Local next Step -path Planning for a Mobile Robot in an Initially Unknown Environment

A new path planning strategy is presented. It is called the local next step method and consists of an iterative process of calculating local targets, moving to them and then recalculating until the goal is reached. It is based on two concepts: the desirability of a direction of motion and the danger associated with that motion. A construct called the local tangent graph which is a tangent graph applied to only local environmental information is used to determine desirable directions. The robot is then constrained to move along this graph. The danger is based on uncertainty in the robot's sensor data and control. It is known how inaccurate robot control is, so using this information, it can be determined how likely it is that the robot will collide with an obstacle during its motion. The danger property is used together with desirability to determine the next point to which the robot should travel. On top of this, there is more global environmental information being used. A search is made of the environment to nd any cul de sacs which may trap the robot. If the robot nds itself inside one (either its initial position is inside one or it moves into one without realising because of incomplete environmental knowledge) then a cul de sac extrication behaviour takes over the robot and changes its priorities from nding the goal to nding an exit.