Survey Categorizes 138 Works on Robot Motion Planning in Dynamic Environments

A new comprehensive survey from Zongyuan Shen and eight co-authors (arXiv:2606.02677) systematically reviews 138 research papers primarily published between 2015 and 2025 on motion planning for robots operating in dynamic environments. Unlike many existing surveys that focus on static settings, this work tackles the critical challenge of real-time path adaptation as surroundings change. The authors categorize methods into five distinct groups based on core concepts: sampling-based planners, graph search algorithms, model predictive control (MPC), learning-based techniques (including supervised learning and reinforcement learning), and classical local planning approaches such as velocity obstacles, potential fields, and dynamic windows. Each category is analyzed for its principles, strengths, and limitations, with particular attention to the unique difficulties posed by moving obstacles and unpredictable agents.

Beyond path planning algorithms, the survey delves into the role of dynamic perception—how robots detect and model moving obstacles using cameras, LiDAR, and event-based sensors. It also highlights three major challenges that persist in dynamic environments: prediction uncertainty (anticipating where obstacles will go), human-robot interaction (navigating safely around people), and the freezing robot problem (when a robot becomes paralyzed by too many conflicting obstacles). By providing a structured taxonomy and detailed comparisons, this survey serves as a valuable roadmap for researchers and engineers looking to choose or develop motion planning systems for autonomous vehicles, drones, service robots, and other platforms that must operate safely amidst constant change.