Beyond Visibility Limits: A DRL-Based Navigation Strategy for Unexpected Obstacles

2025-03-03

AI Summary

Deep reinforcement learning (DRL) navigation in dynamic low-visibility environments suffers from poor safety and frequent collisions due to distance-based reward designs. Method: We propose DRL-NSUO, a novel navigation strategy featuring: (i) LiDAR point-cloud change rate as the core perceptual signal for dynamic environment representation; (ii) a composite reward function constrained by environmental change rate, integrated with curriculum learning to adaptively balance safety and efficiency; and (iii) short-range feature preprocessing with dynamic weight adjustment to enhance sensitivity to nearby obstacles. Results: Evaluated in the BARN simulation framework, DRL-NSUO achieves 94.0% and 91.0% navigation success rates at linear velocities of 0.5 m/s and 1.0 m/s, respectively—significantly outperforming conventional obstacle inflation methods. Moreover, it substantially reduces collision rates in human-robot collaborative scenarios, demonstrating superior robustness and safety under dynamic, low-visibility conditions.

Technology Category

• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Search and Optimization: Learning to Search
• Planning, Routing, and Scheduling: Learning for Planning and Scheduling

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

Distance-based reward mechanisms in deep reinforcement learning (DRL) navigation systems suffer from critical safety limitations in dynamic environments, frequently resulting in collisions when visibility is restricted. We propose DRL-NSUO, a novel navigation strategy for unexpected obstacles that leverages the rate of change in LiDAR data as a dynamic environmental perception element. Our approach incorporates a composite reward function with environmental change rate constraints and dynamically adjusted weights through curriculum learning, enabling robots to autonomously balance between path efficiency and safety maximization. We enhance sensitivity to nearby obstacles by implementing short-range feature preprocessing of LiDAR data. Experimental results demonstrate that this method significantly improves both robot and pedestrian safety in complex scenarios compared to traditional DRL-based methods. When evaluated on the BARN navigation dataset, our method achieved superior performance with success rates of 94.0% at 0.5 m/s and 91.0% at 1.0 m/s, outperforming conservative obstacle expansion strategies. These results validate DRL-NSUO's enhanced practicality and safety for human-robot collaborative environments, including intelligent logistics applications.

Problem

• Improves robot navigation safety in dynamic environments
• Enhances obstacle detection with LiDAR data preprocessing
• Balances path efficiency and safety using composite rewards

Innovation

• DRL-NSUO uses LiDAR data rate change.
• Composite reward function with dynamic weights.
• Short-range LiDAR preprocessing enhances obstacle sensitivity.