Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

HumanDiffusion: A Vision-Based Diffusion Trajectory Planner with Human-Conditioned Goals for Search and Rescue UAV

Faryal Batool, Iana Zhura, Valerii Serpiva, Roohan Ahmed Khan, Ivan Valuev, Issatay Tokmurziyev, Dzmitry Tsetserukou

TL;DR

HumanDiffusion tackles autonomous, human-aware UAV navigation in time-critical SAR settings without relying on explicit maps. It combines a YOLO-based perception module with a conditional diffusion trajectory generator to produce pixel-space trajectories conditioned on RGB imagery and inferred start–goal pairs, which are then projected into 3D for execution. Trained on simulated A*-generated trajectories, the approach demonstrates a pixel-space MSE of $0.02$ on a test set and an overall real-world success rate of $80\%$ across two indoor scenarios. The results indicate diffusion-based planning as a practical, robust alternative to classical navigation methods for dynamic human–robot collaboration in emergency response.

Abstract

Reliable human--robot collaboration in emergency scenarios requires autonomous systems that can detect humans, infer navigation goals, and operate safely in dynamic environments. This paper presents HumanDiffusion, a lightweight image-conditioned diffusion planner that generates human-aware navigation trajectories directly from RGB imagery. The system combines YOLO-11--based human detection with diffusion-driven trajectory generation, enabling a quadrotor to approach a target person and deliver medical assistance without relying on prior maps or computationally intensive planning pipelines. Trajectories are predicted in pixel space, ensuring smooth motion and a consistent safety margin around humans. We evaluate HumanDiffusion in simulation and real-world indoor mock-disaster scenarios. On a 300-sample test set, the model achieves a mean squared error of 0.02 in pixel-space trajectory reconstruction. Real-world experiments demonstrate an overall mission success rate of 80% across accident-response and search-and-locate tasks with partial occlusions. These results indicate that human-conditioned diffusion planning offers a practical and robust solution for human-aware UAV navigation in time-critical assistance settings.

HumanDiffusion: A Vision-Based Diffusion Trajectory Planner with Human-Conditioned Goals for Search and Rescue UAV

TL;DR

HumanDiffusion tackles autonomous, human-aware UAV navigation in time-critical SAR settings without relying on explicit maps. It combines a YOLO-based perception module with a conditional diffusion trajectory generator to produce pixel-space trajectories conditioned on RGB imagery and inferred start–goal pairs, which are then projected into 3D for execution. Trained on simulated A*-generated trajectories, the approach demonstrates a pixel-space MSE of on a test set and an overall real-world success rate of across two indoor scenarios. The results indicate diffusion-based planning as a practical, robust alternative to classical navigation methods for dynamic human–robot collaboration in emergency response.

Abstract

Reliable human--robot collaboration in emergency scenarios requires autonomous systems that can detect humans, infer navigation goals, and operate safely in dynamic environments. This paper presents HumanDiffusion, a lightweight image-conditioned diffusion planner that generates human-aware navigation trajectories directly from RGB imagery. The system combines YOLO-11--based human detection with diffusion-driven trajectory generation, enabling a quadrotor to approach a target person and deliver medical assistance without relying on prior maps or computationally intensive planning pipelines. Trajectories are predicted in pixel space, ensuring smooth motion and a consistent safety margin around humans. We evaluate HumanDiffusion in simulation and real-world indoor mock-disaster scenarios. On a 300-sample test set, the model achieves a mean squared error of 0.02 in pixel-space trajectory reconstruction. Real-world experiments demonstrate an overall mission success rate of 80% across accident-response and search-and-locate tasks with partial occlusions. These results indicate that human-conditioned diffusion planning offers a practical and robust solution for human-aware UAV navigation in time-critical assistance settings.
Paper Structure (20 sections, 5 equations, 3 figures)

This paper contains 20 sections, 5 equations, 3 figures.

Table of Contents

  1. Introduction
  2. Related Works
  3. Human-Centered UAV Navigation for Search and Rescue
  4. UAV Navigation and Planning
  5. Diffusion Models for Trajectory Planning
  6. Vision--Language Models for UAV Reasoning
  7. System Architecture
  8. Perception and Human Goal Inference
  9. Diffusion-Based Trajectory Generator
  10. Forward Diffusion Process.
  11. Reverse Denoising Process.
  12. Training Objective.
  13. Training Pipeline and Dataset Generation
  14. Experimental Evaluation
  15. Results and Failure Analysis
  16. ...and 5 more sections

Figures (3)

  • Figure 1: Comparison between diffusion predicted and annotated ground truth trajectories.
  • Figure 2: Human detection results for Scenario 1 with the corresponding diffusion-planned trajectories and the executed 3D flight path showing start position and goal updates.
  • Figure 3: Human detection results for Scenario 2 with the corresponding diffusion-planned trajectories and the executed 3D flight path showing start position and goal updates.