Motion Control of High-Dimensional Musculoskeletal System with Hierarchical Model-Based Planning
Keywords: Model predictive control, High-dimensional embodied system
Abstract: Controlling high-dimensional nonlinear systems presents significant challenges in biological and robotic applications due to the large state and action spaces. While deep reinforcement learning has emerged as the leading approach, it suffers from computationally-intensive and time-consuming, and are not scalable to wide varieties of tasks that each require significant manual tuning. This paper introduces Model Predictive Control with Morphology-aware Proportional Control (MPC$^2$), a novel hierarchical model-based algorithm that addresses these challenges. By integrating a sampling-based model predictive controller for target posture planning with a morphology-aware proportional controller for actuator coordination, our algorithm achieves stable movement control of a 700-actuator musculoskeletal model without training. We show that MPC$^2$ enables zero-shot high-dimensional motion control across diverse movement tasks, such as standing, walking on varying terrains, and sports motion imitation. It can be incorporated into optimal cost function design to automatically optimize the objective, reducing the reliance on traditional reward engineering methods. This work presents a major advancement in (near) real-time control for complex dynamical systems.
Primary Area: applications to robotics, autonomy, planning
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 4268
Loading