Reactive trajectory planning for robotic manipulators in shared dynamic workspace

With the rise of collaborative robotics, the ability of a robotic manipulator to plan and adapt to dynamic and changing scenarios is taking on particular importance. Commonly used state-of-the-art motion planners are unable to satisfy such requirements, as they lack the necessary reactiveness. In this paper, we propose the adoption of nonlinear model predictive control (NMPC) to generate collision-free trajectories, which are then tracked by a low-level controller. We formulate the trajectory generation as an optimization problem in the NMPC framework, proposing a set of obstacle avoidance constraints that prioritise safety as a primary requirement. The proposed algorithm is integrated with a custom perception pipeline and tested in both simulations and on a real collaborative pick-and-place manipulation task, proving that our algorithm achieves reactive motion planning for robotic arms.