Autonomous Navigation in Dynamic Human Environments with an Embedded 2D LiDAR-based Person Tracker

TL;DR

This work tackles safe autonomous navigation in dynamic human environments by presenting an embedded real-time pipeline that fuses 2D LiDAR-based human detection, multi-object tracking, and local planning. It employs DR-SPAAM for detection, a tailored Norfair-based tracker with constant-velocity Kalman filtering, and the Timed-Elastic-Band (TEB) planner within the ROS Navigation Stack, achieving real-time operation at 20 Hz on a Jetson Xavier NX. Experimental validation on a Unitree A1 quadruped with three new real-world datasets yields a high MOTA (up to 85.45% on average for Config-3) and demonstrates improved collision avoidance when tracking information informs planning. The approach emphasizes modularity and transferability of components, contributing to safer human-robot cohabitation in shared spaces and providing a foundation for future enhancements in detection, tracking accuracy, and planning under dynamic hazards.

Abstract

In the rapidly evolving landscape of autonomous mobile robots, the emphasis on seamless human-robot interactions has shifted towards autonomous decision-making. This paper delves into the intricate challenges associated with robotic autonomy, focusing on navigation in dynamic environments shared with humans. It introduces an embedded real-time tracking pipeline, integrated into a navigation planning framework for effective person tracking and avoidance, adapting a state-of-the-art 2D LiDAR-based human detection network and an efficient multi-object tracker. By addressing the key components of detection, tracking, and planning separately, the proposed approach highlights the modularity and transferability of each component to other applications. Our tracking approach is validated on a quadruped robot equipped with 270° 2D-LiDAR against motion capture system data, with the preferred configuration achieving an average MOTA of 85.45% in three newly recorded datasets, while reliably running in real-time at 20 Hz on the NVIDIA Jetson Xavier NX embedded GPU-accelerated platform. Furthermore, the integrated tracking and avoidance system is evaluated in real-world navigation experiments, demonstrating how accurate person tracking benefits the planner in optimizing the generated trajectories, enhancing its collision avoidance capabilities. This paper contributes to safer human-robot cohabitation, blending recent advances in human detection with responsive planning to navigate shared spaces effectively and securely.

Figures (4)

• Figure 1: Unitree A1 robot used in this work with the additional backpack system.
• Figure 2: Pipelined execution of detector (inference time $T_{det}^{i}$) and tracker (update time $T_{track}^{i}$) for each LiDAR scan $i$, received regularly with period $T_{scan}$. The overall detection and tracking pipeline update latency is $T_{lat}^{i}$.
• Figure 3:
• Figure 6: Comparative navigation experiments illustrating the robot and human trajectories from the Vicon Motion Capture system. TEB is configured without our tracker (top) where it fails to avoid collision, and with Config-3 (bottom) where it successfully performs an early avoidance maneuver.