Today is a very special day for us. We have successfully tested the autonomous steering system for the self-driving cart. Here is a video.
For the past 27 days (not counting weekend only 19 days), Michael and I have been working on the autonomous steering system for the self-driving golf cart. Frankly, I have never been more genuinely excited about an engineering project. Our challenges extend from engineering a mechanism to control the steering with an electric motor, to tuning the software so it handles driving smoothly. Today, most of these components worked very well.
The testing site is located behind the library. The road extends from the edge of the English building all the way to the track. The vehicle handled more than 4 significant turns with no human steering input at all. On a relatively straight road, I was able to avoid running into cars and the side of the road for more than 200 meters. (Until it got lost).
After some extensive testing, we have noticed a couple of things.
- The vehicle tends to stay on the side of the road. But it follows the road well.
- The vehicle works generally well under normal lighting and good color contrast.
- The vehicle tends to correct itself a little bit too late.
- The system doesn’t work very well with no lane markings. When encountering dirt/ground with no snow and no grass, the system tends to fail.
The project is far from being finished. We are just getting started. We will try to address the problems above. Now, we also need to integrate more machine learning models into the system. Most of them are already trained. The next engineering challenge is to interface with the breaking system electronically, which will allow us to test the cruise control program.
Thanks again to Emily Richardson, Peter Nilsson and Megan Hayes-Golding and everyone else for their tremendous help. Please leave a comment below if you have any questions. You can contact me directly at firstname.lastname@example.org. Thanks!