Haiyan Zhao
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Haiyan Zhao has a total of 23 co-authors. Here are the top co-authors by number of shared publications:
- Bingzhao Gao
Tongji University
12 shared publications - Hong Chen
Tongji University
11 shared publications - Xinghao Lu
College of Communication Engineering, Jilin University, Changchun, China
10 shared publications - Hongbin Xie
College of Communication Engineering, Jilin University, Changchun, China
7 shared publications - Chengcheng Xu
College of Communication Engineering, Jilin University, Changchun, China
4 shared publications - Kang Sun
College of Communication Engineering, Jilin University, Changchun, China
2 shared publications
Latest Publications
- 19 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Electronics
- Vol. 73 (5) , 7530-7541
Adaptive Weighting Control for Electric Vehicles Based on Thermal States and Motion Demands
To enhance adaptive coordination between vehicle stability and motor thermal protection, a novel adaptive control framework is proposed in this article. The method integrates vehicle motion states and motor thermal models into a unified nonlinear model predictive control (MPC) structure. Specifically, adaptive weighting strategies including vehicle stability index and motor thermal index are designed to dynamically adjust the cost function coefficients based on real-time thermal and vehicle conditions, enabling the controller to flexibly prioritize competing objectives in response to changing states and thermal stresses, thereby significantly improving control adaptability. Additionally, a soft constraint mechanism based on motor temperature trends is proposed to explicitly incorporate anticipated thermal behavior into the cost function as a predictive penalty term, which guides the controller to regulate control inputs in advance, effectively reducing overheating risks before critical thresholds are reached. Furthermore, vehicle stability, motor safety, and input smoothness are jointly addressed through coordinated optimization of torque allocation and steering commands, ensuring consistent and reliable performance under various conditions. Finally, the hardware-in-the-loop (HIL) results demonstrate the effectiveness of the proposed method in ensuring vehicle stability while maintaining thermal safety.
- 19 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Electronics
- Vol. 73 (5) , 7530-7541
Adaptive Weighting Control for Electric Vehicles Based on Thermal States and Motion Demands
To enhance adaptive coordination between vehicle stability and motor thermal protection, a novel adaptive control framework is proposed in this article. The method integrates vehicle motion states and motor thermal models into a unified nonlinear model predictive control (MPC) structure. Specifically, adaptive weighting strategies including vehicle stability index and motor thermal index are designed to dynamically adjust the cost function coefficients based on real-time thermal and vehicle conditions, enabling the controller to flexibly prioritize competing objectives in response to changing states and thermal stresses, thereby significantly improving control adaptability. Additionally, a soft constraint mechanism based on motor temperature trends is proposed to explicitly incorporate anticipated thermal behavior into the cost function as a predictive penalty term, which guides the controller to regulate control inputs in advance, effectively reducing overheating risks before critical thresholds are reached. Furthermore, vehicle stability, motor safety, and input smoothness are jointly addressed through coordinated optimization of torque allocation and steering commands, ensuring consistent and reliable performance under various conditions. Finally, the hardware-in-the-loop (HIL) results demonstrate the effectiveness of the proposed method in ensuring vehicle stability while maintaining thermal safety.
- 31 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Systems, Man, and Cybernetics: Systems
- Vol. 56 (1) , 350-361
An Adaptive Trajectory Planning Method of Autonomous Vehicles Integrating Multiple Tasks
In order to improve the environmental adaptation and safety of autonomous vehicles trajectory planning process in a complex driving environment, a novel trajectory planning method which meets the requirement of multidriving tasks and adapts to various driving conditions is proposed in this article. In the trajectory planning method, the optimal control problem considering multiple driving tasks is established based on the constructed performance function and constraint analysis of different driving tasks to ensure the accurate realization of driving tasks. Besides, the neural network empirical model, precollision detection model, and trajectory evaluation model are designed by the consideration of selecting the optimal planning parameters in different driving conditions to enhance the adaptability to traffic environment. The advantage of the proposed method is that it not only meets the requirements of a variety of driving tasks, but also able to select the optimal planning parameters according to different traffic conditions while existing methods usually only meet single planning task, such as lane change, and has the fixed and rigid parameter selection. Four different typical scenarios are given to verify the effectiveness of the proposed method and the results show that the proposed trajectory planning method is able to ensure the safety of the vehicle and adapt to different traffic environments flexibly.
- 31 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Systems, Man, and Cybernetics: Systems
- Vol. 56 (1) , 350-361
An Adaptive Trajectory Planning Method of Autonomous Vehicles Integrating Multiple Tasks
In order to improve the environmental adaptation and safety of autonomous vehicles trajectory planning process in a complex driving environment, a novel trajectory planning method which meets the requirement of multidriving tasks and adapts to various driving conditions is proposed in this article. In the trajectory planning method, the optimal control problem considering multiple driving tasks is established based on the constructed performance function and constraint analysis of different driving tasks to ensure the accurate realization of driving tasks. Besides, the neural network empirical model, precollision detection model, and trajectory evaluation model are designed by the consideration of selecting the optimal planning parameters in different driving conditions to enhance the adaptability to traffic environment. The advantage of the proposed method is that it not only meets the requirements of a variety of driving tasks, but also able to select the optimal planning parameters according to different traffic conditions while existing methods usually only meet single planning task, such as lane change, and has the fixed and rigid parameter selection. Four different typical scenarios are given to verify the effectiveness of the proposed method and the results show that the proposed trajectory planning method is able to ensure the safety of the vehicle and adapt to different traffic environments flexibly.
- 18 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Informatics
- Vol. 21 (9) , 6979-6989
A Centralized Fault Prevention Method for Vehicle Longitudinal-Lateral-Vertical Control Considering Motor Thermal Protection
In order to prevent the overheating of motors and enhance vehicle stability, a centralized fault prevention control method based on model predictive control that incorporates a motor thermal model is proposed. The motor thermal model is separated into heat generation and heat dissipation components, precisely capturing the impact of motor torque on temperature variations. By dynamically adjusting motor torque distribution, it effectively regulates motor temperature, extends motors lifespan, and addresses the research gap in the coupling of motor thermal management and vehicle motion control. Besides, it coordinates the control of motor torques, steering angles, and active suspension forces within a unified centralized architecture, thereby avoiding response delays and performance conflicts that may arise from distributed control framework, achieving a balance between vehicle dynamics and thermal protection. Finally, the effectiveness of the proposed method in preventing motor overheating and controlling vehicle stability while maintaining ride comfort is validated through hardware-in-the-loop testing.
- 18 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Informatics
- Vol. 21 (9) , 6979-6989
A Centralized Fault Prevention Method for Vehicle Longitudinal-Lateral-Vertical Control Considering Motor Thermal Protection
In order to prevent the overheating of motors and enhance vehicle stability, a centralized fault prevention control method based on model predictive control that incorporates a motor thermal model is proposed. The motor thermal model is separated into heat generation and heat dissipation components, precisely capturing the impact of motor torque on temperature variations. By dynamically adjusting motor torque distribution, it effectively regulates motor temperature, extends motors lifespan, and addresses the research gap in the coupling of motor thermal management and vehicle motion control. Besides, it coordinates the control of motor torques, steering angles, and active suspension forces within a unified centralized architecture, thereby avoiding response delays and performance conflicts that may arise from distributed control framework, achieving a balance between vehicle dynamics and thermal protection. Finally, the effectiveness of the proposed method in preventing motor overheating and controlling vehicle stability while maintaining ride comfort is validated through hardware-in-the-loop testing.
- 10 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Intelligent Transportation Systems
- Vol. 26 (8) , 11850-11863
A Deep Reinforcement Learning Method for Autonomous Driving Integrating Multi-Modal Fusion
Due to the input of high-dimensional data in end-to-end autonomous driving, if the feature extraction network is trained online from scratch, it will lead to difficulties in learning and decision-making. To solve this challenging problem, this paper proposes a predefined multi-modal image method to improve decision-making accuracy and a shared framework to quickly train all networks. The inputs are composed of various observations including bird’s-eye view, state, and front view, rather than traditional sensor raw data. They can be filtered out ineffectual features based on human prior knowledge to avoid over-exploration with invalid changes, such as light, trees, mountains, etc. Combined with the processed representations, predefined images are formed by embedding the trajectory, forbidden line, etc. The method can assist agents to capture the coupling relationship between the actions and states, allowing networks to positively adjust weights to quickly obtain expected rewards. Considering that embedding a feature extraction network into reinforcement learning networks will lead to repeated cumulative calculations, a shared framework is proposed to independently compress features while jointly participates in online training to reduce computing costs. This module dynamically combines the policy and critic to update its weights, which overcomes decision-making problems caused by inaccurate the latent feature sequence during pre-training and fine-tuning methods. Finally, the interactive environment is constructed on a realistic driving simulator CARLA, and the fusion of different modal states is explored. The results showed that the multi-modal fusion can explore to earn maximum rewards, and the proposed methods are effective for the training.
- 10 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Intelligent Transportation Systems
- Vol. 26 (8) , 11850-11863
A Deep Reinforcement Learning Method for Autonomous Driving Integrating Multi-Modal Fusion
Due to the input of high-dimensional data in end-to-end autonomous driving, if the feature extraction network is trained online from scratch, it will lead to difficulties in learning and decision-making. To solve this challenging problem, this paper proposes a predefined multi-modal image method to improve decision-making accuracy and a shared framework to quickly train all networks. The inputs are composed of various observations including bird’s-eye view, state, and front view, rather than traditional sensor raw data. They can be filtered out ineffectual features based on human prior knowledge to avoid over-exploration with invalid changes, such as light, trees, mountains, etc. Combined with the processed representations, predefined images are formed by embedding the trajectory, forbidden line, etc. The method can assist agents to capture the coupling relationship between the actions and states, allowing networks to positively adjust weights to quickly obtain expected rewards. Considering that embedding a feature extraction network into reinforcement learning networks will lead to repeated cumulative calculations, a shared framework is proposed to independently compress features while jointly participates in online training to reduce computing costs. This module dynamically combines the policy and critic to update its weights, which overcomes decision-making problems caused by inaccurate the latent feature sequence during pre-training and fine-tuning methods. Finally, the interactive environment is constructed on a realistic driving simulator CARLA, and the fusion of different modal states is explored. The results showed that the multi-modal fusion can explore to earn maximum rewards, and the proposed methods are effective for the training.
- 3 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 25 (13) , 25605-25614
Decision-Making Method Combining Path-Constraint-Based Action Compensation for Autonomous Vehicles
In order to improve the performance and reliability of the decision-making process for autonomous vehicles, a decision-making method using deep reinforcement learning (DRL) and combining path constraints-based action compensation is proposed in this paper. In the decision-making method, an action compensation module based on path constraints is developed, which enhances the reliability of the decision of vehicle and ensures that the final output actions meet safety and comfort requirements. Additionally, the traditional reinforcement learning task is divided into three subtasks, and the switching between the subtasks is realized by the scene recognition module. It reduces the complexity of the task and improves the training efficiency and sample utilization. Moreover, rewards and penalties are designed within path constraints, incorporating the distance from the lane center and the heading deviation angle. This can provide efficient and timely feedback for the vehicle during environmental interactions, accelerating the training process and improving decision-making performance. Finally, experiments are conducted using the CARLA simulation platform. The results show that the proposed method not only enables reliable decision-making but also improves the performance of vehicle, ensuring smooth control of the vehicle through the driving task, which verifies the effectiveness of the proposed method.
- 3 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 25 (13) , 25605-25614
Decision-Making Method Combining Path-Constraint-Based Action Compensation for Autonomous Vehicles
In order to improve the performance and reliability of the decision-making process for autonomous vehicles, a decision-making method using deep reinforcement learning (DRL) and combining path constraints-based action compensation is proposed in this paper. In the decision-making method, an action compensation module based on path constraints is developed, which enhances the reliability of the decision of vehicle and ensures that the final output actions meet safety and comfort requirements. Additionally, the traditional reinforcement learning task is divided into three subtasks, and the switching between the subtasks is realized by the scene recognition module. It reduces the complexity of the task and improves the training efficiency and sample utilization. Moreover, rewards and penalties are designed within path constraints, incorporating the distance from the lane center and the heading deviation angle. This can provide efficient and timely feedback for the vehicle during environmental interactions, accelerating the training process and improving decision-making performance. Finally, experiments are conducted using the CARLA simulation platform. The results show that the proposed method not only enables reliable decision-making but also improves the performance of vehicle, ensuring smooth control of the vehicle through the driving task, which verifies the effectiveness of the proposed method.
- 16 May 2025
- Published by Elsevier in Expert Systems with Applications
- Vol. 287
Safe MBPO-based speed and lane change method integrating decision-making and control for autonomous vehicles
- 16 May 2025
- Published by Elsevier in Expert Systems with Applications
- Vol. 287
Safe MBPO-based speed and lane change method integrating decision-making and control for autonomous vehicles
- 14 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Transportation Electrification
- Vol. 11 (5) , 11017-11029
Coordinated Electro-Hydraulic Composite Braking Method for Improving Energy Recovery in Emergency Braking
To enhance the braking performance and energy recovery efficiency during emergency braking, an electro-hydraulic composite braking method integrating an antilock braking system (ABS) based on the model predictive control (MPC) is proposed. The proposed method ensures precise control of the tire slip ratio to effectively prevent wheel lock-up while maximizing the braking force generated from tire-road interaction. Leveraging the fast response of motors and the wide controllable range of hydraulic torque, regenerative and friction braking are coordinated within an optimization problem, which enables adaptive adjustment of regenerative and hydraulic braking torques, ensuring efficient energy recovery and braking safety at the same time. A unified optimization framework is designed to simultaneously consider braking torque fluctuation, slip ratio tracking, and braking deceleration, with computational redundancy significantly reduced. Finally, the hardware-in-the-loop (HIL) testing under different conditions is carried out. The results show that the braking distance is reduced by 7.3 m, and the state of charge (SOC) is increased by 5.1%, which demonstrates the effectiveness of the proposed method in ensuring both braking performance and energy recovery efficiency.
- 14 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Transportation Electrification
- Vol. 11 (5) , 11017-11029
Coordinated Electro-Hydraulic Composite Braking Method for Improving Energy Recovery in Emergency Braking
To enhance the braking performance and energy recovery efficiency during emergency braking, an electro-hydraulic composite braking method integrating an antilock braking system (ABS) based on the model predictive control (MPC) is proposed. The proposed method ensures precise control of the tire slip ratio to effectively prevent wheel lock-up while maximizing the braking force generated from tire-road interaction. Leveraging the fast response of motors and the wide controllable range of hydraulic torque, regenerative and friction braking are coordinated within an optimization problem, which enables adaptive adjustment of regenerative and hydraulic braking torques, ensuring efficient energy recovery and braking safety at the same time. A unified optimization framework is designed to simultaneously consider braking torque fluctuation, slip ratio tracking, and braking deceleration, with computational redundancy significantly reduced. Finally, the hardware-in-the-loop (HIL) testing under different conditions is carried out. The results show that the braking distance is reduced by 7.3 m, and the state of charge (SOC) is increased by 5.1%, which demonstrates the effectiveness of the proposed method in ensuring both braking performance and energy recovery efficiency.
- 5 January 2025
- Published by Elsevier in Measurement
- Vol. 245
Target-oriented deformable fast depth estimation based on stereo vision for space object detection
- 5 January 2025
- Published by Elsevier in Measurement
- Vol. 245
Target-oriented deformable fast depth estimation based on stereo vision for space object detection
- 22 October 2024
- Published by Springer Nature in Science China Information Sciences
- Vol. 67 (11) , 1-16
All-day perception for intelligent vehicles: switching perception algorithms based on WBCNet
- 22 October 2024
- Published by Springer Nature in Science China Information Sciences
- Vol. 67 (11) , 1-16
All-day perception for intelligent vehicles: switching perception algorithms based on WBCNet
- 20 August 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 24 (19) , 31466-31480
Multisensor Decision-Level Fusion Network Based on Attention Mechanism for Object Detection
To solve the problem of low accuracy caused by threshold constraints in traditional decision-level fusion methods, this paper proposes a deep learning method based on attention mechanism to fuse the three-dimensional information of sensors. The proposed model based on attention mechanism (AFnet) can improve the accuracy of the detection system without relying on traditional constraints. The AFnet model decouples the correlation between the data by the encoder, and fully utilizes the nonlinear fitting capability of deep learning. The adaptive fusion can be realized under data scale and result bias, which effectively solves the problem caused by traditional methods in the case of vehicle occlusion and overlap. The depth information and object detection networks are combined by embedding, which ensures that cameras can achieve spatial detection of vehicles and overcome the limitations of two-dimensional plane. The redefined clustering method takes into account the spatial position and velocity attribute, which can effectively distinguish high-density overlapping point clouds. Finally, experimental results in NuScenes and Carla show that the proposed fusion method does not rely on traditional rule constraints, and improves the accuracy of object detection. The fusion model of AFnet presents the state of the art on fusion matching accuracy of 99.11%.
- 20 August 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 24 (19) , 31466-31480
Multisensor Decision-Level Fusion Network Based on Attention Mechanism for Object Detection
To solve the problem of low accuracy caused by threshold constraints in traditional decision-level fusion methods, this paper proposes a deep learning method based on attention mechanism to fuse the three-dimensional information of sensors. The proposed model based on attention mechanism (AFnet) can improve the accuracy of the detection system without relying on traditional constraints. The AFnet model decouples the correlation between the data by the encoder, and fully utilizes the nonlinear fitting capability of deep learning. The adaptive fusion can be realized under data scale and result bias, which effectively solves the problem caused by traditional methods in the case of vehicle occlusion and overlap. The depth information and object detection networks are combined by embedding, which ensures that cameras can achieve spatial detection of vehicles and overcome the limitations of two-dimensional plane. The redefined clustering method takes into account the spatial position and velocity attribute, which can effectively distinguish high-density overlapping point clouds. Finally, experimental results in NuScenes and Carla show that the proposed fusion method does not rely on traditional rule constraints, and improves the accuracy of object detection. The fusion model of AFnet presents the state of the art on fusion matching accuracy of 99.11%.