Cooperative Perception for Self-Driving Cars

From late 2018 through August 2020 I worked on cooperative perception for a pair of self-driving cars. The general idea is that one of the cars follows the other, with the follower having only low-cost sensors like cameras and radar. The problem is then to fuse perception information from both vehicles, dealing with network latency and measurement uncertainty while doing data association and filtering.

Here’s a video of the system running in simulation:

We had some physical vehicles that we tested on as well, but I don’t have video from those that I can share.