Unified Planning-Learning Framework for Robust UUV Navigation Under Partial Observability

2026-08-05

AI Summary

This work addresses the insufficient robustness and safety of unmanned underwater vehicle (UUV) navigation in dynamic underwater environments caused by limited perceptual capabilities. To tackle this challenge, we propose an observation-based autonomous navigation framework that relies solely on sonar and depth imagery. The framework integrates a persistent occupancy map, a global planner with clearance constraints, and a risk-aware reinforcement learning local controller. Innovatively, it adopts a planning–learning hybrid architecture: staged supervision is achieved through behavior tree distillation, while a teacher-guided reweighting mechanism calibrated by uncertainty enhances decision safety and training stability under partial observability. Evaluated in high-fidelity NVIDIA Isaac Sim simulations, our method significantly outperforms pure behavior tree and standard reinforcement learning baselines, demonstrating markedly improved navigation robustness and safety in dynamic underwater settings.

Abstract

This paper presents an observation-only autonomy framework for Unmanned Underwater Vehicles (UUVs) navigation in dynamic underwater environments that integrates persistent occupancy mapping, global clearance-aware planning, and risk-aware local control. The proposed pipeline constructs occupancy maps solely from onboard sonar and depth image observations, adapts a clearance-constrained global planner (GP) to provide long-horizon structure, and integrates a reinforcement learning (RL) policy to handle short-range tracking and reactive avoidance. To further support decision-making under partial observability, the system learns a compact latent state representation from onboard sensor data, encoding environmental structure, obstacle dynamics, and uncertainty. Behavior tree (BT) distillation with staged supervision is introduced to improve safety and training stability, while an uncertainty-calibrated distillation mechanism reweights teacher guidance using online latent-model uncertainty, emphasizing uncertain regimes during learning, with time-to-collision (TTC) and clearance cues remaining explicit in planning and local policy features. To demonstrate the efficacy of the framework, a reproducible multi-seed evaluation protocol is established in high-fidelity GPU-accelerated simulation using NVIDIA Isaac Sim, and performance is benchmarked against BT-only and standard RL baselines. The results obtained demonstrate improved robustness and safety under dynamic conditions, thus providing a general pipeline with a unified hybrid planning learning architecture and a reproducible methodology for robust UUV autonomy under partial observability.

Problem

• UUV navigation
• partial observability
• robust autonomy
• dynamic underwater environments
• sensor-based perception

Innovation

• partial observability
• unified planning-learning framework
• occupancy mapping
• risk-aware reinforcement learning
• behavior tree distillation