Phase-based Nonlinear Model Predictive Control for Humanoid Walking Stabilization with Single and Double Support Time Adjustments

2025-06-04

AI Summary

Bipedal robots struggle to jointly optimize single-support phase (step duration) and double-support phase (DSP) duration under external disturbances while maintaining balance. Method: This paper proposes a phase-driven nonlinear model predictive control (NMPC) framework that enables online co-optimization of step duration and DSP duration for the first time. The method integrates phase-parameterized modeling, zero-moment point (ZMP) dynamics constraints, and real-time gait re-planning—replacing conventional stepwise coordination or heuristic DSP adjustment with disturbance-aware, dynamic phase-adaptive re-planning. Contribution/Results: Simulation results demonstrate significantly improved disturbance rejection compared to heuristic and sequential coordination approaches. Real-robot experiments validate robust stability in challenging scenarios, including walking on compliant terrain and recovering from impulsive lateral push disturbances at stance. The framework achieves seamless, constraint-satisfying gait adaptation without pre-defined DSP heuristics, marking a departure from traditional bipedal locomotion control paradigms.

Technology Category

• Intelligent Robots: Motion and Path Planning
• Planning, Routing, and Scheduling: Replanning and Plan Repair
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Balance control for humanoid robots has been extensively studied to enable robots to navigate in real-world environments. However, balance controllers that explicitly optimize the durations of both the single support phase, also known as step timing, and the Double Support Phase (DSP) have not been widely explored due to the inherent nonlinearity of the associated optimization problem. Consequently, many recent approaches either ignore the DSP or adjust its duration based on heuristics or on linearization techniques that rely on sequential coordination of balance strategies. This study proposes a novel phase-based nonlinear Model Predictive Control (MPC) framework that simultaneously optimizes Zero Moment Point~(ZMP) modulation, step location, step timing, and DSP duration to maintain balance under external disturbances. In simulation, the proposed controller was compared with two state-of-the-art frameworks that rely on heuristics or sequential coordination of balance strategies under two scenarios: forward walking on terrain emulating compliant ground and external push recovery while walking in place. Overall, the findings suggest that the proposed method offers more flexible coordination of balance strategies than the sequential approach, and consistently outperforms the heuristic approach. The robustness and effectiveness of the proposed controller were also validated through experiments with a real humanoid robot.

Problem

• Optimizes ZMP modulation and step adjustments for humanoid balance
• Addresses nonlinearity in single and double support phase timing
• Improves robustness against external disturbances and compliant terrain

Innovation

• Phase-based nonlinear MPC for humanoid walking
• Simultaneous optimization of ZMP and step parameters
• Flexible coordination of balance strategies