Research Analyzer
← Back ICRA 2026

Future-Oriented Navigation: Dynamic Obstacle Avoidance with One-Shot Energy-Based Multimodal Motion Prediction

Ze Zhang, Georg Hess, Junjie Hu, Emmanuel Dean, Lennart Svensson, Knut Akesson

PDF

AI summary

Key figure (auto-extracted from paper)
A one-shot energy-based neural network predicts multimodal obstacle motions, which are grouped and fed into a proactive MPC planner to enable safe, deadlock-free navigation for mobile robots in dynamic environments.
Dynamic obstacle avoidance energy-based learning multimodal motion prediction model predictive control mobile robots warehouse automation

Problem

AMRs struggle with collision-free navigation in dynamic environments due to unpredictable obstacle movements and the computational limits of traditional prediction-planning pipelines, often triggering the "Freezing Robot Problem".

Approach

The framework combines a fast energy-based neural network for one-shot multimodal motion prediction with a grouped, proactive Model Predictive Control (MPC) planner that treats predicted futures as soft constraints to guide safe fleet navigation.

Key results

  • Stable energy-based learning framework with modified loss and PELU activation for fast multimodal prediction
  • Proactive MPC controller that groups predicted obstacle futures to reduce computation and prevent freezing
  • Superior navigation performance over state-of-the-art methods in dynamic warehouse and hospital scenarios
  • Effective multi-robot fleet coordination and collision-free path planning in busy industrial settings

Why it matters

Provides a robust, computationally efficient navigation solution for AMRs in dynamic industrial settings, advancing safe human-aware robotics and warehouse automation.

Abstract

This paper proposes an integrated approach for the safe and efficient control of mobile robots in dynamic and uncertain environments. The approach consists of two key steps: one-shot multimodal motion prediction to anticipate motions of dynamic obstacles and model predictive control to incorporate these predictions into the motion planning process. Motion prediction is driven by an energy-based neural network that generates high-resolution, multi-step predictions in a sin- gle operation. The prediction outcomes are further utilized to create geometric shapes formulated as mathematical constraints. Instead of treating each dynamic obstacle individually, predicted obstacles are grouped by proximity in an unsupervised way to improve performance and efficiency. The overall collision- free navigation is handled by model predictive control with a specific design for proactive dynamic obstacle avoidance. The proposed approach allows mobile robots to navigate effectively in dynamic environments. Its performance is accessed across various scenarios that represent typical warehouse settings. The results demonstrate that the proposed approach outperforms other existing dynamic obstacle avoidance methods.

Index terms

Human-Aware Motion Planning Collision Avoidance Deep Learning Methods

Related papers