Research | Chenzhu Wang

Proactive Crash Risk Prediction and Prevention

Mon, 01 Jan 0001 00:00:00 +0000

Real-time risk and crash-type prediction

Connected-vehicle trajectories and risky-driving events offer information beyond fixed traffic detectors. My work links these observations with traffic states to predict both crash occurrence and crash type.

The published BiLSTM-Transformer study examines spatiotemporal prediction using connected-vehicle data. Machine learning with causal mediation provides a complementary way to investigate the pathways behind estimated risk.

Behavior-aware prevention

Current work examines driver-group differences, perception, and interactions between roadway environment and driving behavior. The submitted BusMEP manuscript addresses bus safety-risk prediction; it is not a study of autonomous buses.

A separate submitted deep-ensemble and causal-learning study focuses on multi-source freeway safety assessment.

Safety and operations

Speed-management measures of effectiveness connect risk evidence with operational decisions. My FDOT PI proposal is pending approval, while related published studies examine tunnel crash severity and congestion duration and post-crash speed forecasting.

Methods: spatiotemporal sequence learning, causal mediation, interpretable models, and multi-source data integration.

Crash Mechanisms, Human Factors, and Roadway Design

Mon, 01 Jan 0001 00:00:00 +0000

Crash and injury mechanisms

Safety relationships can change across road users, locations, and periods. My work uses spatial models, temporal analysis, and random-parameter approaches to examine this heterogeneity, rather than assuming one average effect transfers to every setting.

Intersection crash frequency and pedestrian injury severity across vehicle movements are complementary examples.

Human factors and vulnerable road users

Behavioral evidence helps explain how risks develop. I study driver perception and reaction, workload, pedestrian distraction, and vehicle-pedestrian interactions.

The vehicle-pedestrian interactions review connects crash analysis with conflict assessment. Related intersection interaction research examines right-turning vehicles and pedestrians using conflict and crash datasets.

Roadway design and environmental context

My earlier engineering practice and plateau-road research inform a continuing interest in how physical road conditions interact with human capabilities. Studies of perception-reaction time and minimum horizontal curve radius link behavioral and psychophysiological evidence with design questions.

Methods: causal analysis, Bayesian and random-parameter models, spatial effects, behavioral measurement, and psychophysiological assessment.

Autonomous Driving, ADAS, and CAV Safety

Mon, 01 Jan 0001 00:00:00 +0000

I examine ADAS and connected and automated vehicle safety through crash-data analysis, simulation, and vehicle-dynamics-informed evaluation. Driving assistance and automated driving are distinct capabilities; safety claims must be tied to the technology and operating conditions being evaluated.

Safety-critical decision-making and recovery

Current collaborative research asks how vehicles can recover from developing hazards. The submitted ALARM study concerns local action safety boundaries and pre-crash safety recovery. A separate controller-state recovery study examines resilience during CAV-HDV on-ramp merging.

These are submitted research manuscripts, not claims of validated deployment performance.

Cyber-physical resilience and vehicle dynamics

Communication failures and cyberattacks can affect vehicle behavior through physical constraints. The submitted tire-force risk-envelope study addresses CACC platoons under cyberattacks and reduced-friction roadways.

Related work includes a mixed-traffic cyberattack simulation platform and connected-platoon impact assessment. These studies connect information disturbances with vehicle dynamics and collision risk.

ADAS effectiveness and infrastructure compatibility

The published ADAS effectiveness study evaluates safety using real-world crash records.

As PI of an awarded FHWA project, I lead work on roadway design and infrastructure factors influencing ADAS technologies for mitigating rural roadway departure crashes. This complements the technology-focused work with a road-engineering perspective.

Methods: crash-data evaluation, simulation, causal analysis, vehicle-dynamics constraints, and safety-boundary assessment.

Multimodal Traffic Scene Understanding

Mon, 01 Jan 0001 00:00:00 +0000

From traffic video to safety evidence

My published vision-language and generative-model review organizes research on interpreting traffic videos and examines grounding, temporal consistency, and verification.

The aim is to connect semantic descriptions with observable traffic evidence. Model-generated explanations and scenarios require checks against the underlying events and physical constraints.

Datasets and timing-sensitive evaluation

Two separate SAVeD research outputs address complementary questions:

The SAVeD dataset preprint concerns first-person ADAS near-miss and crash video data.
The submitted SAVeD deployment-oriented protocol concerns candidate-event warnings, video and structured signals, and the timing of crash-risk escalation.

They are distinct works, rather than a replacement title for the same record.

Occluded pedestrian crossing

The submitted multimodal risk-recognition study addresses unsignalized, occluded pedestrian crossing scenarios and graded prevention. It connects this research area with vulnerable-road-user safety and proactive risk assessment.

Verification and research resources

Benchmark and protocol design must distinguish observed events from generated scenarios, and warning evaluation must consider event timing. Related work includes the SAVeD dataset, warning protocol, ALARM benchmark, and MSC-MTSim simulation platform.

Methods: multimodal learning, vision-language models, structured scene representations, temporal event analysis, and evidence-based evaluation.