GUARD: Toward a Compromise between Traditional Control and Learning for Safe Robot Systems

2025-09-27

AI Summary

Traditional control methods for robotic systems offer high safety guarantees but lack adaptability, whereas learning-based approaches provide flexibility yet lack formal safety assurances. Method: This paper proposes a risk-aware real-time decision-making framework that integrates reactive model predictive contour control (RMPCC), ICP registration, and probabilistic kernel-based active learning to jointly model uncertainty propagation and dynamic obstacle avoidance. Contribution/Results: Its key innovation lies in embedding uncertainty quantification directly into a reactive MPC architecture, enabling verifiable risk-averse behavior under ambiguous safety definitions. Experimental results demonstrate that the method achieves millisecond-level response times while significantly improving obstacle avoidance success rates and robustness in complex environments. It effectively balances real-time performance, safety assurance, and environmental adaptability, showing strong potential for practical deployment in safety-critical robotic applications.

Technology Category

• Intelligent Robots: Behavior Learning & Control
• Natural Language Processing: Safety and Robustness
• Reasoning under Uncertainty: Sequential Decision Making

Application Category

• 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
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

This paper presents the framework extbf{GUARD} ( extbf{G}uided robot control via extbf{U}ncertainty attribution and prob extbf{A}bilistic kernel optimization for extbf{R}isk-aware extbf{D}ecision making) that combines traditional control with an uncertainty-aware perception technique using active learning with real-time capability for safe robot collision avoidance. By doing so, this manuscript addresses the central challenge in robotics of finding a reasonable compromise between traditional methods and learning algorithms to foster the development of safe, yet efficient and flexible applications. By unifying a reactive model predictive countouring control (RMPCC) with an Iterative Closest Point (ICP) algorithm that enables the attribution of uncertainty sources online using active learning with real-time capability via a probabilistic kernel optimization technique, emph{GUARD} inherently handles the existing ambiguity of the term extit{safety} that exists in robotics literature. Experimental studies indicate the high performance of emph{GUARD}, thereby highlighting the relevance and need to broaden its applicability in future.

Problem

• Combining traditional control with learning for safe robot collision avoidance
• Addressing ambiguity in safety definitions within robotics applications
• Balancing traditional methods and learning algorithms for flexible systems

Innovation

• Combining traditional control with uncertainty-aware perception
• Using active learning for real-time collision avoidance
• Unifying reactive MPC with probabilistic kernel optimization