robotics, moving intelligence

steering, acceleration signals for control of vehicle, and the resulting trajectory of the
vehicle will be close to optimal. During trials (the number depends on the problem and
the algorithm used) the system will improve performance and the resulting trajectory
will be close to optimal. The freedom of choice of the criteria function makes the
method applicable to a variety of problems. The ability to derive a control strategy only
from trial/error experience makes the system capable of semantic closure. These are
very strong advantages of this method.

Fig. 5. Structure of the adaptive critic controller (Liao et al., 2003).

7.3 Creative Learning Structure

It is assumed that we can use a kinematic model of a mobile robot to provide a simulated
experience to construct a value function in the critic network and to design a kinematic
based controller for the action network. A proposed diagram of creative learning algorithm
is shown in Figure 6. In this proposed diagram, there are six important components: the task
control center, the dynamic knowledge database, the critic network, the action network, the
model-based action and the utility funtion. Both the critic network and action network can
be constructed by using any artificial neural networks with sigmoidal function or radial
basis function (RBF). Furthermore, the kinematic model is also used to construct a model-
based action in the framework of adaptive critic-action approach. In this algorithm, dynamic
databases are built to generalize the critic network and its training process and provide
evironmental information for decision making. It is especially critical when the operation of
mobile robots is in unstructured environments. Furthermore, the dynamic databases can
also used to store environmental parameters such as Global Position System (GPS) way
points, map information, etc. Another component in the diagram is the utility function for a
tracking problem (error measurement). In the diagram, Xk, Xkd, Xkd+1 are inputs and Y is the
ouput and J(t), J(t+1) is the critic function with respect to time.
Mobile Robotics, Moving Intelligence 107

Criteria Adaptive critic learning

filters system

…
J(t+1) Xdk+1
Critic n

Critic 2 γ
- Utility
Dynamic Critic Network function
(Critic)
Knowledge - -
Database Critic 1 J(t)

Xdk+1
Action Y
Model-
Network based
Action
Xk
Xk Xdk
-
Task Z-1
Control
Center

Fig. 6. Proposed creative learning algorithm structure.

7.4 Dynamic Knowledge Database (DKD)

The dynamic databases contain domain knowledge and can be modified to permit adaptation to
a changing environment. Dynamic knowledge databases may be called a “neurointerface” in a
dynamic filtering system based on neural networks (NNs) and serves as a “coupler” between a
task control center and a nonlinear system or plant that is to be controlled or directed. The
purpose of the coupler is to provide the criteria function for the adaptive critic learning system
and filter the task strategies commanded by the task control center. The proposed dynamic
database contains a copy of the model (or identification). Action and critic networks are utilized
to control the plant under nominal operation, as well as make copies of a set of parameters (or
scenario) previously adapted to deal with a plant in a known dynamic environment. The
database also stores copies of all the partial derivatives required when updating the neural
networks using backpropagation through time13 . The dynamic database can be expanded to
meet the requirements of complex and unstructured environments.
The data stored in the dynamic database can be uploaded to support offline or online
training of the dynamic plant and provide a model for identification of nonlinear dynamic
environment with its modeling function. Another function module of the database
management is designed to analyze the data stored in the database including the sub-task
optima, pre-existing models of the network and newly added models. The task program
module is used to communicate with the task control center. The functional structure of the
proposed database management system (DBMS) is shown in Figure 7. The DBMS can be
customized from an object-relational database.
108 Mobile Robotics, moving intelligence

Modeling …

Task
Control
Dynamic Adaptive
Database Critic

…
Task
Analysis Program

Fig. 7. Functional structure of dynamic database.

7.5 Task Control Center (TCC)

The task control center (TCC) can build task-level control systems for the creative learning
system. By "task-level", we mean the integration and coordination of perception, planning
and real-time control to achieve a given set of goals (tasks). TCC provides a general task
control framework, and it is to be used to control a wide variety of tasks. Although the TCC
has no built-in control functions for particular tasks (such as robot path planning
algorithms), it provides control functions, such as task decomposition, monitoring, and
resource management, that are common to many applications. The particular task built-in
rules or criteria or learning J functions are managed by the dynamic database controlled
with TCC to handle the allocation of resources. The dynamic database matches the
constraints on a particular control scheme, sub-tasks or environment allocated by TCC.
The task control center acts as a decision-making system. It integrates domain knowledge or
criteria into the database of the adaptive learning system. The task control architecture for
mobile robots provides a variety of control constructs that are commonly needed in mobile
robot applications, and other autonomous mobile systems. The goal of the architecture is to
enable autonomous mobile robot systems to easily specify hierarchical task-decomposition
strategies, such as how to navigate to a particular location, or how to collect a desired
sample, or how to follow a track in an unstructured environment. This can include temporal
constraints between sub-goals, leading to a variety of sequential or concurrent behaviors.
TCC schedules the execution of planned behaviors, based on those temporal constraints
acting as a decision-making control center.
Integrating the TCC with the adaptive critic learning system and interacting with the
dynamic database, the creative learning system provides both task-level and real-time
control or learning within a single architectural framework. Through interaction with
human beings to attain the input information for the system, the TCC could decompose the
task strategies to match the dynamic database for the rules of sub-tasks by constructing a
distributed system with flexible mechanisms, which automatically provide the right data at
the right time. The TCC also provides orderly access to the resources of the dynamic
database with built-in learning mechanisms according to a queue mechanism. This is the
inter-process communication capability between the task control center and the dynamic
database. The algorithm on how to link the task control center and the dynamic database is
currently done by the human designers.

7.6 Creative Learning Controller for Intelligent Robot Control

Creative learning may be used to permit exploration of complex and unpredictable
environments, and even permit the discovery of unknown problems, ones that are not yet
Mobile Robotics, Moving Intelligence 109

recognized but may be critical to survival or success. By learning the domain knowledge,
the system should be able to obtain the global optima and escape local optima. The
method attempts to generalize the highest level of human learning – imagination. As a
ANN robot controller, the block diagram of the creative controller can be presented as
shown in Figure 8.
Experience with the guidance of a mobile robot has motivated this study and has progressed
from simple line following to the more complex navigation and control in an unstructured
environment. The purpose of this system is to better understand the adaptive critic learning
theory and move forward to develop more human-intelligence-like components into the
intelligent robot controller. Moreover, it should extend to other applications. Eventually,
integrating a criteria knowledge database into the action module will develop a powerful
adaptive critic learning module.

Creative
Controller

Primary τ
Yd Controller +
Y
Robot
+ +
Secondary
Controller
--

Sensors

Fig. 8. Block diagram of creative controller.

A creative controller is designed to integrate domain knowledge or criteria database and the
task control center into the adaptive critic neural network controller. It provides a needed
and well-defined structure for autonomous mobile robot application. In effect, it replaces a
human doing remote control. We have used the intelligent mobile robot as the test-bed for
the creative controller.
The task control center of the creative learning system can be considered hierarchically as
follows:
” Mission for robot – e.g. mobile robot
” Task for robot to follow – J : task control
” Track for robot to follow
” Learn non-linear system model- model discovery
” Learn unknown parameters

7.7 Adaptive Critic System Implementation

Adaptive critic system and NN. In order to develop the creative learning algorithm
addressed above, we have taken a bottom-up approach to implement adaptive critic
controllers by first using neural network for on-line or off-line learning methods. (Campos
and Lewis, 1999) Then the proposed dynamic knowledge database and task control center
are added with some to be realized in future research projects.
110 Mobile Robotics, moving intelligence

Tuning algorithm and stability analysis. For linear time invariant systems it is straightforward
to examine stability by investigating the poles in the s-plane. However, stability of nonlinear
dynamic systems is much more complex, thus the stability criteria and tests are much more
difficult to apply than those for linear time invariant systems (Stubberud, A.R. and S.C.
Stubberud, (2000)). For general nonlinear continuous time systems, the state space model is

x& = f [ x (t ), u (t )]
y = g[ x (t ), u (t )] (8)

where the nonlinear differential equation is in state variable form, x(t) is the state vector and
u(t) is the input and the second equation y(t) is the output of the system.

7.8 Creative Controller and Nonlinear Dynamic System

For a creative controller, the task control center and the dynamic database are not time-variable
systems; therefore, the adaptive critic learning component determines the stability of the creative
controller. As it is discussed in the previous section, the adaptive critic learning is based on critic
and action network designs, which are originated from artificial neural network (ANN), thus
stability of the system is determined by the stability of the neural networks (NN) or convergence
of the critic network and action network training procedure.
The creative controller is a nonlinear system. It is not realistic to explore all the possibilities of
nonlinear systems and prove that the controller is in a stable state. We have used both robot arm
manipulators and mobile robot models to examine a large class of problems known as tracking in
this study. The objective of tracking is to follow a reference trajectory as closely as possible. This
may also be called optimal control since we optimize the tracking error over time.

7.9 Critic and Action NN Weights Tuning Algorithm

In adaptive critic learning controller, both the critic network and the action network use
multilayer NN. Multilayer NN are nonlinear in the weights V and so weight tuning
algorithms that yield guaranteed stability and bounded weights in closed-loop feedback
systems have been difficult to discover until a few years ago.

7.10 Urban Rescue Scenarios

Suppose a mobile robot is used for urban rescue as shown in Figure 9. It waits at a start
location until a call is received from a command center. Then it must go rescue a person.
Since it is in an urban environment, it must use the established roadways. Along the
roadways, it can follow pathways. However, at intersections, it must choose between
various paths to go to the next block. Therefore, it must use different criteria at the corners
than along the track. The overall goal is to arrive at the rescue site in minimum time. To
clarify the situation consider the following steps:
1. Start location – the robot waits at this location until it receives a task command to
go to a certain location.
2. Along the path, the robot follows a road marked by lanes. It can use a
minimum mean square error between its location and the lane location during
this travel.
3. At intersections, the lanes disappear but a database gives a GPS waypoint and the
location of the rescue goal.
Mobile Robotics, Moving Intelligence 111

This example requires the use of both continuous and discrete tracking, a database of known
information and multiple criteria optimization. It is possible to add a large number of real-world
issues including position estimation, perception, obstacles avoidance, communication, etc.

E F T Destination

J2 G
C B

S
Error

Start J1 A D

Fig. 9. Simple urban rescue site.

In an unstructured environment as shown in Figure 9, we assume that information collected
about different positions of the environment could be available to the mobile robot,
improving its overall knowledge. As any robot moving autonomously in this environment
must have some mechanism for identifying the terrain and estimating the safety of the
movement between regions (blocks), it is appropriate for a coordination system to assume
that both local obstacle avoidance and a map-building module are available for the robot
which is to be controlled. The most important module in this system is the adaptive system
to learn about the environment and direct the robot action.
A Global Position System (GPS) may be used to measure the robot position and the distance
from the current site to the destination and provide this information to the controller to make its
decision on what to do at next move. The GPS system or other sensors could also provides the
coordinates of the obstacles for the learning module to learn the map, and then aid in avoiding
the obstacles when navigating through the intersections A, B or G, D to destination T.
Task control center. The task control center (TCC) acts a decision-making command center. It
takes environmental perception information from sensors and other inputs to the creative
controller and derives the criteria functions. We can decompose the robot mission at the urban
rescue site shown as Figure 9 into sub-tasks as shown in Figure 10. Moving the robot between the
intersections, making decisions is based on control-center-specified criteria functions to minimize
the cost of mission. It’s appropriate to assume that J1 and J2 are the criteria functions that the task
control center will transfer to the learning system at the beginning of the mission from the Start
point to Destination (T). J1 is a function of t related to tracking error. J2 is to minimize the
distance of the robot from A to T since the cost is directly related to the distance the robot travels.
• From Start (S) to intersection A: robot follow the track SA with the J1 as objective
function
112 Mobile Robotics, moving intelligence

• From intersection A to B or D: which one will be the next intersection, the control
center takes both J1 and J2 as objective functions.

Urban

Follow a track

Local Navigating
Navigating to A

Fig. 10. Mission decomposition diagram.

Dynamic databases. Dynamic databases would store task-oriented environment knowledge,
adaptive critic learning parameters and other related information for accomplishing the mission.
In this scenario, the robot is commanded to reach a dangerous site to conduct a rescue task. The
dynamic databases saved a copy of the GPS weight points S, A, B, C, D, E, F, G and T. The map
for direction and possible obstacle information is also stored in the dynamic databases. A copy of
the model parameters can be saved in the dynamic database as shown in the simplified database
Figure 11. The action model will be updated in the dynamic database if the current training
results are significantly superior to the previous model stored in the database.
Database fields
Field Description
MODEL_ID Action model ID
MODEL_NAME Action model name
UTILITY_FUN Utility function
CRITERIA_FUN Criteria function
… …
Adaptive Critic Training Parameters
INPUT_CRITIC Input to critic network
DELT_J J(t+1)-J(t)
… …
Fig. 11. Semantic dynamic database structure.

Robot learning module. Initial plans such as road tracking and robot navigating based on
known and assumed information can be used to incrementally revise the plan as new
information is discovered about the environment. The control center will create criteria
functions according to the revised information of the world through the user interface.
These criteria functions along with other model information of the environment will be
input to the learning system. There is a data transfer module from the control center to the
learning system as well as a module from the learning system to the dynamic database. New
knowledge is used to explore and learn, training according to the knowledge database
information and then decide which to store in the dynamic database and how to switch the
criteria. The simplest style in the adaptive critic family is heuristic dynamic programming
(HDP). This is NN on-line adaptive critic learning. There is one critic network, one action
network and one model network in the learning structure. U(t) is the utility function. R is
Mobile Robotics, Moving Intelligence 113

the critic signal as J (criteria function). The learning structure and the parameters are saved a
copy in the dynamic database for the system model searching and updating.

8. Conclusion
A new concept for mobile intelligence is presented in this paper. A series of experimental robots
named the Bearcats, have been constructed at the University of Cincinnati over the past several
years. This experience has evolved into our current, creative control design that is presented in
this paper. Fortunately, our intelligent robots have been able to use our increasingly capable
computer controls in which multi-threaded, distributed computing is now easily available.
The theory presented shows how to design intelligent robots that are capable of adapting, learning
and predicting. This is a step toward understanding the semiotic closure exhibited by biological
creatures and a further step toward appreciating the wonderful capabilities of human intelligence.

9. References
Alhaj Ali, S. M. (2003). Technologies for autonomous navigation in unstructured outdoor
environments, University of Cincinnati, Ohio.
Baker, N. C.; MacKenzie, D. C. & Ingalls, S. A. (1992). Development of an autonomous aerial
vehicle: A case study. Applied Intelligence, Vol. 2, No. 3, (1992) 271-298.
Baratoff, G.; Toepfer, C. & Neumann, H. (2000). Combined space-variant maps for optical-
flow-based navigation. Biological Cybernetics, Vol. 83, Issue 3, (Aug. 2000) 199-209.
Bellman, R.E. ( 1957) Dynamic Programming, Princeton University Press, Princeton, N.J.
Campos, J. and F.L. Lewis. (1999) Adaptive Critic Neural Network for Feedforward
Compensation. Proc. of American Control Conference.
Chen, K. H. & Tsai, W. H. (2000). Vision-based obstacle detection and avoidance for
autonomous land vehicle navigation in outdoor roads. Automation in Construction,
Vol. 10, Issue 1, (Nov. 2000) 1-25.
Choi, S. K. (1993). Design of advanced underwater robotic vehicle and graphic workstation.
IEEE Int. Conf. on Robotics and Automation, pp. 99-105, Atlanta, GA, 1993.
Choi, S. K.; Yuh, J. & Takashige, G. Y. (1995). Development of the omnidirectional intelligent
navigator. IEEE Robotics & Automation Magazine, Vol. 2, No. 1, (1995) 44-51.
Cicirelli, G.; Distante, D.; Orazio, T. D. & Attolico, G. (1998). Learning actions from vision-
based positioning in goal-directed navigation, Proceedings of the 1998 IEEE.RSJ
International Conference on Intelligent Robots and Systems, Vo. 3, pp. 1715-1720, 1998.
Corradini, M. L.; Leo, T. & Orlando, G. (2002). Experimental testing of a discrete-time sliding
mode controller for trajectory tracking of a wheeled mobile robot in the presence of
skidding effects. Journal of Robotic Systems, Vol. 19, Issue 4, (April 2002) 177-188.
Corradini, M. L. & Orlando, G. (2001). Robust tracking control of mobile robots in the
presence of uncertainties in the dynamical model. Journal of Robotic Systems, Vol. 18,
Issue 6, (June 2001) 317-323.
Daxwanger, W. A. & Schmidt, G. (1998). Neuro-fuzzy posture estimation for visual vehicle
guidance, Proceedings of the 1998 IEEE International Joint Conference on Neural
Networks, Vol. 3, pp. 2086-2091, 1998.
Devy, M.; Chatila, R. ; Fillatreau, P.; Lacroix, S. & Nashashibi, F. (). On autonomous
navigation in a natural environment. Robotics and Autonomous Systems, Vol. 16,
(Nov. 1995) 5-16.
114 Mobile Robotics, moving intelligence

de Wit, C. C.; Siciliano, B. & Bastin, G. (1996). Theory of Robot Control. Springer-Verlag
London Limited, Britain.
Dong, W.; Liang Xu, W. & Huo, W. (1999). Trajectory tracking control of dynamic non-
holonomic systems with unknown dynamics. International Journal of Robust and
Nonlinear Control, Vol. 9, Issue 13, (November 1999) 905-922.
Dracopoulos, D. C. (1998). Robot path planning for maze navigation, Proceedings of the
1998 IEEE International Joint Conference on Neural Networks, Vol. 3, pp. 2081-
2085, 1998.
Dudek, G. & Jenkin, M. (2000). Computational Principles of Mobile Robotics, Cambridge
University Press, ISBN 0-521-56876-5, United Kingdom.
Feruson, J. & Pope, A. (1995). THESEUS: Multipurpose Canadian AUV, Sea Technology Magazine.
Fuji, T.; Tamaki, U. & Kuroda, Y. (1993). Mission execution experiments with a newly
developed AUV 'The Twin-Burger', Proc. 8th Int. Symp. Unmanned Untethered
Submersible Technology, pp. 92-105, Durham, NC, 1993.
Ghaffari, M., Liao, X., Hall, E. (2004) A Model for the Natural Language Perception-based
Creative Control of Unmanned Ground Vehicles, Proceedings of the SPIE 2004
Conference.
Golnazarian, W. & Hall, E. L. (2000). Intelligent Industrial Robots, In: Handbook of Industrial
Automation, 499-527, Marcel Dekker, Inc., New York.
Greenhouse, L. & Norris, J. (2002). Control system architecture for military robotics. The NIST
Virtual Timeline, Information Services Division, National Institute of Standards
and Technology. Available online: http://museum.nist.gov/exhibits/timeline/
printerFriendly.cfm? itemId=38.
Grudic G. Z. & Lawrence, P. D. (1998). Nonparametric learning approach to vision based
mobile robot localization, Proceedings of the 1998 IEEE/RSJ International Conference on
Intelligent Robots and Systems, Vol. 2, pp. 724-729, 1998.
Honda, (2005). Honda Debuts New ASIMO, Honda Motor Co., Ltd, Available online:
http://world.honda.com/news/2005/c051213.html.
Janet, J. A.; Luo, R. C. & Kay, M. G. (1997). Autonomous mobile robot global motion
planning and geometric beacon collection using traversability vectors. IEEE
Transactions on Robotics and Automation, Vol. 13, No. 1, (Feb. 1997) 132-140.
Jarvis, R. (1996). An all-terrain intelligent autonomous vehicle with sensor-fusion-based
navigation capabilities. Control Eng. Practice, Vol. 4, No. 4, (April 1996) 481-486.
Jiang, Z.-P.; Lefeber, E. & Nijmeijer, H. (2001). Saturated stabilization and tracking of a
nonholonomic mobile robot. Systems and Control Letters, Vol. 42, Issue 5, (April
2001) 327-332.
Kim, K. H. C. & Hyung, S. (1998). Mobile robot navigation based on optimal via-point
selection method, Proceeding of the 1998 IEEE/RSJ International Conference on
Intelligent Robots and Systems, Vol. 2, pp. 1242-1247, 1998.
Koh, K. C. & Cho, H. S. (1999). A smooth path tracking algorithm for wheeled mobile robots
with dynamic constraints. Journal of Intelligent and Robotic Systems, Vol. 24, Issue 4,
(April 1999) 367-385.
Kortenkamp, D.; Bonasso, R. P. & Murphy, R. (1998). Artificial Intelligence and Mobile Robots,
AAAI Press/The MIT Press, ISBN 0-262-61137-6, USA.
Krishna K. M. & Kalra, P. M. (2001). Perception and remembrance of the environment
during real-time navigation of a mobile robot. Robotics and Autonomous Systems,
Vol. 37, (Oct. 2001.) 25-51.
Mobile Robotics, Moving Intelligence 115

Kurazume, R. & Hirose, S. (2000). Development of a cleaning robot system with cooperative
positioning system. Autonomous Robots, Vol. 9, Issue 3, (Dec. 2000) 237-246.
Kurd, S. & Oguchi, K. (1997). Neural network control for automatic guided vehicles using
discrete reference markers, Industry Applications Society, Thirty-Second IAS Annual
Meeting, Vol. 2, pp. 886-891, 1997.
Lewis, F. L.; Jagannathan, S. & Yesildirek, A. (1999). Neural Network Control of Robot Manipulators
and Nonlinear Systems, Taylor and Francis Ltd, T. J. International Ltd, Padstow, UK.
Lewis, M. A.; Fagg, A. H. & Bekey, G. A. (1993). The USC autonomous flying vehicle: An
experiment in real-time behavior-based control. Proc. IEEE/RSJ IROS, Yokohama,
Japan, 1993.
Lin, C. H. & Wang, L. L. (1997). Intelligent collision avoidance by fuzzy logic control.
Robotics and Autonomous Systems, Vol. 20, (April 1997) 61-83.
Liao, X. and Hall.E. (2002) Beyond Adaptive Critic - Creative Learning for Intelligent
Autonomous Mobile Robots, Proceedings of Intelligent Engineering Systems Through
Artificial Neural Networks, ANNIE, in Cooperation with the IEEE Neural Network
Council, St. Louis - Missouri.
Liao, X., et al. (2003) Creative Control for Intelligent Autonomous Mobile Robots, Proceedings
of Intelligent Engineering Systems Through Artificial Neural Networks, ANNIE, in
Cooperation with the IEEE Neural Network Council, St. Louis - Missouri.
Marco, D. B.; Healey, A. J. & McGhee, R. B. (1996). Autonomous underwater vehicles:
Hybrid control of mission and motion. Autonomous Robots, Vol. 3, Issue 2, (May
1996) 169-186.
Martinez, E. & Torras, C. (2001). Qualitative vision for the guidance of legged robots in
unstructured environments. Pattern Recognition, Vol. 34, Issue 8, (August 2001)
1585-1599.
Meystel, A. M. & Albus, J. S. (2002). Intelligent Systems: Architecture, Design, and Control, John
Wiley & Sons, Inc., New York.
Montaner, M. B. & Ramirez-Serrano, A. (1998). Fuzzy knowledge-based controller design for
autonomous robot navigation. Expert Systems with Applications, Vol. 14, Issue 1-2,
(January 1998) 179-186.
Mora, T. E. & Sanchez, E. N. (1998). Fuzzy logic-based real-time navigation controller for a
mobile robot, Proceeding of the 1998 IEEE/RSJ International Conference on Intelligent
Robots and Systems, Vol. 1, pp. 612-617, 1998.
Premvuti, S. & Wang, J. (1996). Relative position localizing system for multiple autonomous
mobile robots in distributed robotic system: System design and simulation. Robotics
and Autonomous Systems, Vol. 18, Issue 3, (August 1996) 319-326.
Rajagopalan, R. & Barakat, N. (1997). Velocity control of wheeled mobile robots using
computed torque control and its performance for a differentially driven robot.
Journal of Robotic Systems, Vol. 14, Issue 4, (April 1997) 325-340.
Shim, H.-S. & Sung, Y.-G. (2003). Asymptotic control for wheeled mobile robots with driftless
constraints. Robotics and Autonomous Systems, Vol. 43, Issue 1, (April 2003) 29-37.
Space and Naval Warfare Systems Center, San Diego (SSC San Diego). (2002). Robotics at
Space and Naval Warfare Systems Center. The Space and Naval Warfare Systems
Center, San Diego. Available online: http://www.spawar.navy.mil/robots.
Steer, B.; Kloske, J.; Garner, P.; LeBlanc, L. & Schock, S. (1993). Towards sonar based
perception and modelling for unmanned untethered underwater vehicles, Proc.
IEEE Int. Conf. Robotics and Automation, Vol. 2, pp. 112-116, Atlanta, GA, 1993.
Stubberud, A.R. and Stubberud, S.C., (2000) Stability, in Handbook of Industrial Automation,
R.L. Shell and E.L. Hall, Editors. 2000, MARCEL DEKKER, INC.: New York.
 116 Mobile Robotics, moving intelligence

Toda, M.; Kitani, O.; Okamoto, T. & Torii, T. (1999). Navigation method for a mobile robot
via sonar-based crop row mapping and fuzzy logic control. Journal of Agricultural
Engineering Research, Vol. 72, Issue 4, (April 1999) 299-309.
Topalov, A. V.; Kim, J.-H. & Proychev, T. P. (1998). Fuzzy-net control of non-holonomic
mobile robot using evolutionary feedback-error-learning. Robotics and Autonomous
Systems, Vol. 23, Issue 3, (April 1998) 187-200.
Torras, C. (1995). Robot adaptivity. Robotics and Autonomous systems, Vol. 15, Issue 1-2, (July
1995) 11-23.
Vázquez, F. & Garcia, E. (1994). A local guidance method for low-cost mobile robots using
fuzzy logic. Annual Review in Automatic Programming, Vol. 19, (1994) 203-207.
Vercelli, G. & Morasso, P. (1998). Recognition and classification of path features with self-
organizing maps during reactive navigation, Proceeding of the 1998 IEEE/RSJ
International Conference on Intelligent Robots and Systems, Vol. 3, pp. 1437-1442, 1998.
Watanabe, K. L.; Izumi, K.; Tang, J. & Han, F. (1998). Rotational control of an omnidirectional mobile
robot using a fuzzy servo controller. Advanced Robotics, Vol. 12, Issue 3, (1998) 171-189.
Werbos, P. J. (1999). New designs for universal stability in classical adaptive control and
reinforcement learning, Proceedings of the International Joint Conference on Neural
Networks IJCNN, pp. 2292-2295, Vol. 4, USA, 1999.
Wijesoma, W.; Khaw, P. P. & Teoh, E. K. (1999). Control and navigation of an outdoor AGV
using fuzzy reasoning, Proceedings of IEEE/IEEJ/JSAI International Conference on
Intelligent Transportation Systems, pp. 544-549, 1999.
Wong, W. (2005). SCI-FI robots edge closer to reality. Electronic Design, Vol. 53, Issue 14,
(June 2005) 65-68, ISSN 0013-4872.
Wu, S.-F.; Mei, J.-S. & Niu, P.-Y. (2001). Path guidance and control of a guided wheeled
mobile robot. Control Engineering Practice, Vol. 9, Issue 1, (January 2001) 97-105.
Xue, H. & Cheung, E. (1996). Learning control for position tracking of active suspension of
high-speed AGV via neural network, Proceedings of 1996 IEEE Conference on
Emerging Technologies and Factory Automation, Vol. 2, pp. 523-527, 1996.
Yahja, A.; Sanjiv, S. & Stentz, A. (2000). An efficient on-line path planner for outdoor mobile
robots. Robotics and Autonomous systems, Vol. 32, Issue 2-3, (Aug. 2000) 129-143.
Yan, X.; lyengar, S. S. & Brener, N. E. (1997). An event driven integration reasoning scheme
for handling dynamic threats in an unstructured environment. Artificial Intelligence,
Vol. 95, Issue 1, (August 1997) 169-186.
Yu, H.; Chi, C.; Su, T. & Bi, Q. (2001). Hybrid evolutionary motion planning using follow
boundary repair for mobile robots. Journal of Systems Architecture, Vol. 47, Issue 7,
(July 2001) 635-647.
Yun-Su, H. & Yuta, S. (1996). Trajectory tracking control for navigation of the inverse
pendulum type self-contained mobile robot. Robotics and Autonomous Systems, Vol.
17, Issue 1-2, (April 1996) 65-80.
Zhang, M. & Hirschorn, R. M. (1997). Discontinuous feedback stabilization of nonholonomic
wheeled mobile robots. Dynamics and Control, Vol. 7, Issue 2, (April 1997) 155-169.
Zhang, Y.; Chung, J. H. & Velinsky, S. A. (2003). Variable structure control of a differentially steered
wheeled mobile robot. Journal of Intelligent and Robotic Systems, Vol. 36, Issue 3, 301-314.
Zhu, Z. Y.; Yang, S.; Shi, D. & Xu, G. (1998). Better road following by integrating omni-view
images and neural nets, Proceedings of the 1998 IEEE International Joint Conference on
Neural Networks, Vol. 2, pp. 974-979.

View publication stats