Mikhail Katliar

nomagic

Tuesday, June 08, 2021, 11:00

online

In industrial applications, it is sometimes required to move a robotic manipulator through a sequence of waypoints in minimal time. We consider the case when the waypoints are known beforehand, but their sequence is determined by random events that happen after the motion has started. The optimization is performed for all possible waypoint sequences taking their probabilities into account. The problem is constrained by the position, velocity, and acceleration limits of robot joints. Extra constraints such as lateral gravito-inertial acceleration limit, velocity direction, and kinematic acceleration direction of the end effector can also be applied. The OCP is discretized by direct multiple shooting method, and numerically solved using CasADi and Ipopt. We demonstrate our method by calculating optimal trajectories for a realistic setting, and present a comparison between an optimized motion of the robot and a motion performed by linear robot moves. Finally, we discuss possible improvements of the method.