Learning-Based Motion Planning for Dynamic Environments: From Foundational Algorithms to Emerging Paradigms

This survey reviews learning-based motion planning for dynamic environments, focusing primarily on work from 2015 to 2025. It organizes methods by the role learning plays in the planning pipeline: direct policy learning, learning-augmented classical planning, hybrid planning, and training enhancement. The paper also examines observation representations, prediction uncertainty, interaction modeling, planner integration, safety constraints, and training strategies. It concludes by identifying challenges including the sim-to-real gap, safe and certifiable planning, dense crowd navigation, perception-planning coupling, and embodied AI. The abstract does not provide new benchmark results.