Fare: Failure Resilience in Learned Visual Navigation Control

2025-10-28

AI Summary

In visual navigation, imitation learning (IL) policies suffer from unpredictable failures under out-of-distribution (OOD) conditions and lack intrinsic capabilities for autonomous anomaly detection and recovery. This paper introduces Fare, the first unified framework integrating OOD-aware anomaly detection, fine-grained failure attribution, and adaptive recovery—without requiring explicit failure annotations. Its core contributions are: (1) a causally inspired, OOD-aware failure attribution module that provides interpretable localization of root causes (e.g., occlusion, abrupt illumination changes, map misalignment); and (2) a lightweight, plug-and-play recovery policy library compatible with diverse IL architectures. Extensive experiments across multiple challenging visual navigation benchmarks demonstrate that Fare significantly enhances long-horizon navigation robustness—improving failure detection accuracy by 32.7% and average recovery success rate by 41.5%. Moreover, it exhibits strong cross-model generalization in recovery performance.

Technology Category

• Machine Learning: Imitation Learning & Inverse Reinforcement Learning
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Computer Vision: Visual Reasoning & Symbolic Representations

Application Category

• Responsible Web: Human-perceived consequences of algorithmic deployment on the web
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

While imitation learning (IL) enables effective visual navigation, IL policies are prone to unpredictable failures in out-of-distribution (OOD) scenarios. We advance the notion of failure-resilient policies, which not only detect failures but also recover from them automatically. Failure recognition that identifies the factors causing failure is key to informing recovery: e.g. pinpointing image regions triggering failure detections can provide cues to guide recovery. We present Fare, a framework to construct failure-resilient IL policies, embedding OOD-detection and recognition in them without using explicit failure data, and pairing them with recovery heuristics. Real-world experiments show that Fare enables failure recovery across two different policy architectures, enabling robust long-range navigation in complex environments.

Problem

• Improving failure resilience in imitation learning policies
• Enabling automatic failure detection and recovery mechanisms
• Developing robust visual navigation in out-of-distribution scenarios

Innovation

• Embeds OOD detection and recognition without failure data
• Pairs policies with recovery heuristics for resilience
• Enables robust navigation across different policy architectures