Safe Planning in Unknown Environments using Conformalized Semantic Maps

2025-09-29

AI Summary

Semantic planning in perception-uncertain, unknown environments—i.e., reaching a goal while maintaining category-aware distances from multiple unknown semantic regions—remains challenging. Existing approaches either neglect uncertainty or rely on known sensor models and noise priors. Method: We propose the first sensor-model-agnostic and distribution-free semantic obstacle-avoidance planner. It constructs a semantic map online and employs conformal prediction to quantify perception uncertainty in a model- and distribution-independent manner. This uncertainty quantification is rigorously integrated into a safety-critical planning framework, guaranteeing a user-specified lower bound on task success probability. Results: Experiments across diverse unknown environments demonstrate that our method consistently achieves the target success rate, significantly outperforming baselines. It combines theoretical soundness—via provable safety guarantees—with strong empirical robustness, enabling reliable deployment under unknown sensing conditions.

Technology Category

• Planning, Routing, and Scheduling: Planning under Uncertainty
• Reasoning under Uncertainty: Uncertainty Representations
• Machine Learning: Calibration & Uncertainty Quantification

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

This paper addresses semantic planning problems in unknown environments under perceptual uncertainty. The environment contains multiple unknown semantically labeled regions or objects, and the robot must reach desired locations while maintaining class-dependent distances from them. We aim to compute robot paths that complete such semantic reach-avoid tasks with user-defined probability despite uncertain perception. Existing planning algorithms either ignore perceptual uncertainty - thus lacking correctness guarantees - or assume known sensor models and noise characteristics. In contrast, we present the first planner for semantic reach-avoid tasks that achieves user-specified mission completion rates without requiring any knowledge of sensor models or noise. This is enabled by quantifying uncertainty in semantic maps - constructed on-the-fly from perceptual measurements - using conformal prediction in a model- and distribution-free manner. We validate our approach and the theoretical mission completion rates through extensive experiments, showing that it consistently outperforms baselines in mission success rates.

Problem

• Planning robot paths in unknown environments with semantic uncertainty
• Ensuring safe distances from unknown objects during navigation tasks
• Achieving mission completion guarantees without sensor model knowledge

Innovation

• Uses conformal prediction for uncertainty quantification
• Constructs semantic maps online from perceptual data
• Achieves user-specified mission completion rates