Page 343 - Autonomous Mobile Robots

332 Autonomous Mobile Robots

the extended Kalman filter (EKF) approach to SLAM is described. For such
feature-based approaches, the chapter outlines the essential operations that are
required for the successful use of SLAM in uncertain environments. These
operations include the ability to extract features from the raw sensor data for
inclusion into existing maps, to distinguish between new features and previ-
ously detected features which should be associated with known or previously
detected features, to be able to determine the robot’s location and correct erro-
neous map information in the presence of ever-increasing uncertainty (i.e., loop
closing and relocation techniques). The chapter also describes techniques to
handle the required operations for robust performance of SLAM on robots.
For the mapping of large environments, the method of local map joining is
described. The application of the local map joining method to multi-robot
mapping, where the relative locations between robots is not known, is also
presented.
The success of online robot localization and construction of maps paves
the way for more elaborate planning as the robot maneuvers through the
unknown environment. The area of path planning has been studied intens-
ively over the years, and is mainly concerned about the generation of a suitable
path to the goal, taking into account the obstacles present within the environ-
ment. Chapter 10, the second chapter of this part, discusses the incorporation
of internal constraints — namely kinematic, dynamic constraints, and vis-
ibility constraints — into motion planning. These additional constraints are
especially important in systems of embodied mobile robots. Following an over-
view of conventional classes of approaches to motion planning, the chapter
examines the use of randomized sampling techniques for motion planning of
robots subjected to kinematic and dynamic constraints. The effect of visibil-
ity constraints on motion planning, together with several solution techniques,
is investigated through the use of three representative visibility-based plan-
ning problems — guarding art galleries, online indoor exploration, and target
tracking.
The last chapter of the section examines cooperative motion planning and
control in multi-robot systems. This is a natural extension of single robot motion
planning, since autonomous systems are seldom made up of a single robot. The
planned motions of each robot will no longer be solely to obtain a collision
free path, but will also be shaped by the positions of other robots within the
team. The control of a robot’s path such that it maintains specific relative dis-
tances from others, relates to multi-robot formation control, and is treated in
detail within the chapter. Specifically, due to the prevalence of nonholonomic
robots in real-world applications (Part II), the chapter examines the formation
control and stability of teams of nonholonomic robots using formation con-
trol graphs, where different formations are achieved through the creation or
deletion of edges between robots. Optimization-based control of formations
is also investigated, with the focus on an off-line optimization process based

© 2006 by Taylor & Francis Group, LLC

FRANKL: “dk6033_c009” — 2006/3/31 — 16:43 — page 332 — #2