[Paper Review] Highly Dynamic Quadruped Locomotion via Whole-Body Impulse Control and Model Predictive Control

Donghyun Kim, Jared Di Carlo+3 more|arXiv (Cornell University)|Sep 14, 2019

Robotic Locomotion and Control21 references216 citations

TL;DR

A hybrid MPC and Whole-Body Impulse Control (WBIC) framework enables highly dynamic quadruped locomotion with aerial phases, tested on MIT Mini-Cheetah to achieve 3.7 m/s running and multiple gaits.

ABSTRACT

Dynamic legged locomotion is a challenging topic because of the lack of established control schemes which can handle aerial phases, short stance times, and high-speed leg swings. In this paper, we propose a controller combining whole-body control (WBC) and model predictive control (MPC). In our framework, MPC finds an optimal reaction force profile over a longer time horizon with a simple model, and WBC computes joint torque, position, and velocity commands based on the reaction forces computed from MPC. Unlike existing WBCs, which attempt to track commanded body trajectories, our controller is focused more on the reaction force command, which allows it to accomplish high speed dynamic locomotion with aerial phases. The newly devised WBC is integrated with MPC and tested on the Mini-Cheetah quadruped robot. To demonstrate the robustness and versatility, the controller is tested on six different gaits in a number of different environments, including outdoors and on a treadmill, reaching a top speed of 3.7 m/s.

Motivation & Objective

• Develop a versatile control scheme for highly dynamic quadruped locomotion with aerial phases and short stance times.
• Enable switching between gaits by altering a high-level gait scheduler and footstep planning.
• Integrate model predictive control with a full-body impulse controller to handle underactuation and ground contact forces.
• Demonstrate robustness and versatility on real hardware across diverse terrains and gaits.

Proposed method

• Use Model Predictive Control (MPC) to compute optimal ground reaction force profiles over a gait cycle using a lumped-mass model.
• Develop Whole-Body Impulse Control (WBIC) to translate MPC forces into high-frequency joint position, velocity, and torque commands.
• Formulate a convex MPC by simplifying dynamics for tractable optimization and keeping a full-body dynamics model in WBIC.
• Employ a null-space projection based prioritized task execution within WBIC for body stabilization and swing foot control.
• Utilize a QP to refine reaction forces and floating-base accelerations while respecting contact constraints and friction cones.

Experimental results

Research questions

• RQ1 Can a two-layer MPC-WBIC control framework achieve and sustain highly dynamic running with aerial phases on a quadruped robot?
• RQ2 How does using MPC-provided reaction forces as targets in WBIC compare to tracking MPC-calculated CoM trajectories?
• RQ3 Can the approach handle multiple gaits and terrains in real hardware without task-specific tuning?
• RQ4 What throughput and stability can be achieved with the integrated controller on Mini-Cheetah hardware?
• RQ5 What are the effects of relaxing floating-base dynamics in the WBIC QP on performance during flight phases?

Key findings

• Achieved a top running speed of 3.7 m/s on Mini-Cheetah, with observed maximum 4 m/s but stability limits at 3.7 m/s.
• Demonstrated robustness across six different gaits in outdoor and treadmill settings.
• MPC runs at 30 Hz and WBIC at 500 Hz on the onboard computer, enabling high-frequency feedback.
• WBIC primarily tracks MPC-derived reaction forces rather than CoM trajectories to handle flights and underactuation.
• The controller fully utilizes hardware capabilities, with joint velocities and torques approaching actuator limits during high-speed running.