Battery Pack Design and Transient Performance Modeling for High-Power Legged Robots

Abstract

Legged robotics has recently shifted toward advanced optimization-based control methods, such as Model Predictive Control (MPC), to generate agile and energy-efficient locomotion. By casting the control problem as an optimization task, robotic systems can account for complex robot dynamics and operational constraints, including joint limits and actuator capabilities. However, high-performance maneuvers also demand rigorous consideration of onboard battery constraints. This work presents an empirically derived lithium-ion battery model that captures transient voltage sag and time-dependent internal battery state, enabling more accurate prediction of feasible power delivery. Additionally, a custom high-power battery pack was designed to meet the power demands of the MIT Humanoid, emphasizing power density, safety, and maintainability. Although the work presented in this thesis does not integrate the battery model into a trajectory optimization framework, it establishes the foundation for future research that aims to couple battery and robot dynamics in robot control. Ultimately, this approach will facilitate safer and more capable legged robots by ensuring that planned trajectories respect both physical and electrochemical constraints.