The goal of this research is to produce a general and versatile path planner that can be applied to a wide range of complex situations including both multiple redundant robots and wheeled vehicles subject to obstacle avoidance and other constraints. Currently, path planning is a human--intensive endeavor with long and arduous teaching of the robot path by a human operator in a simulated work environment. Besides the cost and quality considerations of such practice, planning for certain complex systems, such as multifingered robot hands or tractors with multiple trailers, are beyond the capability of most humans. The approach of this research project is to first pose the path planning problem as a finite-time nonlinear control problem, and then transform it to a static root-finding problem in an augmented space which can be iteratively solved. The approach will be extended to address issues related to convergence, penalty function selection, optimality, and robustness and sensitivity with respect to the kinematic model. The algorithm will be systematically applied, in both simulation and experimentation n, to a number of fairly complex case studies involving both redundant manipulators and wheeled vehicles which have traditionally been treated as different cases.
