This thesis is concerned with sensor-based motion planning problem for general robot-sensor systems, where an eye type of sensor, or a range sensor, is mounted on a robot with non-trivial geometry and kinematics. In this kind of system, unlike a simple mobile robot, where the robot can move and what it should sense have a very complex relationship. The next sensing action of the sensor should be planned to give maximum information of C-space, the robot's configuration space, instead of the knowledge of physical space the sensor actually scans.; The notion of C-space entropy was recently introduced as a measure of the knowledge of C-space for view planning problems of this kind [31, 34, 35]. The next best sensing action of the sensor is then the one that maximizes the expected C-space entropy reduction (MER criterion). Algorithms based on MER criterion for a point sensor showed great improvement over former simple physical space based geometrical strategies. (Abstract shortened by UMI.)
展开▼
