Legged Robot Control

One of ARCAD lab's research focuses is humanoid robot control, where we leverage the unique ability of these machines to navigate human-designed environments. As interest in humanoid robotics grows, we explore advanced control techniques, particularly model predictive control (MPC). Our research aims to synthesize and stabilize dynamic motions on physical humanoid platforms, pushing the boundaries of their capabilities and real-world applications.

Humanoid Related Publications:

• Orientation-aware model predictive control with footstep adaptation for dynamic humanoid walking

• Dynamic walking with footstep adaptation on the mit humanoid via linear model predictive control

• Humanoid Self-Collision Avoidance Using Whole-Body Control with Control Barrier Functions

Quadruped Related Publications:

• Representation-Free Model Predictive Control for Dynamic Motions in Quadrupeds

• qpSWIFT: A real-time sparse quadratic program solver for robotic applications

• Real-time model predictive control for versatile dynamic motions in quadrupedal robots

Custom ROBOT Hardware

Our lab has the capability to design and fabricate custom robot hardware that fulfills a unique set of specifications of legged robots. Building upon the Quasi-Direct-Drive (QDD) actuator, we design and integrate hardware, ranging from custom actuators to novel mechanisms, to push the boundaries of robot actuation technology. We aim to enhance the mobility, efficiency, and overall functionality of legged robots in diverse environments.

Hardware Related Publications:

• Design and development of the MIT humanoid: A dynamic and robust research platform

• Design and Experimental Implementation of a Quasi-Direct-Drive Leg for Optimized Jumping

• Hoppy: An open-source kit for education with dynamic legged robots

dynamic motion planning

Long horizon planning for highly dynamic motion is key to pushing the boundary of capapble legged robots in unstructured environments. This line of research focuses on combining advanced optimization and sampling methods to synthesize parkour motions that navigates complex terrain while respecting dynamical constraints of the robot.

Planning Related Publications:

• Hybrid Sampling/Optimization-based Planning for Agile Jumping Robots on Challenging Terrains

• Centroidal-momentum-based trajectory generation for legged locomotion

• Kinodynamic Motion Planning for Multi-Legged Robot Jumping via Mixed-Integer Convex Program

• Single Leg Dynamic Motion Planning with Mixed-Integer Convex Optimization