Risk-Aware Safe Feedback Motion Planning in Gaussian Splatting World
Huanxiang Wang, Guoxiang Zhao · 2024
This paper proposes a novel risk-aware feedback motion planner designed to guide robots safely to their target region in an environment cluttered by randomly shaped obstacles. Leveraging the capabilities of 3D Gaussian Splatting (GS) for environmental perception, the planner can effectively identify obstacles and maximize the clearance between them. Our plan-ner stands out for its computational efficiency and enhanced safety compared to conventional reconstruction-infused motion planners and they are confirmed by numerical experiments in complex environments.