Jinnan Lu

End-to-End Autonomous Driving System: Design and Implement: 09.2023- 04.2024

Project Introduction: Develop an vision-based end-to-end autonomous driving system and conduct on-road tests.

Project Description: 

1. Drive the Neolix vehicle in the available industrial park and collect the driving data. 

2. Construct neural networks by modifying ResNet and Vision Transformer and build a navigation system based on the A* algorithm.

3. Train the models with the collected data, evaluate the model, deploy the model on the unmanned vehicle, and conduct tests.

Combining Monte Carlo Tree Search and Deep Reinforcement Learning in Behavior Planning for Autonomous Driving: 09.2022- 03.2023

Project Introduction: Combining DRL and MCTS to train ego vehicle to safely pass through intersection scenario.

Project Description:

1. Use SUMO software to create an intersection scenario and introduce random traffic flow to simulate the traffic flow in real world.

2. Build a DQN agent and train the DQN agent to enable the ego vehicle to safely arrive at the destination from the starting point.

3. Build MCTS and replace the random strategy of MCTS with the trained DQN, so that DQN can lead the search direction of MCTS.

4. Evaluate the models and conduct simulation tests. 

Deep Reinforcement Learning in Behavior Planning for Autonomous Driving: 03/2022- 09/2022

Project Introduction: Use DRL to train ego vehicle to safely pass through roundabout scenario.

Project Description:

1. Use SUMO software to create a roundabout scenario and introduce random traffic flow to simulate the traffic flow in real world.

2. Build a DQN agent and train the DQN agent to enable the ego vehicle to safely arrive at the destination from the starting point.

4. Evaluate the models and conduct simulation tests.