ASTECH
Adaptive and Safe Torque Estimation and Control for Humanoids

Project Description

Official name: Passivity-Preserving Adaptive Control for Safe and Compliant Humanoid Loco-Manipulation

Humanoid robots operating in narrow, cluttered, and evolving environments must combine precise manipulation and multi-contact locomotion with safe and compliant interaction. Unexpected contacts, modeling errors, friction, and flexible surfaces can generate large interaction forces or destabilize the robot, while simply reducing control gains improves compliance at the cost of precision. This project develops passivity-preserving adaptive whole-body control methods that reconcile these conflicting requirements. The proposed framework combines hierarchical optimization for dynamically feasible multi-objective motion generation with robust torque control whose convergence and compliance are adapted online while preserving passivity. Using Lyapunov-based analysis, the project aims to provide formal guarantees of stability and interaction safety even under external disturbances and compliant contacts, using primarily kinematic feedback from encoders and inertial sensors rather than joint torque sensing. The approach will explicitly address floating-base dynamics, balance, contact flexibility, and multi-contact locomotion, and will be validated in simulation and on humanoid platforms including HRP-5P, RHP-Friends, and H1. Representative experiments will combine precise manipulation, locomotion, support contacts, and disturbances in environments inspired by large-scale construction and assembly.

PI in JRL

Mehdi Benallegue

Period

2026.4 - 2029.3

Funded by

JSPS Grant-in-Aid for Scientific Research (C)