The problem with motorized wheelchairs is that they are large, clumsy and difficult to control. This is especially true if the driver has severely reduced capabilities. What we want is a wheelchair that can take instructionsfrom the driver and then based on its understanding of the environment, construct a plan that will take the user to the intended destination. The user should be able to sit in a room, tell the wheelchair that he wants to be in another room and the wheelchair should take him there as quickly and smoothly as possible.

The planner presented in this thesis uses a randomized bi-directional tree search. It builds two trees, one from the start state and one from the goal state by randomly sampling the control space of the robot. Each node is a state and each edge is a control input to the robot.

http://liu.diva-portalsmash/record.jsf?pid=diva2:19807