Research

Research interests: Driving Simulator Experiment, Human-Computer Interaction, Trajectory Prediction, and Traffic Simulation.

RESEARCH EXPERIENCE

A driver trajectory prediction method addresses the false alarm of the active safety system by considering the influence of the driver’s abnormal emotions.
The prediction model takes into account both the physical movement features and cognitive features, and the accuracy of the prediction method is verified through realistic driving simulator experiments.
The stimulus-organism-response theory is referred to describe the influence mechanism of abnormal emotions on driving behavior. The driver’s dynamic cognitive features are further extracted.
Compared with advanced models, the proposed trajectory prediction method has significantly lower errors, especially in the scenarios of unprotected left turns and the sudden braking of the front car.
This method can be integrated into the active safety system, making the system better adapt to drivers in abnormal emotions, and effectively reducing false alarms.

A microscopic traffic simulation calibration algorithm based on parallel computing to solve the problem of long time and low efficiency of the heuristic calibration algorithms.
A microscopic simulation model was built on SUMO. Parallel computing technology is applied to heuristic calibration algorithms so that the simulation model can accurately reproduce actual traffic flow.
Implemented the parallelization of the genetic algorithm (GA) and particle swarm optimization (PSO) calibration algorithm Following the three steps of parallel framework selection, algorithm bottleneck identification, and subtask load balancing.
The proposed parallel calibration algorithm can shorten the calibration process from 5 hours to less than 1 hour, reducing the calibration time by 80%.
The results of this study help to achieve rapid calibration of microscopic traffic simulation parameters.

Established a multi-style driver model of left-turn vehicle interaction at the intersection based on the actual collected trajectory data to meet the heterogeneity of drivers in the actual traffic environment.
Deployed the established driver model in VTD simulation software and provided a test environment that can adjust the heterogeneity of traffic flow for the autonomous driving algorithm.