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

2026-08-01

AI Summary

This work addresses the challenges of robot motion planning in dynamic environments—particularly those posed by moving obstacles, prediction uncertainty, and multi-agent interactions—by providing a systematic survey of learning-driven planning approaches from 2015 to 2025. It introduces a classification framework centered on the “role of learning,” encompassing paradigms such as direct policy learning, learning-augmented classical planning, hybrid planning, and training-enhanced methods, thereby clarifying the diverse functions of learning within the planning pipeline. By integrating reinforcement learning, imitation learning, classical optimization, predictive models, and safety verification mechanisms, the study critically examines how observation representation, uncertainty handling, interaction modeling, and safety constraints shape method design. Beyond mapping the research landscape, it identifies key open challenges and future directions, including real-world transferability, certifiable safety, and navigation in dense crowds.

Abstract

Motion planning in dynamic environments is a fundamental problem in robotics, aiming to generate safe and efficient paths, trajectories, or control actions in the presence of moving obstacles, uncertain predictions, and multi-agent interactions. It has broad applications in autonomous driving, service robotics, warehouse logistics, human-robot collaboration, crowd navigation, and multi-robot systems. This survey reviews representative works published primarily between 2015 and 2025, with a particular focus on how recent learning-based advances extend, complement, or interact with classical planning foundations. We first revisit classical planning methods as algorithmic foundations and reference frameworks for learning-based extensions. We then propose a role-of-learning taxonomy that categorizes existing methods according to how learning participates in the planning pipeline, including direct policy learning, learning-augmented classical planning, hybrid planning, and training enhancement methods. For each category, we summarize the main problem settings, representative algorithms, key ideas, integration mechanisms, strengths, and limitations. We further analyze how observation representations, prediction uncertainty, interaction modeling, planner integration, safety constraints, and training strategies shape learning-based motion planning in dynamic environments. Finally, we discuss open challenges and future directions, including sim-to-real gap, safe and certifiable planning, dense crowd navigation, perception-planning coupling, and embodied AI.

Problem

• motion planning
• dynamic environments
• moving obstacles
• multi-agent interactions
• prediction uncertainty

Innovation

• learning-based motion planning
• role-of-learning taxonomy
• dynamic environments
• classical planning integration
• safety-aware planning