Hang Zhou
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
Hang Zhou has a total of 67 co-authors. Here are the top co-authors by number of shared publications:
- Chengyuan Ma
Department of Civil and Environmental Engineering, University of Wisconsin–Madison, 1415 Engineering Drive, Madison, WI 53706, USA
11 shared publications Xiaopeng Li
9 shared publications- Xiaopeng Li
University of Wisconsin–Madison
9 shared publications - Keke Long
Department of Civil & Environmental Engineering, University of Wisconsin-Madison, Madison, WI, United States
5 shared publications Ke Ma
The Hong Kong University of Science and Technology
4 shared publicationsPeng Zhang
4 shared publications
Latest Publications
- 5 October 2026
- Published by Elsevier in Transportation Research Part C: Emerging Technologies
- Vol. 194
Robust vehicle routing problem with flexible docking drones in humanitarian logistics
- 5 October 2026
- Published by Elsevier in Transportation Research Part C: Emerging Technologies
- Vol. 194
Robust vehicle routing problem with flexible docking drones in humanitarian logistics
- Open Access
- 4 October 2026
- Published by Elsevier in SSRN Electronic Journal
How Corner Is a Corner Case? Percentile Control for Highway Scenario Generation
Generating corner-case scenarios with appropriate adversity in a simulation environment is critical for testing an autonomous vehicle (AV) software stack’s safety performance before deployment. Existing autonomous-driving scenario generators can enforce specific behavior, adversity, or feasibility conditions, but they provide limited control over how extreme a generated scenario is relative to plausible futures in the same traffic context. This study represents the adversity of a generated scenario as its percentile in the conditional distribution of future risk given the observed history. This view supports calibrated answers to two questions: how ”corner” a generated corner-case scenario is and how its ”cornerness” can be fine-tuned.To this end, we formulate history-conditioned risk-percentile requests and learn a reference risk distribution that maps each requested percentile to a physical risk target. We then use a percentile-conditioned joint diffusion model with sampling-time risk guidance to generate multi-agent futures, together with a reference-based criterion for evaluating percentile realization.Experiments use the minimum post-encroachment time (PET) between the ego and its surrounding vehicles as the risk surrogate on highD. On the primary evaluation set, our method realizes 1,422 of 1,440 requests within a 0.05 percentile tolerance (98.75%), with mean percentile error 0.00673 and PET-target error 0.00991 seconds. The resulting interface connects context-relative risk specification, physical realization, and evaluation through a common risk scale. Project website and videos of generated scenarios are available at https://hhjj233.github.io/CornerPercentile/.
- 4 October 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessHow Corner Is a Corner Case? Percentile Control for Highway Scenario Generation
Generating corner-case scenarios with appropriate adversity in a simulation environment is critical for testing an autonomous vehicle (AV) software stack’s safety performance before deployment. Existing autonomous-driving scenario generators can enforce specific behavior, adversity, or feasibility conditions, but they provide limited control over how extreme a generated scenario is relative to plausible futures in the same traffic context. This study represents the adversity of a generated scenario as its percentile in the conditional distribution of future risk given the observed history. This view supports calibrated answers to two questions: how ”corner” a generated corner-case scenario is and how its ”cornerness” can be fine-tuned.To this end, we formulate history-conditioned risk-percentile requests and learn a reference risk distribution that maps each requested percentile to a physical risk target. We then use a percentile-conditioned joint diffusion model with sampling-time risk guidance to generate multi-agent futures, together with a reference-based criterion for evaluating percentile realization.Experiments use the minimum post-encroachment time (PET) between the ego and its surrounding vehicles as the risk surrogate on highD. On the primary evaluation set, our method realizes 1,422 of 1,440 requests within a 0.05 percentile tolerance (98.75%), with mean percentile error 0.00673 and PET-target error 0.00991 seconds. The resulting interface connects context-relative risk specification, physical realization, and evaluation through a common risk scale. Project website and videos of generated scenarios are available at https://hhjj233.github.io/CornerPercentile/.
- Open Access
- 14 September 2026
- Published by Elsevier in SSRN Electronic Journal
Safety Evaluation of Automated Vehicles Using Markov Chain Aggregation
Automated vehicle (AV) safety evaluation requires the evaluation of the occurrence and evolution of safety-critical rare states. Existing scenario-based evaluation methods use episodes with limited lengths to estimate accident rates, which fail to address how vehicles transition to dangerous states over time. This paper proposes a new perspective that analyzes the steady-state probabilities of states of interest over an infinite driving process, providing a comprehensive analysis of the distribution and evolution of near-collision states. By employing Markov Chain steady-state probability theory, state probability distributions are derived from state transition matrices, which can be directly extracted from tested AV's behavior model, thus avoiding the "curse of rarity". To address the high-dimensional state space issue, aggregate Markov Chain theory is introduced to cluster the state space for effective estimation. We aggregate microscopic states into macroscopic blocks, simplifying the high-dimensional microscopic state transition relationships into macroscopic block transition probability estimation to reduce the complexity. An iterative steady-state sampling method is designed to estimate transition patterns across states of varying danger levels. Simulation experiments compare the proposed method with the Monte Carlo benchmark, showing that the proposed method can estimate dangerous state probabilities using just 0.03% of the sample size and provide more valuable assessment insights than existing approaches. Compared with the state-of-the-art method, our method can reduce the number of AV calls for accident rate detection by two orders of magnitude to 10^5 .
- 14 September 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessSafety Evaluation of Automated Vehicles Using Markov Chain Aggregation
Automated vehicle (AV) safety evaluation requires the evaluation of the occurrence and evolution of safety-critical rare states. Existing scenario-based evaluation methods use episodes with limited lengths to estimate accident rates, which fail to address how vehicles transition to dangerous states over time. This paper proposes a new perspective that analyzes the steady-state probabilities of states of interest over an infinite driving process, providing a comprehensive analysis of the distribution and evolution of near-collision states. By employing Markov Chain steady-state probability theory, state probability distributions are derived from state transition matrices, which can be directly extracted from tested AV's behavior model, thus avoiding the "curse of rarity". To address the high-dimensional state space issue, aggregate Markov Chain theory is introduced to cluster the state space for effective estimation. We aggregate microscopic states into macroscopic blocks, simplifying the high-dimensional microscopic state transition relationships into macroscopic block transition probability estimation to reduce the complexity. An iterative steady-state sampling method is designed to estimate transition patterns across states of varying danger levels. Simulation experiments compare the proposed method with the Monte Carlo benchmark, showing that the proposed method can estimate dangerous state probabilities using just 0.03% of the sample size and provide more valuable assessment insights than existing approaches. Compared with the state-of-the-art method, our method can reduce the number of AV calls for accident rate detection by two orders of magnitude to 10^5 .
- Open Access
- 6 August 2026
- Published by Elsevier in Transportation Research Today
- Vol. 1
Benchmarking Tesla’s Traffic Light and Stop Sign Control: Field dataset and behavior insights
- 6 August 2026
- Published by Elsevier in Transportation Research Today
- Vol. 1
Open AccessBenchmarking Tesla’s Traffic Light and Stop Sign Control: Field dataset and behavior insights
- 1 June 2026
- Published by Elsevier in Accident Analysis & Prevention
- Vol. 234, 108607
Real-time lane-level abnormal traffic detection on freeways using sparse telematics data
- 1 June 2026
- Published by Elsevier in Accident Analysis & Prevention
- Vol. 234, 108607
Real-time lane-level abnormal traffic detection on freeways using sparse telematics data
- Open Access
- 4 April 2026
- Published by Elsevier in SSRN Electronic Journal
Robust Vehicle Routing Problem with Flexible Docking Drones in Humanitarian Logistics
Post-disaster humanitarian logistics face severe operational challenges due to time-critical demand, disrupted transportation infrastructure, and pervasive uncertainty. In recent years, truck--drone cooperative delivery systems have been widely considered as a potential approach to improve accessibility and responsiveness in such environments. However, most existing studies focus on conventional logistics settings and do not fully account for the distinctive characteristics of humanitarian operations, nor do they fully exploit the structural advantages of truck--drone systems. To address this gap, this paper proposes the Robust Vehicle Routing Problem with Flexible Docking Drones (RVRPFDD). The proposed model considers humanitarian logistics delivery using a truck--drone cooperative delivery system and incorporates a flexible docking mechanism that allows drones to be launched from one truck and recovered by another, thereby enhancing routing flexibility under infrastructure disruption. In addition, several humanitarian-specific constraints are explicitly modeled, including a hierarchical delivery deadline structure to reflect heterogeneous delivery urgency and robust energy consumption constraints for drones to account for uncertainty. A mixed-integer linear programming (MILP) formulation is developed for the problem, and an Adaptive Large Neighborhood Search (ALNS) algorithm is designed with tailored repair and destroy operators and a sub-route-level robust feasibility verification to accelerate the search process. Computational experiments show that, on small-scale instances, the proposed ALNS algorithm yields solutions that are at least as good as those of the MILP model and often better, while requiring substantially less computation time. On large-scale instances, ALNS also obtains better solutions than the benchmark heuristic methods. Beyond these algorithmic improvements, experiments also validate the benefits of the flexible docking mechanism and the robust energy consumption. Sensitivity analysis shows that increasing the drone fleet size and battery capacity significantly improves service coverage and delivery timeliness.
- 4 April 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessRobust Vehicle Routing Problem with Flexible Docking Drones in Humanitarian Logistics
Post-disaster humanitarian logistics face severe operational challenges due to time-critical demand, disrupted transportation infrastructure, and pervasive uncertainty. In recent years, truck--drone cooperative delivery systems have been widely considered as a potential approach to improve accessibility and responsiveness in such environments. However, most existing studies focus on conventional logistics settings and do not fully account for the distinctive characteristics of humanitarian operations, nor do they fully exploit the structural advantages of truck--drone systems. To address this gap, this paper proposes the Robust Vehicle Routing Problem with Flexible Docking Drones (RVRPFDD). The proposed model considers humanitarian logistics delivery using a truck--drone cooperative delivery system and incorporates a flexible docking mechanism that allows drones to be launched from one truck and recovered by another, thereby enhancing routing flexibility under infrastructure disruption. In addition, several humanitarian-specific constraints are explicitly modeled, including a hierarchical delivery deadline structure to reflect heterogeneous delivery urgency and robust energy consumption constraints for drones to account for uncertainty. A mixed-integer linear programming (MILP) formulation is developed for the problem, and an Adaptive Large Neighborhood Search (ALNS) algorithm is designed with tailored repair and destroy operators and a sub-route-level robust feasibility verification to accelerate the search process. Computational experiments show that, on small-scale instances, the proposed ALNS algorithm yields solutions that are at least as good as those of the MILP model and often better, while requiring substantially less computation time. On large-scale instances, ALNS also obtains better solutions than the benchmark heuristic methods. Beyond these algorithmic improvements, experiments also validate the benefits of the flexible docking mechanism and the robust energy consumption. Sensitivity analysis shows that increasing the drone fleet size and battery capacity significantly improves service coverage and delivery timeliness.
- 3 February 2026
- Published by Elsevier in Transportation Research Part C: Emerging Technologies
- Vol. 185
Real-time identification of cooperative perception necessity in road traffic scenarios
- 3 February 2026
- Published by Elsevier in Transportation Research Part C: Emerging Technologies
- Vol. 185
Real-time identification of cooperative perception necessity in road traffic scenarios
- 26 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Intelligent Transportation Systems Magazine
- Vol. 18 (2) , 49-54
Urban Vehicle Trajectory Dataset Based on Drone Videos
Along with the advancement of lightweight sensing and processing technologies, drones become powerful tools for traffic data collection. This article proposes a learning-based vehicle tracking framework from drone videos, addressing key challenges such as camera instability and dynamic lighting conditions. A dynamic keyframe adaptation strategy and a lighting mask are incorporated to enhance robustness. The method is validated using drone videos covering the transition from day to night over a 380-m corridor with three intersections. Approximately 5,000 vehicle trajectories were extracted at 10 Hz, capturing the formation and dissipation of the evening rush hour. This dataset is unique in its duration, signal timing information, and corridor-level coverage, making it the first publicly available resource of its kind. Experiments demonstrate the high accuracy of the approach, achieving a precision of 0.983 and a recall of 0.996. The labeled vehicle datasets, code, and resulting trajectory data are publicly available to support further research in traffic analysis and digital twin systems.
- 26 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Intelligent Transportation Systems Magazine
- Vol. 18 (2) , 49-54
Urban Vehicle Trajectory Dataset Based on Drone Videos
Along with the advancement of lightweight sensing and processing technologies, drones become powerful tools for traffic data collection. This article proposes a learning-based vehicle tracking framework from drone videos, addressing key challenges such as camera instability and dynamic lighting conditions. A dynamic keyframe adaptation strategy and a lighting mask are incorporated to enhance robustness. The method is validated using drone videos covering the transition from day to night over a 380-m corridor with three intersections. Approximately 5,000 vehicle trajectories were extracted at 10 Hz, capturing the formation and dissipation of the evening rush hour. This dataset is unique in its duration, signal timing information, and corridor-level coverage, making it the first publicly available resource of its kind. Experiments demonstrate the high accuracy of the approach, achieving a precision of 0.983 and a recall of 0.996. The labeled vehicle datasets, code, and resulting trajectory data are publicly available to support further research in traffic analysis and digital twin systems.