Yijian Duan
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Yijian Duan has a total of 22 co-authors. Here are the top co-authors by number of shared publications:
- Jihong Zhu
Department of Precision Instrument, Tsinghua University, Beijing 100084, China
3 shared publications - Jinlai Zhang
Changsha University of Science and Technology
3 shared publications Jihong Zhu
2 shared publicationsLiwen Meng
College of Mechanical Engineering, Guangxi University, Nanning, Guangxi, China
2 shared publicationsXin Liu
College of Mechanical Engineering, Guangxi University, Nanning, China
2 shared publicationsYanmei Meng
2 shared publications
Latest Publications
- 26 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Vehicular Technology
- Vol. 75 (3) , 3916-3927
Reinforcement Learning Decision and Planning Algorithm Using Posterior Return Heuristic
The complexities and dynamics of modern driving environments have amplified the uncertainties faced by autonomous driving systems, posing significant challenges to their decision-making processes. This paper introduces a novel reinforcement learning framework inspired by posterior return estimation, designed to enhance safety and foresight in robotic decision-making. To overcome the challenges of long-term dependency in reinforcement learning, a tree structure is utilized to record the interaction history between the agent and the environment. The posterior distribution of returns for each state is estimated from this historical data, with a neural network employed to model this distribution for enhanced decision-making. The framework's efficacy is validated through two decision-making and planning scenarios, where it surpasses several commonly used frameworks and demonstrates notably more anticipatory decision-making patterns.
- 26 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Vehicular Technology
- Vol. 75 (3) , 3916-3927
Reinforcement Learning Decision and Planning Algorithm Using Posterior Return Heuristic
The complexities and dynamics of modern driving environments have amplified the uncertainties faced by autonomous driving systems, posing significant challenges to their decision-making processes. This paper introduces a novel reinforcement learning framework inspired by posterior return estimation, designed to enhance safety and foresight in robotic decision-making. To overcome the challenges of long-term dependency in reinforcement learning, a tree structure is utilized to record the interaction history between the agent and the environment. The posterior distribution of returns for each state is estimated from this historical data, with a neural network employed to model this distribution for enhanced decision-making. The framework's efficacy is validated through two decision-making and planning scenarios, where it surpasses several commonly used frameworks and demonstrates notably more anticipatory decision-making patterns.
- 15 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 25 (19) , 36466-36478
Adversarial Patches for Fooling Positive–Negative Obstacle Detectors on Unstructured Roads
Adversarial patches represent a critical form of physical adversarial attacks, posing significant risks to the security of neural network-based object detection systems. Previous research on adversarial patches has predominantly focused on pedestrian detection, facial recognition, and vehicle detection, with limited attention to the detection of positive and negative obstacles on unstructured roads. Moreover, prior studies typically optimize perturbation information while fixing parameters such as the patch’s position and rotation angle, or manipulate parameters such as position and rotation angle while fixing the perturbation information to generate adversarial patches. In this context, we propose a PRP multi-adversarial patch attack method based on a simulated annealing improved differential evolution algorithm (SADE-PRP), designed to deceive unstructured positive and negative obstacle detection systems. Firstly, We propose a PRP-guided multi-adversarial patch framework for visible light-based positive-negative obstacle detection systems. This framework simultaneously considers three features of the adversarial patch: position, rotation angle and perturbation pixels, as opposed to manually setting these parameters like most previous works. Secondly, we employ more robust based simulated annealing improved differential evolution algorithm (SADE), which effectively improved the robustness, achieving higher attack success rates in black-box attack scenarios. Finally, we constructed a dataset of positive-negative obstacles on unstructured roads. Then, we conducted extensive experiments in both digital and physical environments to demonstrate the superiority of the proposed method.
- 15 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 25 (19) , 36466-36478
Adversarial Patches for Fooling Positive–Negative Obstacle Detectors on Unstructured Roads
Adversarial patches represent a critical form of physical adversarial attacks, posing significant risks to the security of neural network-based object detection systems. Previous research on adversarial patches has predominantly focused on pedestrian detection, facial recognition, and vehicle detection, with limited attention to the detection of positive and negative obstacles on unstructured roads. Moreover, prior studies typically optimize perturbation information while fixing parameters such as the patch’s position and rotation angle, or manipulate parameters such as position and rotation angle while fixing the perturbation information to generate adversarial patches. In this context, we propose a PRP multi-adversarial patch attack method based on a simulated annealing improved differential evolution algorithm (SADE-PRP), designed to deceive unstructured positive and negative obstacle detection systems. Firstly, We propose a PRP-guided multi-adversarial patch framework for visible light-based positive-negative obstacle detection systems. This framework simultaneously considers three features of the adversarial patch: position, rotation angle and perturbation pixels, as opposed to manually setting these parameters like most previous works. Secondly, we employ more robust based simulated annealing improved differential evolution algorithm (SADE), which effectively improved the robustness, achieving higher attack success rates in black-box attack scenarios. Finally, we constructed a dataset of positive-negative obstacles on unstructured roads. Then, we conducted extensive experiments in both digital and physical environments to demonstrate the superiority of the proposed method.
- 11 December 2024
- Published by Elsevier in Food and Bioproducts Processing
- Vol. 149, 353-367
Multi-scale convolution and dynamic task interaction detection head for efficient lightweight plum detection
- 11 December 2024
- Published by Elsevier in Food and Bioproducts Processing
- Vol. 149, 353-367
Multi-scale convolution and dynamic task interaction detection head for efficient lightweight plum detection
- Open Access
- 3 October 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
- Vol. 17 (19391404) , 19627-19639
MFSA-Net: Semantic Segmentation With Camera-LiDAR Cross-Attention Fusion Based on Fast Neighbor Feature Aggregation
Given the inherent limitations of Camera-Only and LiDAR-Only Methods in performing semantic segmentation tasks in large-scale complex environments, multimodal information fusion for semantic segmentation has become a focal point of contemporary research. However, significant modal disparities often result in existing fusion-based methods struggling with low segmentation accuracy and limited efficiency in large-scale complex environments. To address these challenges,We propose a semantic segmentation network with Camera-LiDAR cross-attention fusion based on fast neighbor feature aggregation (MFSA-Net), which is better suited for large-scale semantic segmentation in complex environments.Initially, we propose a dual-distance attention feature aggregation module based on rapid 3D nearest neighbor search. This module employs a sliding window method in point cloud perspective projections for swift proximity search, and efficiently combines feature distance and Euclidean distance information to learn more distinctive local features. This improves segmentation accuracy while ensuring computational efficiency. Furthermore, we propose a Cross-Attention Fusion Two-Stream Network Based on Residual, which allows for more effective integration of Camera information into the LiDAR data stream, enhancing both accuracy and robustness. Extensive experimental results on the large-scale point cloud datasets SemanticKITTI and Nuscenes demonstrate that our proposed algorithm outperforms similar algorithms in semantic segmentation performance in large-scale complex environments.
- 3 October 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
- Vol. 17 (19391404) , 19627-19639
Open AccessMFSA-Net: Semantic Segmentation With Camera-LiDAR Cross-Attention Fusion Based on Fast Neighbor Feature Aggregation
Given the inherent limitations of Camera-Only and LiDAR-Only Methods in performing semantic segmentation tasks in large-scale complex environments, multimodal information fusion for semantic segmentation has become a focal point of contemporary research. However, significant modal disparities often result in existing fusion-based methods struggling with low segmentation accuracy and limited efficiency in large-scale complex environments. To address these challenges,We propose a semantic segmentation network with Camera-LiDAR cross-attention fusion based on fast neighbor feature aggregation (MFSA-Net), which is better suited for large-scale semantic segmentation in complex environments.Initially, we propose a dual-distance attention feature aggregation module based on rapid 3D nearest neighbor search. This module employs a sliding window method in point cloud perspective projections for swift proximity search, and efficiently combines feature distance and Euclidean distance information to learn more distinctive local features. This improves segmentation accuracy while ensuring computational efficiency. Furthermore, we propose a Cross-Attention Fusion Two-Stream Network Based on Residual, which allows for more effective integration of Camera information into the LiDAR data stream, enhancing both accuracy and robustness. Extensive experimental results on the large-scale point cloud datasets SemanticKITTI and Nuscenes demonstrate that our proposed algorithm outperforms similar algorithms in semantic segmentation performance in large-scale complex environments.
- Open Access
- 1 July 2022
- Published by SAGE Publications in International Journal of Advanced Robotic Systems
- Vol. 19 (4)
Active obstacle avoidance method of autonomous vehicle based on improved artificial potential field
Aiming at the local minimum point problem in an artificial potential field based on a safe distance model, this article proposes an algorithm for active obstacle avoidance path planning and tracking of autonomous vehicles using an improved artificial potential field. First, a possible road operating condition in which the artificial potential field based on the safety-distance model falls into a local minimum point is studied. Subsequently, an improved artificial potential field method is proposed by introducing the second virtual target attraction potential field, which successfully overcomes the local minimum point problem. Second, a model for autonomous vehicle active obstacle avoidance path planning and tracking based on the improved artificial potential field is established. Finally, MATLAB/CarSim co-simulations were performed under the road conditions of constant- and variable-velocity obstacle vehicles. The simulation results demonstrate that the improved artificial potential field method can effectively solve the local minimum point problem of the artificial potential field based on the safe distance model. Additionally, the safety and stability of autonomous vehicle active obstacle avoidance are improved.
- 1 July 2022
- Published by SAGE Publications in International Journal of Advanced Robotic Systems
- Vol. 19 (4)
Open AccessActive obstacle avoidance method of autonomous vehicle based on improved artificial potential field
Aiming at the local minimum point problem in an artificial potential field based on a safe distance model, this article proposes an algorithm for active obstacle avoidance path planning and tracking of autonomous vehicles using an improved artificial potential field. First, a possible road operating condition in which the artificial potential field based on the safety-distance model falls into a local minimum point is studied. Subsequently, an improved artificial potential field method is proposed by introducing the second virtual target attraction potential field, which successfully overcomes the local minimum point problem. Second, a model for autonomous vehicle active obstacle avoidance path planning and tracking based on the improved artificial potential field is established. Finally, MATLAB/CarSim co-simulations were performed under the road conditions of constant- and variable-velocity obstacle vehicles. The simulation results demonstrate that the improved artificial potential field method can effectively solve the local minimum point problem of the artificial potential field based on the safe distance model. Additionally, the safety and stability of autonomous vehicle active obstacle avoidance are improved.