Dyna-LfLH: Learning Agile Navigation in Dynamic Environments from Learned Hallucination

2024-03-25

arXiv.org

AI Summary

To address the poor generalization and data scarcity challenges of motion planners in dynamic, densely cluttered environments, this paper proposes a self-supervised motion planning learning framework. Methodologically, it introduces “learning-based hallucination” for dynamic obstacle modeling—novelly employing learnable latent variable distributions to physically plausible generate diverse obstacle trajectories, thereby automatically constructing high-difficulty training scenarios and overcoming the fundamental limitation of conventional learning-from-hindsight (LfH) approaches in handling moving obstacles. The framework integrates implicit modeling, an end-to-end planning network, and simulation-to-real co-training. Extensive evaluations on both simulated and real-world robotic platforms demonstrate that our method achieves up to a 25% improvement in task success rate over state-of-the-art learning-based and classical planners, while significantly enhancing navigation safety and agility.

Technology Category

• Intelligent Robots: Motion and Path Planning
• Planning, Routing, and Scheduling: Learning for Planning and Scheduling
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

This paper presents a self-supervised learning method to safely learn a motion planner for ground robots to navigate environments with dense and dynamic obstacles. When facing highly-cluttered, fast-moving, hard-to-predict obstacles, classical motion planners may not be able to keep up with limited onboard computation. For learning-based planners, high-quality demonstrations are difficult to acquire for imitation learning while reinforcement learning becomes inefficient due to the high probability of collision during exploration. To safely and efficiently provide training data, the Learning from Hallucination (LfH) approaches synthesize difficult navigation environments based on past successful navigation experiences in relatively easy or completely open ones, but unfortunately cannot address dynamic obstacles. In our new Dynamic Learning from Learned Hallucination (Dyna-LfLH), we design and learn a novel latent distribution and sample dynamic obstacles from it, so the generated training data can be used to learn a motion planner to navigate in dynamic environments. Dyna-LfLH is evaluated on a ground robot in both simulated and physical environments and achieves up to 25% better success rate compared to baselines.

Problem

• Training motion planners for dense dynamic obstacle navigation
• Overcoming learning inefficiencies in imitation and reinforcement learning
• Generating dynamic obstacles from learned latent distributions

Innovation

• Self-supervised training with dynamic obstacle hallucination
• Generating obstacles from learned latent distribution
• Improving navigation success rate in dynamic environments