Highly Dynamic Quadruped Locomotion via Whole-Body Impulse Control and Model Predictive Control (1909.06586v1)

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.

• The paper introduces a novel hybrid controller that prioritizes reaction force tracking via MPC and WBIC for dynamic quadruped locomotion.
• It demonstrates high performance with the Mini-Cheetah achieving speeds up to 3.7 m/s across challenging environments.
• The integrated approach effectively manages stability during rapid gait switches and under-actuated aerial phases.

Highly Dynamic Quadruped Locomotion via Whole-Body Impulse Control and Model Predictive Control

This paper presents a sophisticated approach to achieving dynamic locomotion in quadruped robots by integrating Whole-Body Control (WBC) with Model Predictive Control (MPC). The authors focus on the Mini-Cheetah robot to validate their proposed control framework, showcasing its potential in navigating complex environments with high-speed dynamics.

The principal challenge addressed by this research is the lack of robust control schemes to manage quadrupedal locomotion, especially during aerial phases, short stance times, and high-speed leg swings. The authors propose a unique controller architecture that gives precedence to reaction force commands rather than merely tracking body trajectory. This emphasis on force tracking allows for more effective regulation of high-speed dynamics, including handling significant periods of under-actuation as seen in aerial phases.

The control framework combines MPC to plan reaction forces over longer time horizons and WBIC to execute these plans by generating joint commands for torque, position, and velocity. The MPC employs a simple model to compute optimal reaction force profiles that consider gait types, speeds, and directions, while WBIC adjusts these to accommodate the robot's full dynamics and achieve body stabilization and swing leg control.

Key empirical performance outcomes include demonstrating various gaits in diverse environments, with the Mini-Cheetah achieving a top speed of 3.7 meters per second, indicative of high dynamism comparable to other leading quadruped robots. The control framework was tested across multiple terrains, evidencing robustness and adaptability.

One of the standout elements of this research is its versatility. The integration of MPC and WBIC not only supports a range of dynamic gaits but also simplifies gait switching through contact sequence adjustment. This versatility implies significant potential for adaptation to different locomotion tasks, including where gait transitions or changes are frequently required.

The paper also emphasizes the robustness of the controller in managing the robot's stability amidst disturbances, which is critical for applications needing reliable autonomous operations in dynamic or unpredictable environments. Such robustness may extend to scenarios involving additional tasks or constraints, like manipulation activities, which could be incorporated by minor controller modifications.

From a theoretical perspective, the integration of WBC with MPC resolves typical single-time-step limitations inherent in conventional WBC approaches, leveraging MPC’s longer horizon predictions. Practically, this method offers a promising solution for developers of autonomous legged robots seeking to enhance mobility and adaptability in non-structured settings.

Looking forward, this hybrid control framework presents a promising foundation for further advancements in AI-driven robotics. Future implementations could explore more complex manipulations or different robotic forms, such as biped locomotion, which can benefit from similar control dynamics. This work paves the way for broad exploration in robust, high-speed robotic locomotion and the autonomous operation of multi-legged robots across varying terrains and tasks.