Reactive planar nonprehensile manipulation with hybrid model predictive control

Abstract

This article presents an offline solution and online approximation to the hybrid control problem of planar non-prehensile manipulation. Hybrid dynamics and underactuation are key characteristics of this task that complicate the design of feedback controllers. We show that a model predictive control approach used in tandem with integer programming offers a powerful solution to capture the dynamic constraints associated with the friction cone as well as the hybrid nature of contact. We introduce the Model Predictive Controller with Learned Mode Scheduling (MPC-LMS), which leverages integer programming and machine learning techniques to effectively deal with the combinatorial complexity associated with determining sequences of contact modes. We validate the controller design through a numerical simulation study and with experiments on a planar manipulation setup using an industrial ABB IRB 120 robotic arm. Results show that the proposed algorithm achieves closed-loop tracking of a nominal trajectory by reasoning in real-time across multiple contact modalities.