01World models
Digital twins &
HD mapping
Reconstruct simulation-ready roads and lane-level vector maps from field and global geospatial data.
About
Hao Zhang
Ph.D. Researcher in Autonomous Driving Systems & Transportation Safety
I build learning, simulation, and mapping systems for safer and more capable autonomous vehicles.
I am a Ph.D. researcher in Civil and Environmental Engineering at Texas A&M University. My work spans autonomous-driving simulation, digital twins, HD map learning, vehicle control, V2X scenario generation, and safety analysis.
My goal is to work on autonomous driving systems that perform well not only in benchmarks, but also in the complexity of real traffic. My transportation-engineering background gives me a system-level perspective on how vehicles interact with roads, infrastructure, human drivers, and safety constraints.
Technical profile
Hao Zhang · AV systems researcher
From world models to
safe autonomous driving.
Model the world→Generate hard cases→Design behavior→Measure safety
Digital twins &
02Hard cases
Scenario
engineering
Build synchronized mixed-traffic worlds and turn road or language intent into repeatable critical tests.
03Vehicle behavior
AV control &
decision-making
Design car-following and lane-change control across continuous and discrete actions.
04Risk measurement
Safety analysis &
validation
Stress-test AVs in high-risk interactions with physics-aware, multidimensional, and stochastic methods.
Data sourcesAV datasets, OSM, LiDAR, imagery
Vector mapsLane topology, semantics, infrastructure
Mixed trafficVehicles, pedestrians, cyclists, signals
Behavior & intentPrediction, control, trajectory forecasting
Risk & safetyMetrics, scenario testing, validation
Selected research
Research for real-world autonomy.
I focus on three practical bottlenecks for autonomous driving: building scalable world models, creating useful simulation scenarios, and evaluating safety under realistic traffic interactions.
01
Virtual Roads, Smarter Safety
A digital-twin framework for reproducing mixed autonomous traffic and measuring safety with synchronized traffic flow, vehicle physics, and perception-level simulation.
Why it matters for AV development
It provides a scenario-to-metric workflow for testing vehicle behavior under realistic geometry, traffic demand, and physical conditions.
EnvironmentLiDAR · OSM · vehicle data
→Co-simulationCARLA · SUMO · PhysX
→ValidationBehavior · dynamics · safety
02
Learning HD Digital Twins from Global Maps
A multi-modal learning framework that transforms satellite imagery and imperfect OpenStreetMap priors into lane-level vector maps with geometry, semantics, and topology.
Why it matters for AV development
It explores a scalable route for bootstrapping and updating HD map coverage without relying exclusively on repeated survey-vehicle collection.
2,341aligned map tiles
0.890geometry F1 at 2 m
0.879adjacency F1
4 citieswithin- and cross-city tests
03
V2XSynth
An LLM-orchestrated, retrieval-augmented framework that converts natural-language traffic intent into communication-aware CARLA scenarios with executable behavior and V2X fault models.
Why it matters for AV development
It reduces the effort required to author diverse simulation scenarios and links communication failures—latency, loss, and jitter—to driving safety outcomes.
Prompt
Describe the traffic, agents, environment, and communication conditions.
LLM + RAG compiler
Retrieve validated behaviors, maps, and V2X modules.
CARLA execution
Run the scenario and collect joint mobility and message telemetry.
Publications
Selected work
20262026
Integrated Automated Car Following and Lane-changing Control Based on a Parameterized Deep Q-Network with Hybrid Action Space
H. Zhang, Z. Li, and Y. Zhou
arXiv preprintVirtual Roads, Smarter Safety: A Digital Twin Framework for Mixed Autonomous Traffic Safety Analysis
H. Zhang, X. Yue, K. Tian, S. Li, K. Wu, Z. Li, D. Lord, and Y. Zhou
IEEE Internet of Things Journal2025
Why Anticipatory Sensing Matters in Commercial ACC Systems Under Cut-in Scenarios
H. Zhang, S. Li, Z. Li, M. Anis, D. Lord, and Y. Zhou
Accident Analysis & Prevention2025
2025On the Collision Risk of Stochastic Traffic: Analytical Approximation in 2D and 3D Spaces for Generic Vehicles
H. Zhang, K. Wu, Z. Li, Z. Tu, D. Lord, and Y. Zhou
Presented at TRB 2026V2XSynth: An LLM-Orchestrated and Retrieval-Augmented V2X Scenario Generator
K. Wu, H. Zhang, P. Li, R. Gan, J. You, Z. Tu, and Y. Zhou
Presented at TRB 20262022
A Temporal Fusion Transformer for Short-term Freeway Traffic Speed Multistep Prediction
H. Zhang, Y. Zou, X. Yang, and H. Yang
NeurocomputingBackground
Education & experience
Texas A&M University
Ph.D. in Civil and Environmental Engineering
Research / Teaching Assistant · 2023–Present
Tongji University
M.S. in Transportation Engineering
Research Assistant · 2020–2023
Central South University
B.S. in Logistics Engineering
2016–2020Contact
Interested in autonomous driving research and engineering.
I welcome conversations about AV simulation, mapping, safety validation, and intelligent mobility systems.
haozhangch9811@gmail.com