AutoPath: Learning Transferable Goal-Conditioned Stochastic Path Prior for Safe Navigation Without Human Demonstrations

2026-07-13

AI Summary

This work addresses the challenge of achieving real-time, demonstration-free, and cross-platform navigation in complex dynamic environments with limited perception, where path multimodality and platform heterogeneity pose significant obstacles. The authors propose a goal-conditioned stochastic path prior that resolves rotational ambiguity through normalized state representations, integrating a geometry-aware polar-coordinate action manifold with risk-sensitive utility shaping to enable structured path sampling and zero-shot transfer. Evaluated in densely cluttered scenes with both static and dynamic pedestrians, the method achieves high success rates and computational efficiency. Notably, a single navigation prior trained on a differential-drive robot is directly transferred—without fine-tuning—to a quadrupedal platform, demonstrating strong generalization and practical applicability across diverse robotic systems.

Abstract

Real-time navigation in cluttered and dynamic environments requires collision-free and dynamically feasible motion under limited perception. However, feasible navigation behaviors are inherently multimodal because multiple paths may exist around obstacles. In this paper, we formulate navigation as learning a transferable goal-conditioned stochastic path prior that models a reusable distribution over goal-aligned geometry-consistent local paths conditioned on local observations. This formulation enables structured sampling of navigation candidates, allowing multiple feasible paths to be explored through sampling without relying on robot-specific motion constraints. To this end, we introduce a goal-aligned canonical state representation that removes in-plane rotational ambiguity and normalizes local geometry with respect to the goal, enabling rotation-invariant path distribution learning. We further develop a structured prior learning framework that parameterizes local paths using a geometry-aware polar action manifold and incorporates risk-sensitive utility shaping with multi-goal distributional rollouts for stable and safety-aware planning. Extensive experiments in dense static environments and dynamic pedestrian scenarios demonstrate that the proposed method achieves consistently high success rates with competitive efficiency while enabling cross-platform transfer of a single path prior learned on differential-drive robots to quadruped platforms without retraining.

Problem

• safe navigation
• multimodal paths
• goal-conditioned planning
• transferable policy
• collision-free motion

Innovation

• goal-conditioned stochastic path prior
• rotation-invariant representation
• geometry-aware polar action manifold
• risk-sensitive planning
• cross-platform transfer