We propose to use potential field based control for generating the navigation commands. The control system takes the potential gradient map generated by the mapping system and generates navigation commands that move the robot along the potential gradient. An inherent component of this system, then, is the robot localization system. We also plan to use dead-reckoning and artificial landmark recognition for robot localization purpose.
We have implemented low level potential field based control in simulation. The simulator takes a potential gradient map as its input and, with the aid of dead reckoning for robot localization, performs navigation from the start point to the goal. An example of a typical run of the simulator in a world with two walls is shown below.
Uri Benhanan, a visiting faculty member from Israel is working on automatic generation of potential gradient maps from overlayed potential maps and occupancy grids.
This page is maintained by Amit Singhal.
Last update: 11/17/97.