Research Analyzer
← Back ICRA 2026

Real-Time Adaptive Motion Planning Via Point Cloud-Guided, Energy-Based Diffusion and Potential Fields

Wondmgezahu Teshome, Kian Behzad, Octavia I. Camps, Michael Everett, Milad Siami, Mario Sznaier

PDF

AI summary

Key figure (auto-extracted from paper)
Integrating energy-based diffusion models with artificial potential fields enables real-time, adaptive trajectory generation directly from point clouds in dynamic environments.
Energy-based diffusion Motion planning Artificial potential fields Point cloud encoding Pursuit-evasion Real-time adaptation

Problem

Existing motion planners struggle to balance global optimality, real-time responsiveness, and map-free operation when navigating environments with both static obstacles and dynamic adversarial agents.

Approach

A hierarchical framework that conditions an energy-based diffusion model on point cloud features for high-level planning, then continuously refines trajectories in real-time using artificial potential fields and minimal denoising steps.

Key results

  • Real-time planning algorithm combining energy-based diffusion with APF
  • Point cloud-based obstacle encoding enabling map-free navigation
  • Compositional sampling for generalization to unseen obstacle configurations
  • Successful hardware implementation and pursuit-evasion demonstrations

Why it matters

Enables robust, real-time robotic navigation for safety-critical applications like crowd evacuation and pursuit-evasion where partial environmental knowledge and rapid adaptation are essential.

Abstract

Motivated by the problem of pursuit-evasion, we present a motion planning framework that combines energy- based diffusion models with artificial potential fields for robust real time trajectory generation in complex environments. Our approach processes obstacle information directly from point clouds, enabling efficient planning without requiring complete geometric representations. The framework employs classifier-free guidance training and integrates local potential fields during sampling to enhance obstacle avoidance. In dynamic scenarios, the system generates initial trajectories using the diffusion model and continuously refines them through potential field-based adap- tation, demonstrating effective performance in pursuit-evasion scenarios with partial pursuer observability.

Index terms

AI-Based Methods Motion and Path Planning Planning under Uncertainty

Related papers