Optimal motion planning for robotic manipulators with dynamic obstacles using mixed-integer linear programming

The task of motion planning for robotic manipulators means to drive an end-effector between designated points in the work area while obstacles are not hit. This contribution investigates the case of dynamic obstacles (like human operators) and the consideration of a performance criterion to be maximized for the motion. The proposed approach maps the dynamics of the manipulator and the obstacles into the C × T-space (spanned by the configuration C and the time T). Within this space, an (sub-)optimal sequence of configurations in the collision-free subspace is determined by mixed-integer linear programming. To achieve sufficient computational efficiency, the optimization task is approached by employing the principles of model predictive control. The paper describes the approach based on the example of a two-link robot interacting with a human operator.