Time-optimal trajectory planning for sequential robotic tasks with unfixed endpoints

Time-optimal robot trajectory planning for point-to-point paths with fully specified endpoints has been investigated by some researchers. Some practical methods have been developed for finding time-optimal trajectories for point-to-point paths for general manipulators.;In this thesis, the existing methods are extended by considering a sequence of connected point-to-point paths, where the end-points are unfixed. The manifolds of the unfixed end-points are approximated by parametric hypersurfaces in joint space, defined in terms of a set of independent free parameters. For an n-joint robot, the dimension of the hypersurface ranges from n for completely unconstrained point (n free parameters) to zero for a fully specified point (no free parameters). The joint space of the robot is tessellated. The manifolds of unfixed endpoints are discretized into a number of points. The method for point-to-point problem is used to estimate the motion time along the path between each pair of end-points. Then, the dynamic programming method is used to find shortest path from initial end-point to final end-point and through every mid-point. The optimal path at this stage is then used as an initial guess for local optimization to find the globally optimal solution. The paths between endpoints are approximated by cubic B-splines. The optimization variables are the B-spline control points, and the free parameters specifying the path endpoints.;The algorithm has been implemented in a computer program in C for a two link robot. The near global time-optimal trajectories have been obtained. (Abstract shortened by UMI.).