Unified Hierarchical MPC in Task Executing for Modular Manipulators across Diverse Morphologies

2025-08-19

AI Summary

Frequent controller retuning is required for modular manipulators due to their variable configurations. Method: This paper proposes a unified hierarchical model predictive control (HMPC) framework comprising two layers: (i) an upper layer that generates reference trajectories via state prediction and employs quadratic linearization (second-order Taylor expansion) to enhance kinematic modeling accuracy while preserving computational efficiency of linear models; and (ii) a lower layer that performs joint-space trajectory optimization, explicitly incorporating kinematic constraints and approximate dynamics. Crucially, the framework eliminates the need for configuration-specific controller parameter tuning. Results: Evaluated on multiple modular manipulator configurations, the HMPC framework achieves high-precision, robust, and smooth trajectory tracking in real-world pick-and-place tasks. It significantly improves control generality across diverse modular architectures and enhances deployment efficiency.

Technology Category

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

Application Category

• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

This work proposes a unified Hierarchical Model Predictive Control (H-MPC) for modular manipulators across various morphologies, as the controller can adapt to different configurations to execute the given task without extensive parameter tuning in the controller. The H-MPC divides the control process into two levels: a high-level MPC and a low-level MPC. The high-level MPC predicts future states and provides trajectory information, while the low-level MPC refines control actions by updating the predictive model based on this high-level information. This hierarchical structure allows for the integration of kinematic constraints and ensures smooth joint-space trajectories, even near singular configurations. Moreover, the low-level MPC incorporates secondary linearization by leveraging predictive information from the high-level MPC, effectively capturing the second-order Taylor expansion information of the kinematic model while still maintaining a linearized model formulation. This approach not only preserves the simplicity of a linear control model but also enhances the accuracy of the kinematic representation, thereby improving overall control precision and reliability. To validate the effectiveness of the control policy, we conduct extensive evaluations across different manipulator morphologies and demonstrate the execution of pick-and-place tasks in real-world scenarios.

Problem

• Adapting control for modular manipulators across diverse morphologies
• Executing tasks without extensive controller parameter tuning
• Maintaining precision near singular configurations and kinematic constraints

Innovation

• Hierarchical Model Predictive Control structure
• High-level MPC predicts future states
• Low-level MPC refines control actions