Survey of Learning-Based Motion Planning in Dynamic Environments (2015-2025)
There’s this new survey out on arXiv (2608.00625) that looks into how learning influences motion planning in changing environments, covering research from 2015 to 2025. It kicks off by revisiting some classic planning strategies and then introduces a classification system that sorts methods by how they incorporate learning, like direct policy learning or traditional methods that are improved with learning techniques. The findings are relevant to areas like autonomous vehicles, service robots, warehouse management, human-robot teamwork, crowd navigation, and systems with multiple robots. The survey aims to show how innovations in learning-based approaches enhance or interact with existing planning methods.
Key facts
- Survey published on arXiv with ID 2608.00625
- Covers works primarily from 2015 to 2025
- Focuses on motion planning in dynamic environments
- Proposes a role-of-learning taxonomy
- Taxonomy includes direct policy learning and learning-augmented classical methods
- Applications include autonomous driving, service robotics, warehouse logistics, human-robot collaboration, crowd navigation, and multi-robot systems
- Revisits classical planning methods as algorithmic foundations
- Announce type is cross
Entities
Institutions
- arXiv