PATH PLANNING FOR 2 PLANAR ROBOTS MOVING IN UNKNOWN ENVIRONMENT

The paper investigates the problem of planning a collision-free path for a 5-link closed chain, which represents two planar robot manipulators cooperately carrying a small object, from a start position to a target position in a two-dimensional (2-D) workspace that is filled with unknown stationary obstacles. It is assumed that 1) shapes of the obstacles are arbitrary, 2) positions of the obstacles are not known a priori, and 3) any point on the robots is subject to collision. In order to obtain obstacle information, sensors are used to explore the immediate surroundings of the robot arms. This problem is very practical and human heuristic are of little help, so that designing an algorithm that guarantees finding a collision-free path, if it exists, becomes an important task. In this approach, the carried object is modeled as a point. Subsequently, two robot arms and the straight line connecting the bases of the robots are modeled as a 5-link closed chain. The path of the closed chain is planned continuously, based on the arms' start position, target position, current position, and the sensory feedback. The general idea is to plan the motion of the closed chain in the configuration space of the closed chain. The configuration space is divided into two 2-D subspaces. In each 2-D subspace, a connected free space is denoted as a free region. Within each free region, Lumelsky's algorithm Bug2 is adopted to find the collision-free path for the closed chain. The connectivity between free regions is discussed. A systematic search method is used to determine if there is an unvisited free region when no collision-free path can be found in visited free regions. The proposed algorithm is nonheuristic and converges, i.e., it can either claim that a path is found or conclude that a collision-free path does not exist.