Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

DPNet: Doppler LiDAR Motion Planning for Highly-Dynamic Environments

Wei Zuo, Zeyi Ren, Chengyang Li, Yikun Wang, Mingle Zhao, Shuai Wang, Wei Sui, Fei Gao, Yik-Chung Wu, Chengzhong Xu

TL;DR

DPNet integrates Doppler LiDAR into closed-loop motion planning by coupling a Doppler-informed obstacle tracker (D-KalmanNet) with a Doppler-tuned MPC (DT-MPC). The approach achieves high tracking frequency and planning accuracy under partial observability, using Doppler velocity rectification and Kalman gain learning to handle noise and model mismatch. DT-MPC enables run-time controller tuning via Doppler-inferred collision checks and ADMM-based optimization, improving robustness in congested, fast-changing environments. Experimental results in CARLA and real Doppler LiDAR datasets demonstrate superior performance and real-time capability on edge hardware, highlighting the practical impact for autonomous navigation."

Abstract

Existing motion planning methods often struggle with rapid-motion obstacles due to an insufficient understanding of environmental changes. To address this limitation, we propose integrating motion planners with Doppler LiDARs which provide not only ranging measurements but also instantaneous point velocities. However, this integration is nontrivial due to the dual requirements of high accuracy and high frequency. To this end, we introduce Doppler Planning Network (DPNet), which tracks and reacts to rapid obstacles using Doppler model-based learning. Particularly, we first propose a Doppler Kalman neural network (D-KalmanNet) to track the future states of obstacles under partially observable Gaussian state space model. We then leverage the estimated motions to construct a Doppler-tuned model predictive control (DT-MPC) framework for ego-motion planning, enabling runtime auto-tuning of the controller parameters. These two model-based learners allow DPNet to maintain lightweight while learning fast environmental changes using minimum data, and achieve both high frequency and high accuracy in tracking and planning. Experiments on both high-fidelity simulator and real-world datasets demonstrate the superiority of DPNet over extensive benchmark schemes.

DPNet: Doppler LiDAR Motion Planning for Highly-Dynamic Environments

TL;DR

DPNet integrates Doppler LiDAR into closed-loop motion planning by coupling a Doppler-informed obstacle tracker (D-KalmanNet) with a Doppler-tuned MPC (DT-MPC). The approach achieves high tracking frequency and planning accuracy under partial observability, using Doppler velocity rectification and Kalman gain learning to handle noise and model mismatch. DT-MPC enables run-time controller tuning via Doppler-inferred collision checks and ADMM-based optimization, improving robustness in congested, fast-changing environments. Experimental results in CARLA and real Doppler LiDAR datasets demonstrate superior performance and real-time capability on edge hardware, highlighting the practical impact for autonomous navigation."

Abstract

Existing motion planning methods often struggle with rapid-motion obstacles due to an insufficient understanding of environmental changes. To address this limitation, we propose integrating motion planners with Doppler LiDARs which provide not only ranging measurements but also instantaneous point velocities. However, this integration is nontrivial due to the dual requirements of high accuracy and high frequency. To this end, we introduce Doppler Planning Network (DPNet), which tracks and reacts to rapid obstacles using Doppler model-based learning. Particularly, we first propose a Doppler Kalman neural network (D-KalmanNet) to track the future states of obstacles under partially observable Gaussian state space model. We then leverage the estimated motions to construct a Doppler-tuned model predictive control (DT-MPC) framework for ego-motion planning, enabling runtime auto-tuning of the controller parameters. These two model-based learners allow DPNet to maintain lightweight while learning fast environmental changes using minimum data, and achieve both high frequency and high accuracy in tracking and planning. Experiments on both high-fidelity simulator and real-world datasets demonstrate the superiority of DPNet over extensive benchmark schemes.

Paper Structure

This paper contains 16 sections, 17 equations, 7 figures, 4 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. Problem Formulation
  4. Doppler Planning Network
  5. Doppler KalmanNet for Tracking
  6. Doppler Velocity Rectification
  7. Kalman Gain Learning
  8. Doppler-Tuned MPC for Planning
  9. Doppler-Inferred Collision Check
  10. Run-Time Auto-Tuning
  11. Solution by ADMM
  12. Experiments
  13. End-to-End Evaluation in Carla
  14. Evaluation on Real-World Datasets
  15. More Quantitative Results
  16. ...and 1 more sections

Figures (7)

  • Figure 1: Doppler model-based learning enables DPNet to understand and traverse highly-dynamic environments.
  • Figure 2: The proposed DPNet system, which consists of D-KalmanNet and DT-MPC modules.
  • Figure 3: Doppler velocity rectification.
  • Figure 4: Qualitative analysis of robot motions (in ROS-Rviz and Carla dosovitskiy2017carla views) and corresponding control commands. Static obstacles are marked as green boxes, while dynamic ones are red boxes with arrows indicating their moving directions. The blue line is the path connecting start and goal points. The dots in front of dynamic obstacles represent motion predictions.
  • Figure 5: Clutter levels. D (S) refers to dynamic (static) obstacles.
  • ...and 2 more figures