TrajBooster: Boosting Humanoid Whole-Body Manipulation via Trajectory-Centric Learning

2025-09-15

AI Summary

Imitation learning suffers from insufficient robustness and generalization in long-horizon, high-precision whole-body manipulation for humanoid robots due to error accumulation. To address this, we propose KORR: a Koopman-operator-based framework that constructs a linear time-invariant latent space for global modeling of state evolution. Within this space, a residual policy performs closed-loop correction—overcoming the limitations of conventional local correction strategies. KORR integrates imitation learning initialization, residual policy optimization, and Koopman-dynamics-guided control, significantly enhancing trajectory stability and cross-task generalization. Evaluated on furniture assembly tasks, KORR achieves a +28.6% improvement in task success rate over baseline methods and demonstrates superior robustness to disturbances, particularly in extended-duration fine manipulation scenarios.

Technology Category

• Intelligent Robots: Manipulation
• Machine Learning: Imitation Learning & Inverse Reinforcement Learning
• Planning, Routing, and Scheduling: Replanning and Plan Repair

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems

Abstract

Imitation learning (IL) enables efficient skill acquisition from demonstrations but often struggles with long-horizon tasks and high-precision control due to compounding errors. Residual policy learning offers a promising, model-agnostic solution by refining a base policy through closed-loop corrections. However, existing approaches primarily focus on local corrections to the base policy, lacking a global understanding of state evolution, which limits robustness and generalization to unseen scenarios. To address this, we propose incorporating global dynamics modeling to guide residual policy updates. Specifically, we leverage Koopman operator theory to impose linear time-invariant structure in a learned latent space, enabling reliable state transitions and improved extrapolation for long-horizon prediction and unseen environments. We introduce KORR (Koopman-guided Online Residual Refinement), a simple yet effective framework that conditions residual corrections on Koopman-predicted latent states, enabling globally informed and stable action refinement. We evaluate KORR on long-horizon, fine-grained robotic furniture assembly tasks under various perturbations. Results demonstrate consistent gains in performance, robustness, and generalization over strong baselines. Our findings further highlight the potential of Koopman-based modeling to bridge modern learning methods with classical control theory. For more details, please refer to https://jiachengliu3.github.io/TrajBooster.

Problem

• Addresses imitation learning limitations in long-horizon manipulation tasks
• Improves robustness and generalization for unseen scenarios via global dynamics
• Enables precise whole-body control under perturbations using trajectory refinement

Innovation

• Uses Koopman operator for global dynamics modeling
• Guides residual policy with latent state predictions
• Enables stable long-horizon manipulation under perturbations