Xiaohui Hou
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Xiaohui Hou has a total of 34 co-authors. Here are the top co-authors by number of shared publications:
- Minggang Gan
Institute of Artificial Intelligence, Minzu University of China, Beijing, China
5 shared publications - Junzhi Zhang
Tsinghua University
4 shared publications - Shiyue Zhao
School of Vehicle and Mobility, Tsinghua University, Beijing, China
4 shared publications - Jie Chen
Beijing Institute of Technology
3 shared publications Wei Wu
3 shared publications- Yuan Ji
School of Mechanical and Aerospace Engineering, Nanyang Technological University, Singapore
3 shared publications
Latest Publications
- 21 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Transportation Electrification
- Vol. PP (4) , 1
Signal Temporal Logic Constrained Motion Control of Electric Vehicles under Slippery Roads Using an MPPI Approach
Driving on icy roads poses significant challenges for automated electric vehicles (AEVs) motion planning and control due to uncertain dynamics, multiple constraints, and temporal rules satisfying problems.This paper proposes a hybrid data-driven Model Predictive Path Integral (MPPI) control framework with temporal logic constraints for real-time safe motion planning and trajectory tracking on slippery roads. To handle modeling uncertainties and mismatches in vehicle dynamics, a hybrid model is constructed by combining a nominal physics-based baseline with a 1-Lipschitz residual neural network pretrained via the collected vehicle data. This model captures both nominal dynamics and unmodeled residual effects, providing a reliable system representation of AEVs for controller design with robustness performance guaranteed. The pretrained model is embedded into a sampling-based predictive controller, where state constraints and task specifications expressed by temporal logic formulas are incorporated into the cost function. So that, our control strategy ensures that the generated trajectories adhere to both safety constraints and high-level behavioral objectives with real time solving guaranteed. Real electric vehicle experiments validate the effectiveness of the proposed method and demonstrate its superiority in achieving safe and reliable planning and tracking control on icy roads.
- 21 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Transportation Electrification
- Vol. PP (4) , 1
Signal Temporal Logic Constrained Motion Control of Electric Vehicles under Slippery Roads Using an MPPI Approach
Driving on icy roads poses significant challenges for automated electric vehicles (AEVs) motion planning and control due to uncertain dynamics, multiple constraints, and temporal rules satisfying problems.This paper proposes a hybrid data-driven Model Predictive Path Integral (MPPI) control framework with temporal logic constraints for real-time safe motion planning and trajectory tracking on slippery roads. To handle modeling uncertainties and mismatches in vehicle dynamics, a hybrid model is constructed by combining a nominal physics-based baseline with a 1-Lipschitz residual neural network pretrained via the collected vehicle data. This model captures both nominal dynamics and unmodeled residual effects, providing a reliable system representation of AEVs for controller design with robustness performance guaranteed. The pretrained model is embedded into a sampling-based predictive controller, where state constraints and task specifications expressed by temporal logic formulas are incorporated into the cost function. So that, our control strategy ensures that the generated trajectories adhere to both safety constraints and high-level behavioral objectives with real time solving guaranteed. Real electric vehicle experiments validate the effectiveness of the proposed method and demonstrate its superiority in achieving safe and reliable planning and tracking control on icy roads.
- 8 February 2026
- Published by Elsevier in Expert Systems with Applications
- Vol. 313
ZPD-guided adversarial learning for safety-critical autonomous driving
- 8 February 2026
- Published by Elsevier in Expert Systems with Applications
- Vol. 313
ZPD-guided adversarial learning for safety-critical autonomous driving
- 16 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Vehicular Technology
- Vol. 75 (4) , 5505-5515
Interactive Motion Planning Using Genetic-Enhanced Reinforcement Learning With Backup Strategy
Navigating dense traffic environments, particularly during unprotected left-turn maneuvers, poses a critical challenge for autonomous vehicles due to the uncertainty of traffic participants, the trade-off between safety and efficiency, and dynamic interaction between vehicles. Existing methods usually fail to meet the constraints in changing environments or result in overly cautious behavior. This paper proposes a Genetic-Enhanced Reinforcement Learning with Backup Strategy (GERL-BS) to address these challenges by effectively navigating unprotected left turns in dense traffic. By leveraging the evolutionary capabilities of a risk-concerned Genetic Algorithm (GA), GERL-BS enhances the exploration of safety critical states by distilling essential interaction knowledge and control strategies, overcoming the limitations of Reinforcement Learning (RL) in such high-risk scenarios. Additionally, the Safety-Enhanced Contingency Backup (SECB) Module introduces a reward augmentation mechanism to ensure safety in complex and uncertain environments. To address the diversity and variability in vehicle interactions, we introduced the Heterogeneous Intelligent Driver Model (H-IDM), designed to simulate the heterogeneous behaviors of neighboring vehicles. Comprehensive evaluations in dense traffic scenarios with varying vehicle configurations demonstrate the proposed GERL BS controller's superior safety, adaptability, and effectiveness.
- 16 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Vehicular Technology
- Vol. 75 (4) , 5505-5515
Interactive Motion Planning Using Genetic-Enhanced Reinforcement Learning With Backup Strategy
Navigating dense traffic environments, particularly during unprotected left-turn maneuvers, poses a critical challenge for autonomous vehicles due to the uncertainty of traffic participants, the trade-off between safety and efficiency, and dynamic interaction between vehicles. Existing methods usually fail to meet the constraints in changing environments or result in overly cautious behavior. This paper proposes a Genetic-Enhanced Reinforcement Learning with Backup Strategy (GERL-BS) to address these challenges by effectively navigating unprotected left turns in dense traffic. By leveraging the evolutionary capabilities of a risk-concerned Genetic Algorithm (GA), GERL-BS enhances the exploration of safety critical states by distilling essential interaction knowledge and control strategies, overcoming the limitations of Reinforcement Learning (RL) in such high-risk scenarios. Additionally, the Safety-Enhanced Contingency Backup (SECB) Module introduces a reward augmentation mechanism to ensure safety in complex and uncertain environments. To address the diversity and variability in vehicle interactions, we introduced the Heterogeneous Intelligent Driver Model (H-IDM), designed to simulate the heterogeneous behaviors of neighboring vehicles. Comprehensive evaluations in dense traffic scenarios with varying vehicle configurations demonstrate the proposed GERL BS controller's superior safety, adaptability, and effectiveness.
- Open Access
Human-Centric Spatial Cognition Detecting System Based on Drivers’ Electroencephalogram Signals for Autonomous Driving
Existing autonomous driving systems face challenges in accurately capturing drivers’ cognitive states, often resulting in decisions misaligned with drivers’ intentions. To address this limitation, this study introduces a pioneering human-centric spatial cognition detecting system based on drivers’ electroencephalogram (EEG) signals. Unlike conventional EEG-based systems that focus on intention recognition or hazard perception, the proposed system can further extract drivers’ spatial cognition across two dimensions: relative distance and relative orientation. It consists of two components: EEG signal preprocessing and spatial cognition decoding, enabling the autonomous driving system to make more contextually aligned decisions regarding the targets drivers focus on. To enhance the detection accuracy of drivers’ spatial cognition, we designed a novel EEG signal decoding method called a Dual-Time-Feature Network (DTFNet). This approach integrates coarse-grained and fine-grained temporal features of EEG signals across different scales and incorporates a Squeeze-and-Excitation module to evaluate the importance of electrodes. The DTFNet outperforms existing methods, achieving 65.67% and 50.65% accuracy in three-class tasks and 84.46% and 70.50% in binary tasks. Furthermore, we investigated the temporal dynamics of drivers’ spatial cognition and observed that drivers’ perception of relative distance occurs slightly later than their perception of relative orientation, providing valuable insights into the temporal aspects of cognitive processing.
Open AccessHuman-Centric Spatial Cognition Detecting System Based on Drivers’ Electroencephalogram Signals for Autonomous Driving
Existing autonomous driving systems face challenges in accurately capturing drivers’ cognitive states, often resulting in decisions misaligned with drivers’ intentions. To address this limitation, this study introduces a pioneering human-centric spatial cognition detecting system based on drivers’ electroencephalogram (EEG) signals. Unlike conventional EEG-based systems that focus on intention recognition or hazard perception, the proposed system can further extract drivers’ spatial cognition across two dimensions: relative distance and relative orientation. It consists of two components: EEG signal preprocessing and spatial cognition decoding, enabling the autonomous driving system to make more contextually aligned decisions regarding the targets drivers focus on. To enhance the detection accuracy of drivers’ spatial cognition, we designed a novel EEG signal decoding method called a Dual-Time-Feature Network (DTFNet). This approach integrates coarse-grained and fine-grained temporal features of EEG signals across different scales and incorporates a Squeeze-and-Excitation module to evaluate the importance of electrodes. The DTFNet outperforms existing methods, achieving 65.67% and 50.65% accuracy in three-class tasks and 84.46% and 70.50% in binary tasks. Furthermore, we investigated the temporal dynamics of drivers’ spatial cognition and observed that drivers’ perception of relative distance occurs slightly later than their perception of relative orientation, providing valuable insights into the temporal aspects of cognitive processing.
- 31 December 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Intelligent Transportation Systems
- Vol. 26 (3) , 4178-4191
Equipping With Cognition: Interactive Motion Planning Using Metacognitive-Attribution Inspired Reinforcement Learning for Autonomous Vehicles
This study introduces the Metacognitive-Attribution Inspired Reinforcement Learning (MAIRL) approach, designed to address unprotected interactive left turns at intersections—one of the most challenging tasks in autonomous driving. By integrating the Metacognitive Theory and Attribution Theory from the psychology field with reinforcement learning, this study enriches the learning mechanisms of autonomous vehicles with human cognitive processes. Specifically, it applies Metacognitive Theory’s three core elements—Metacognitive Knowledge, Metacognitive Monitoring, and Metacognitive Reflection—to enhance the control framework’s capabilities in skill differentiation, real-time assessment, and adaptive learning for interactive motion planning. Furthermore, inspired by Attribution Theory, it decomposes the reward system in RL algorithms into three components: 1) skill improvement, 2) existing ability, and 3) environmental stochasticity. This framework emulates human learning and behavior adjustment, incorporating a deeper cognitive emulation into reinforcement algorithms to foster a unified cognitive structure and control strategy. Contrastive tests conducted in various intersection scenarios with differing traffic densities demonstrated the superior performance of the proposed controller, which outperformed baseline algorithms in success rates and had lower collision and timeout incidents. This interdisciplinary approach not only enhances the understanding and applicability of RL algorithms but also represents a meaningful step towards modeling advanced human cognitive processes in the field of autonomous driving.
- 31 December 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Intelligent Transportation Systems
- Vol. 26 (3) , 4178-4191
Equipping With Cognition: Interactive Motion Planning Using Metacognitive-Attribution Inspired Reinforcement Learning for Autonomous Vehicles
This study introduces the Metacognitive-Attribution Inspired Reinforcement Learning (MAIRL) approach, designed to address unprotected interactive left turns at intersections—one of the most challenging tasks in autonomous driving. By integrating the Metacognitive Theory and Attribution Theory from the psychology field with reinforcement learning, this study enriches the learning mechanisms of autonomous vehicles with human cognitive processes. Specifically, it applies Metacognitive Theory’s three core elements—Metacognitive Knowledge, Metacognitive Monitoring, and Metacognitive Reflection—to enhance the control framework’s capabilities in skill differentiation, real-time assessment, and adaptive learning for interactive motion planning. Furthermore, inspired by Attribution Theory, it decomposes the reward system in RL algorithms into three components: 1) skill improvement, 2) existing ability, and 3) environmental stochasticity. This framework emulates human learning and behavior adjustment, incorporating a deeper cognitive emulation into reinforcement algorithms to foster a unified cognitive structure and control strategy. Contrastive tests conducted in various intersection scenarios with differing traffic densities demonstrated the superior performance of the proposed controller, which outperformed baseline algorithms in success rates and had lower collision and timeout incidents. This interdisciplinary approach not only enhances the understanding and applicability of RL algorithms but also represents a meaningful step towards modeling advanced human cognitive processes in the field of autonomous driving.
- 24 December 2024
- Published by World Scientific Pub Co Pte Ltd in Unmanned Systems
- Vol. 14 (01) , 143-156
Equipping with Human Cognition: Driver Intention Recognition with Multimodal Information Fusion
To address the issue of autonomous driving systems’ inability to timely detect danger and respond in urban environments, an intention recognition system integrating driver cognitive information was developed. By utilizing the drivers’ electroencephalogram (EEG), eye movement and operational data, the system identifies driver intentions in a type of hazardous scenario. Initially, a driver-in-the-loop simulation platform was used for data collection, followed by experiments and data preprocessing to create a multimodal fused dataset through feature-level fusion. Models based on multilayer perceptron (MLP), convolutional neural network (CNN) and transformer were then developed to predict emergency braking and steering evasion intentions. The transformer-based model, with multimodal data fusion, achieved the best performance with an accuracy of 93.02%, significantly surpassing EEG-only (80.80%) and eye movement and operational data-only models (78.46%). This highlights the transformer’s superior ability to capture complex spatiotemporal correlations in multimodal data. Additionally, pre-extracted EEG frequency domain features could improve model performance, though less significantly than changes in model architecture. Embedding this system into autonomous driving systems is expected to enhance their ability to quickly and accurately recognize and respond to dangerous scenarios.
- 24 December 2024
- Published by World Scientific Pub Co Pte Ltd in Unmanned Systems
- Vol. 14 (01) , 143-156
Equipping with Human Cognition: Driver Intention Recognition with Multimodal Information Fusion
To address the issue of autonomous driving systems’ inability to timely detect danger and respond in urban environments, an intention recognition system integrating driver cognitive information was developed. By utilizing the drivers’ electroencephalogram (EEG), eye movement and operational data, the system identifies driver intentions in a type of hazardous scenario. Initially, a driver-in-the-loop simulation platform was used for data collection, followed by experiments and data preprocessing to create a multimodal fused dataset through feature-level fusion. Models based on multilayer perceptron (MLP), convolutional neural network (CNN) and transformer were then developed to predict emergency braking and steering evasion intentions. The transformer-based model, with multimodal data fusion, achieved the best performance with an accuracy of 93.02%, significantly surpassing EEG-only (80.80%) and eye movement and operational data-only models (78.46%). This highlights the transformer’s superior ability to capture complex spatiotemporal correlations in multimodal data. Additionally, pre-extracted EEG frequency domain features could improve model performance, though less significantly than changes in model architecture. Embedding this system into autonomous driving systems is expected to enhance their ability to quickly and accurately recognize and respond to dangerous scenarios.
- 24 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Intelligent Transportation Systems
- Vol. 25 (11) , 17602-17613
Cross-Observability Optimistic-Pessimistic Safe Reinforcement Learning for Interactive Motion Planning With Visual Occlusion
This study focuses on the motion planning and risk evaluation of unprotected left turns at occluded intersections for autonomous vehicles. In this paper, we present an interactive motion planning controller that combines Cross-Observability Optimistic-Pessimistic Safe Reinforcement Learning (COOP-SRL) and Nonlinear Model Predictive Control (NMPC), with consideration of the uncertain potential risk of occluded zone, the trade-off between safety and efficiency, and the dynamic interaction between vehicles. The proposed COOP-SRL algorithm integrates fully and partially observable policies through cross-observability soft imitation learning to leverage the expert guidance and improve learning efficiency. Moreover, the optimistic exploration policy and pessimism safe constraint are adopted to provide an adaptive safe strategy without hindering the exploration during learning process. Finally, the evaluations of the proposed controller were conducted in occluded intersection scenarios with various traffic density level, which indicate that the proposed method outperforms both the optimization-based and learning-based baselines in qualitative and quantitative indexes.
- 24 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Intelligent Transportation Systems
- Vol. 25 (11) , 17602-17613
Cross-Observability Optimistic-Pessimistic Safe Reinforcement Learning for Interactive Motion Planning With Visual Occlusion
This study focuses on the motion planning and risk evaluation of unprotected left turns at occluded intersections for autonomous vehicles. In this paper, we present an interactive motion planning controller that combines Cross-Observability Optimistic-Pessimistic Safe Reinforcement Learning (COOP-SRL) and Nonlinear Model Predictive Control (NMPC), with consideration of the uncertain potential risk of occluded zone, the trade-off between safety and efficiency, and the dynamic interaction between vehicles. The proposed COOP-SRL algorithm integrates fully and partially observable policies through cross-observability soft imitation learning to leverage the expert guidance and improve learning efficiency. Moreover, the optimistic exploration policy and pessimism safe constraint are adopted to provide an adaptive safe strategy without hindering the exploration during learning process. Finally, the evaluations of the proposed controller were conducted in occluded intersection scenarios with various traffic density level, which indicate that the proposed method outperforms both the optimization-based and learning-based baselines in qualitative and quantitative indexes.
- 4 March 2023
- Published by SAGE Publications in Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering
- Vol. 238 (7) , 1861-1879
Constraint performance slip ratio control for vehicles with distributed electrohydraulic brake-by-wire system
Slip ratio control of current passenger vehicles is achieved by the traditional hydraulic distributed unit through its pump-valve structure, such as the antilock brake system module. This type of hardware largely limits the essential braking pressure regulating ability due to the pump’s jittering backpressure and valves’ poor controllability and further deteriorates the slip ratio control performance. This study proposes a novel electrohydraulic brake-by-wire system based on distributed boosters (Dbooster) as new slip ratio regulating actuators. The Dbooster’s dynamics are modeled and combined with slip braking dynamics into an integrated slip ratio control scheme. Based on this, a finite-time prescribed performance control method with a finite-time disturbance observer is designed to achieve constraint slip ratio regulating performance. Validations are conducted in simulation and hardware-in-the-loop tests. Furthermore, the results show that this study provides a practical solution for more accurate slip ratio regulation.
- 4 March 2023
- Published by SAGE Publications in Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering
- Vol. 238 (7) , 1861-1879
Constraint performance slip ratio control for vehicles with distributed electrohydraulic brake-by-wire system
Slip ratio control of current passenger vehicles is achieved by the traditional hydraulic distributed unit through its pump-valve structure, such as the antilock brake system module. This type of hardware largely limits the essential braking pressure regulating ability due to the pump’s jittering backpressure and valves’ poor controllability and further deteriorates the slip ratio control performance. This study proposes a novel electrohydraulic brake-by-wire system based on distributed boosters (Dbooster) as new slip ratio regulating actuators. The Dbooster’s dynamics are modeled and combined with slip braking dynamics into an integrated slip ratio control scheme. Based on this, a finite-time prescribed performance control method with a finite-time disturbance observer is designed to achieve constraint slip ratio regulating performance. Validations are conducted in simulation and hardware-in-the-loop tests. Furthermore, the results show that this study provides a practical solution for more accurate slip ratio regulation.
- 25 March 2022
- Published by Elsevier in ISA Transactions
- Vol. 130, 629-654
Crash mitigation controller for unavoidable T-bone collisions using reinforcement learning
- 25 March 2022
- Published by Elsevier in ISA Transactions
- Vol. 130, 629-654
Crash mitigation controller for unavoidable T-bone collisions using reinforcement learning