Page 168 - Autonomous Mobile Robots

152 Autonomous Mobile Robots

4.2 LANDMARK-BASED NAVIGATION
In a landmark-based navigation system, the robot relies on its onboard sensors
to detect and recognize landmarks in its environment to determine its position.
Thisnavigationsystemverymuchdependsonthekindofsensorsbeingused,the
types of landmarks, and the number of landmarks available. For instance,
Sugihara [9] used a single camera on a robot to detect the identical points in the
3
environment and then adopted an O(n lg n) algorithm to find the position and
orientation of the robot such that each ray pierces at least one of the n points in
the environment. An extended version was proposed in References 10 and 11,
respectively. The localization based on distinguishable landmarks in the envir-
onment has been researched in Reference 12, in which the localization error
varies depending on the configuration of landmarks. Apart from vision systems,
other sensors have been widely used in position estimation, including laser [3],
odometry [13], ultrasonic beacons [6], GPS [7], IR [12], and sonars [14]. Since
no sensor is perfect and landmarks may change, none of these approaches
is adequate for a mobile robot to operate autonomously in the real world.
A landmark-based navigation system needs the integration of multiple sensors
to achieve robustness and cope with uncertainties in both sensors and land-
mark positions. This motivates us to pursue a hybrid approach to the problem
by integrating multiple sensors and different kinds of landmarks in a unified
framework.
In general, the accuracy of the position estimation in a landmark-based
navigation system is affected by two major problems. The first problem is that
the navigation system cannot work well when landmarks accidentally change
their positions. If natural landmarks are used in the navigation process, their
positions must be prestored into the environment map so that it is possible for
a mobile robot to localize itself during its operation. The second problem is
that sensory measurements are noisy when the robot moves on an uneven floor
surface or changes the speed frequently. The accuracy of robot positioning
degrades gradually, and sometimes becomes unacceptable during a continuous
operation. Therefore, re-calibration is needed from time to time and it becomes
a burden for real-world applications.
To effectively solve these problems, we propose a novel landmark-based
navigation system that is able to:

• Initialize its position through triangulation when necessary
• Update its internal landmark model when the position of landmarks
is changed or new landmarks become available
• Localize the robot position by integrating data from odometry, laser
scanner, sonar, and vision
Figure 4.1 shows the block diagram of our navigation system that is able
to implement concurrent localization and map building automatically. It is

© 2006 by Taylor & Francis Group, LLC

FRANKL: “dk6033_c004” — 2006/3/31 — 16:42 — page 152 — #4