Perpendicular Autonomous Parking with Lidar Detection

Perpendicular Autonomous Parking with Lidar Detection

This project aims to achieve autonomous parking and pedestrian avoidance on a perpendicular parking spot with the help of optimization-based collision avoidance planning algorithm (OBCA) [1,2], MPC controller and Lidar detection.

First, the planning algorithm provides a path to a desired spot within a certain safety region based on differential GPS data. Then, the MPC controller tracks the path to ensure that the car parks into the spot while the on-board Lidar looks out for any potential obstacles or pedestrians that may cause a collision. If such pedestrian is present, the controller will stop the car and resume its course once the path is cleared.

The following clip shows the results in three scenarios: one without pedestrian, one where someone crosses the path when the car is going straight and one where someone crosses the path during the turn. These scenarios were chosen to guarantee safety under traditional parking settings.

Contact: Minglong Li (minglong.li@berkeley.edu), Xiaojing (George) Zhang

Collaborators: Yoonho Jang, Joohyung (Luke) Suh

Supervisors: Vijay Govindarajan, Xiaojing (George) Zhang

References

[1] X. Zhang, A. Liniger, F. Borrelli, Optimization-based Collision Avoidance, arXiv:1711.03449, 2017

[2] https://github.com/XiaojingGeorgeZhang/OBCA

Differential GPS for MPC-based Path Following