Zhengwei Bai
Ph.D. Student, in Electrical and Computer Engineering, University of California, Riverside
Transportation System Research (TSR) Group at Center for Environmental Research and Technology
Education:
M.S., Electrical Information and Engineering, Beijing Jiaotong University
B.E., Electrical Information and Engineering, Beijing Jiaotong University
I am currently a 2nd-year Ph.D. student in Electrical and Computer Engineering at UC, Riverside, co-advised by
Dr. Matthew J. Barth and
Dr. Guoyuan Wu. Before I started my Ph.D. study at UCR,
I received my Master's and Bachalor's degrees under the supervision of
Dr. Baigen Cai and Dr. Wei Shangguan at Beijing Jiaotong University.
My current research mainly focus on computer vision for enabling autonomous driving, e.g., 3D object detection, cooperative perception, multi-object tracking, etc.
Below are the strengths of mine:
1084 Columbia Avenue
Riverside, CA 92507
Email:
zbai012@ucr.edu
Find me at:
ResearchGate
GoogleScholar
YouTube
Slected Publications
Infrastructure-Based Object Detection and Tracking for Cooperative Driving Automation: A Survey
Zhengwei Bai , Guoyuan Wu, Xuewei Qi, Yongkang Liu, Kentaro Oguchi, Matthew J. Barth
arXiv preprint arXiv:2201.11871
[Paper], [CITE]
Cyber Mobility Mirror for Enabling Cooperative Driving Automation: A Co-Simulation Platform
Zhengwei Bai , Guoyuan Wu, Xuewei Qi, Kentaro Oguchi, Matthew J. Barth
101st Annual Meeting for Transportation Research Board (TRB2022)
[Paper], [CITE]
Hybrid Reinforcement Learning-Based Eco-Driving Strategy for Connected and Automated Vehicles at Signalized Intersections
Zhengwei Bai , Peng Hao, Wei Shangguan, Baigen Cai, Matthew J. Barth
System Architecture, Simulator View, Intersection View,
Accepted by IEEE Transactions on Intelligent Transportation Systems , 2022, doi: 10.1109/TITS.2022.3145798.
[Results Video with introduction on YouTube.]
Deep Reinforcement Learning Based High-level Driving Behavior Decision-making under Heterougeneous Traffic
Zhengwei Bai , Baigen Cai, Wei Shangguan, Linguo Chai
[Paper Link], [PDF], [CITE]
Accepted by Chinese Control Conference 2019
arXiv:1902.05772v2 [cs.LG]
[Results Video]
Deep Learning Based Motion Planning For Autonomous Vehicle Using Spatiotemporal LSTM Network
Zhengwei Bai , Baigen Cai, Wei Shangguan, Linguo Chai
[Paper Link], [PDF], [CITE]
2018 Chinese Automation Congress (CAC), Xi'an, China. , doi: 10.1109/CAC.2018.8623233
Oral presentation on the session of Unmanned Control System, [PPT]
[Results Video]
Design and Implementation of Novel Wireless Field Strength Test System
Zhengwei Bai
[Project Introduction]
Undergraduate Thesis
MAJOR PROJECTS
Cyber Mobility Mirror (CMM)
Advisors: Dr. Guoyuan Wu, Dr. Xuewei Qi, Dr. Yongkang Liu, Dr. Kentaro Oguchi, Dr. Wei Matthew Barth
We focus on building a 3D reconstruction mirror in cyber world based on the perception information from traffic mobilities to support cooperative driving automation (CDA). My research mainly focus on object detection and cooperative perception, which are briefly introduced as follows:
- To propose 3D object detection, classification, tracking and reconstruction methods.
- To build up a real-time field operational system to perceive the traffic objects and distribute and display the reconstruction information for enabling CDA.
- To develop different simulation paltform to support various kinds of perception requirements and model training.
Reinforcement Learning Based Eco-Approach and Departure (EAD) System.
Advisors: Dr. Peng Hao, Dr. Guoyuan Wu, Prof. Wei Matthew Barth
My focus is to propose a RL-based eco-driving strategy under intersections with mixed traffic. My main contributions are shown as follows:
- Propose a reinforcement learning framework that can deal with logically complex driving task, such as passing a signalized intersection with mixed traffic.
- Design and Develop a Unity-based reinforcement learning simulator: signalized intersection scenario with CAVs and human-driven vehicles.
- Design a time-efficient and eco-friendly driving strategy for CAVs by combining vision data, on-board sensor data and V2I communication.
Intelligent Transportation Information Management System
Advisors: Prof. Baigen Cai, Prof. Wei Shangguan, Dr. Dakai Yang
A large-scale transportation information management project I worked on in 2018. My focus is to design and develop several software modules including applications and APIs. My main contributions are shown as follows:
- Designed and Developed Five Software Modules of the ITIMS including bus, taxi, intersection, traffic signal,s and flow detection modules
- Developed several high-reliable multi-source data API (Application Programming Interface) which connects to major transportation information database such as The Traffic Police Database in Zhuhai.
- Management and maintenance overall service infrastructure utilizing remote controller(Using Teamviewer, Xshell and Xftp), Oracle & Mysql database.
Machine Learning Based Autonomous Vehicle Control Methods
Advisors: Prof. Baigen Cai, Prof. Wei Shangguan, Dr. Linguo Chai
A large-scale heterougeneous traffic simulation project I am currently working since 2018. My focus is to explore robust and accurate control methods for CAV in heterougeneous traffic situations. My main contributions are shown as follows:
- Proposed a deep neural network called spatiotemporal LSTM to generalize the steering angle output by fitting in the raw image data.
- Proposed a deep Reinforcement Learning (deep RL) based high-level driving behavior decision-making algorithm.
- Developed a simulation environment based on the Unity3D Engine for the training and testing processes of the deep RL algorithm.
High-speed Railway Based BeiDou Fusion Positioning Performance Test
Advisors: Prof. Baigen Cai, Prof. Debiao Lu, Prof. Wei Jiang
A testing about the performance of the GPS, BeiDou and IMU fusion positioning under high-speed motion scenario (350km/h). My main contributions are shown as follows:
- Setup the test environment on the train and using device SPAN, UB380 recorded IMU and GPS/BeiDou information separately.
- Designed the automatic driving test scenarios based on site characteristics (such as the car-flowing scenario, overtaking scenario, obstacles recognition, etc).
- Wrote the most part (over 80%) of the whole proposal (20134 words in total) and made a presentation to the local transportation department.
China Railway Urumqi Railway Administration Group Co.,Ltd – Field Strength Test system
Advisors: Prof. Baigen Cai, Prof. Wei Shangguan
A Field Strength Test System project that I worked on my last year of Undergraduate (2017). The main purpose of the project is to design and develope a system to test, record, and analyze the wireless field strength along the railway. My main contributions are shown as follows:
- Design and developed a MFC framwork based windows software (about 15000 lines of C++ code) for the novel wireless field strength test (FST) system .
- Developed and tested the hardware system (collecting and packaging the sensor data such as ODO, GPS and TAX) of the FST system.
- Testing the whole FST system between Urumqi railway station and Akesu railway station (2018km in total).
Major Competitions
ASCE T&DI Artificial Intelligence Student Competition, 2021
Honorable Mention
Cyber Mobility Mirror: Roadside LiDAR-based Real-time Object 3D Detection, Tracking and Reconstruction
Presentation Video
"Nokia Cup" Innovation Competition Final, 2016
1st Prize
Novel CTCS (Chinese Train Control System) simulation system
More information
"BJTU Huiguang Cup" Graduate Academic Culture Festival Essay Competition, 2018
2nd Prize
Design and Implementation of Novel Wireless Field Strength Test System
