Feedback MPC for Torque-Controlled Legged Robots

Abstract

The computational power of mobile robots is currently insufficient to achieve torque level whole-body Model Predictive Control (MPC) at the update rates required for complex dynamic systems such as legged robots. This problem is commonly circumvented by using a fast tracking controller to compensate for model errors between updates. In this work, we show that the feedback policy from a Differential Dynamic Programming (DDP) based MPC algorithm is a viable alternative to bridge the gap between the low MPC update rate and the actuation command rate. We propose to augment the DDP approach with a relaxed barrier function to address inequality constraints arising from the friction cone. A frequency-dependent cost function is used to reduce the sensitivity to high-frequency model errors and actuator bandwidth limits. We demonstrate that our approach can find stable locomotion policies for the torque-controlled quadruped, ANYmal, both in simulation and on hardware.

Rolling in the Deep – Hybrid Locomotion for Wheeled-Legged Robots Using Online Trajectory Optimization

“…Fortunately, research in traditional legged locomotion offers solutions to bridge this gap. The quadrupedal robot ANYmal (without wheels) performs highly dynamic motions using MPC [16], [17] and TO [18], [19] approaches. Impressive results are shown by MIT Cheetah, which performs blind locomotion over stairs [20] and jumps onto a desk with the height of 0.76 m [21].…”

Perceptive Model Predictive Control for Continuous Mobile Manipulation

“…Modeling the tracking task as a quadratic cost instead of a constraint greatly improves the robustness against disturbances [5]. In this work, we take inspiration from Grandia et al and introduce soft inequality constraints with Relaxed Barrier Functions (RBF) to the MPC cost [10].…”

Context-aware robotic arm using fast embedded model predictive control

“…When dealing with constrained dynamical systems, model predictive control (MPC) is a highly capable control framework [10]. MPC can take into account nonlinear system dynamics, input and state constraints present in most mechanical and robotics systems, as well as obstacle/collision avoidance constraints [11], [12]. Another benefit of MPC is the ability to take into account the behaviour of other environmental sub-systems, such as human motion [13], [14].…”

“…In robotics, MPC has been used on applications such as UAVs, swarm formation, walking, humanoid and mobile robots, [11], [12], [17]. MPC for obstacle avoidance for manipulation tasks has been explored in [18], [19].…”