Feature-Sensitive Motion Planning

Motion planning consists of finding a sequence of motions for a robot (movable object) to move from a start configuration to a goal configuration without colliding with any obstacle. PRMs create a roadmap that paths can be extracted from. To create the roadmap, nodes are generated in the environment at random. Each node represents a possible position the robot can move to. Only feasible nodes are stored in the roadmap. Feasibility of nodes is usually determined through a collision detection test. The nodes in the roadmap are connected with local planners. One local planner that is currently used is Rotate At S. This planner rotates at a given value between configurations before connecting them. Rotating between connections helps to avoid collision. The original implementation of Rotate At S only allows one rotation between the pair of nodes. We implemented a method that will allow several rotations between connections. Several rotations between connections may increase the chance of finding connections. Feature-Sensitive Motion Planning is a method of motion planning that divides a given environment into sections and uses a planner that will create the best roadmap for that section. Once a roadmap is generated in each section they are combined to create one map. The methods currently available to connect roadmaps are the most expensive step in Feature-Sensitive Motion Planning. They can be optimized. One of these methods makes a list of nodes from each section. Each node on the list tries to make a connection with every node on the other list. We will create a method that will be faster and more efficient. It will generate a list of nodes for each section and try to connect it to the closest nodes from the other list. “Feature-Sensitive Motion Planning”, Horton et al. DMP, Parasol Lab, Texas A&M, Summer 2005 1