Catching Objects with a Robot Arm using Model Predictive Control

This paper presents a model predictive control (MPC)-based planning and control approach for catching objects in flight with a robotic arm. The core of the approach is to combine the three elementary tasks of the catching process, namely predicting the flight trajectory, determining the catching pose and the motion planning and control of the robot in one optimization problem. Thereto, a time-optimal problem formulation is chosen with additional robot-specific inequality constraints. Based on a parametric description of the flight parabola, terminal equality constraints are defined ensuring that the end effector position lies on the flight parabola with an orientation in tangential direction of the trajectory. The approach is successfully applied in simulation and experiments in real-time for a 7-degrees-of-freedom (DOF) robot arm with the nonlinear model predictive control toolbox GRAMPC.