Research Experience

Risk-Guided Adversarial Policy Optimization

We proposed a risk-guided adversarial policy optimization method for safety-critical scenario generation, which generates more diverse and plausible scenarios more efficiently.

Adversarial Traffic Participants Behavior Model

We proposed an adversarial traffic participant behavior model combining traffic prior and reinforcement learning, which solves the limitation that adversarial scenario generation can only be applied to specific working conditions.

Safety-Critical Scenario Generation Method Combining Dynamic and Static Scenario Parameters

We proposed a safety-critical scenario generation method combining dynamic and static scenario parameters, which greatly improves the efficiency of scenario generation.

Improve the Diversity of Safety-Critical Scenarios

We are working on quantifying the diversity of scenarios and improving the diversity in safety-critical scenario generation for a comprehensive evaluation of autonomous vehicles.

Improve the Naturalness and Plausibility of Safety-Critical Scenarios with Naturalistic Driving Data

We are working on combining naturalistic driving data with adversarial scenario generation to improve the naturalness and plausibility of the generated scenario.

Safety Analysis and Evaluation in Complex Scenarios

We proposed an approach for the safety analysis of autonomous vehicles from complex safety-critical scenario data, which can intuitively reveal the distribution and characteristics of safety-critical scenarios for any given algorithm.

Safety Boundary Identification in Long-Tailed Environment

We proposed a safety boundary online identification method that learns from the test data under a long-tailed environment.

Data Closed Loop

We reviewed the research related to data-closed loop and proposed a self-evolution mechanism for the autonomous driving algorithm.

Self-Evolve Mechanism for Decision-Making Algorithms

We are working on the closed-loop self-evolve mechanism for autonomous driving algorithms under safety-critical scenarios.

Mixed-Reality Testing

We are working on constructing a mixed-reality simulation test platform by combining vehicle hardware-in-the-loop and generated virtual traffic scenarios.

Design of New Generation Chassis Vehicle and Analysis of Its Dynamic Performance

I designed a new chassis with four-wheel independent drive and four-wheel steering.

I designed an algorithm for the torque distribution and steering angle control of each wheel considering maneuverability and stability.

I conducted the vehicle dynamics simulation and its performance analysis based on Matlab/Simulink and CarSim.

Design of Automated Lane-Changing System for Intelligent Vehicle and Its Hardware-in-the-Loop Testing

I designed a lane-change decision-making algorithm combining risk and efficiency and a lane-changing trajectory planning and control algorithm considering driving comfort and stability.

I designed a hardware-in-the-loop testing platform, and tested and evaluated the designed automated lane-changing system.