Donglai Xue
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Donglai Xue has a total of 12 co-authors. Here are the top co-authors by number of shared publications:
- Fei Gao
Zhejiang University
2 shared publications - Chao Xu
Zhejiang University
1 shared publication Jiaxi Xing
Huzhou Institute of Zhejiang University, Huzhou, China
1 shared publication- Jingping Wang
Institute of Cyber-Systems and Control, College of Control Science and Engineering, Zhejiang University, Hangzhou, China
1 shared publication - Mengze Tian
Institute of Cyber-Systems and Control, College of Control Science and Engineering, Zhejiang University, Hangzhou, China
1 shared publication - Qianhao Wang
State Key Laboratory of Industrial Control Technology, Institute of Cyber-Systems and Control, Zhejiang University, Hangzhou, China
1 shared publication
Latest Publications
- 18 June 2025
- Published by American Association for the Advancement of Science (AAAS) in Science Robotics
- Vol. 10 (103) , eads4551
Hierarchically depicting vehicle trajectory with stability in complex environments
The rapid development of autonomous robots has resulted in marked societal and economic benefits. However, enabling robots to navigate complex environments with human-like agility remains a formidable challenge. Unlike robots, humans excel at pathfinding because of their superior spatial awareness and their ability to leverage experience. Inspired by these observations, we designed a neural network to simulate the intuitive pathfinding abilities of humans, integrating global environmental information and previous experiences to identify feasible pathways. Experiments demonstrated that, unlike traditional algorithms whose efficiency deteriorates in complex settings, the proposed method maintains stable computational performance. To further enhance motion quality, we introduce a numerically stable spatiotemporal trajectory optimizer with a unique bilayer polynomial trajectory representation in flat space. This optimization leverages differential flatness to enhance efficiency and fundamentally eliminates singularities in the original problem, thereby robustly converging to continuous and feasible motion even in complex maneuvering scenarios. Our hierarchical motion planner, validated through large-scale maze experiments, combines front-end path planning with back-end trajectory refinement, achieving robust and efficient navigation. We anticipate that our planner will advance stable navigation for robots in complex environments, thereby propelling the progress of robotic autonomy.
- 18 June 2025
- Published by American Association for the Advancement of Science (AAAS) in Science Robotics
- Vol. 10 (103) , eads4551
Hierarchically depicting vehicle trajectory with stability in complex environments
The rapid development of autonomous robots has resulted in marked societal and economic benefits. However, enabling robots to navigate complex environments with human-like agility remains a formidable challenge. Unlike robots, humans excel at pathfinding because of their superior spatial awareness and their ability to leverage experience. Inspired by these observations, we designed a neural network to simulate the intuitive pathfinding abilities of humans, integrating global environmental information and previous experiences to identify feasible pathways. Experiments demonstrated that, unlike traditional algorithms whose efficiency deteriorates in complex settings, the proposed method maintains stable computational performance. To further enhance motion quality, we introduce a numerically stable spatiotemporal trajectory optimizer with a unique bilayer polynomial trajectory representation in flat space. This optimization leverages differential flatness to enhance efficiency and fundamentally eliminates singularities in the original problem, thereby robustly converging to continuous and feasible motion even in complex maneuvering scenarios. Our hierarchical motion planner, validated through large-scale maze experiments, combines front-end path planning with back-end trajectory refinement, achieving robust and efficient navigation. We anticipate that our planner will advance stable navigation for robots in complex environments, thereby propelling the progress of robotic autonomy.
- 10 July 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 9 (9) , 7429-7436
Towards Dense and Accurate Radar Perception via Efficient Cross-Modal Diffusion Model
Millimeter wave (mmWave) radars have attracted significant attention from both academia and industry due to their capability to operate in extreme weather conditions. However, they face challenges in terms of sparsity and noise interference, which hinder their application in the field of micro aerial vehicle (MAV) autonomous navigation. To this end, this letter proposes a novel approach to dense and accurate mmWave radar point cloud construction via cross-modal learning. Specifically, we introduce diffusion models, which possess state-of-the-art performance in generative modeling, to predict LiDAR-like point clouds from paired raw radar data. We also incorporate the most recent diffusion model inference accelerating techniques to ensure that the proposed method can be implemented on MAVs. We validate the proposed method through extensive benchmark comparisons and real-world experiments, demonstrating its superior performance and generalization ability.
- 10 July 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 9 (9) , 7429-7436
Towards Dense and Accurate Radar Perception via Efficient Cross-Modal Diffusion Model
Millimeter wave (mmWave) radars have attracted significant attention from both academia and industry due to their capability to operate in extreme weather conditions. However, they face challenges in terms of sparsity and noise interference, which hinder their application in the field of micro aerial vehicle (MAV) autonomous navigation. To this end, this letter proposes a novel approach to dense and accurate mmWave radar point cloud construction via cross-modal learning. Specifically, we introduce diffusion models, which possess state-of-the-art performance in generative modeling, to predict LiDAR-like point clouds from paired raw radar data. We also incorporate the most recent diffusion model inference accelerating techniques to ensure that the proposed method can be implemented on MAVs. We validate the proposed method through extensive benchmark comparisons and real-world experiments, demonstrating its superior performance and generalization ability.