Autonomous driving is developing rapidly. A lot of breakthroughs of autonomous driving have emerged in both academy and industry. However, many traffic accidents related to autonomous driving also occur and cause people’s concern on the safety issue of AV. To ensure safety and reliability, rigorous test and simulation is required before AV can really drive on road. For AV test and simulation, realistic data is an essential component. Comprehensive, multi-regime and sufficient self-driving data would definitely help the AV development.
This project is to analyze Waymo and Lyft data, two newly opened datasets comprising of Level- 5 self-driving cars ( https://waymo.com/open/, https://level5.lyft.com/dataset/). Diverse sources of data are covered, including camera, lidar and radar. Our task is to identify the vehicle information from such multi-modal data, that is the position and velocity of each surroundings car. The first step is to explore the existing label in the dataset, which would already provide enough information for our traffic research. The second step, if time permits, is to apply the state-of-art detection algorithm to identify more car information. In the second step, computer vision techniques and statistics inference would be used.
Direct Supervisor: Sharon Di
Hours per week: 20 hr/week
Undergraduate students (especially female undergraduates) are welcome to work on this project during the summer. The student involved in this project will develop codes and algorithms for data analysis and modeling. Students with good computer and coding skills are preferred. Skill requirements are:
1. Familiar with reading and writing data.
2. Mastering Python programing is required.
3. Experience in processing image and lidar data is preferred. Using python to process and
transform image and lidar data, and generate figures, graphs, tables, or statistical models.
Desired outcome: codes and algorithms for data analysis