Nearly a year ago, I was in the process of applying to grad schools. I needed a total of 3 recommendation letters and I already had 2 with people that have worked with me in a research environment. I needed one final letter and to make this letter carry a lot of value in the admissions process, I needed to do some research for my professor. I had met this man ( I will refer to him as Professor Doe from here on out) in my linear systems class the previous spring semester. Professor Doe and I had a very strong professor-student relationship. I did well on his exams and participated in class (even when it made me look foolish) . He and I would always banter back and forth over silly conversations. Overall, we liked each other a lot. This was his first semester at my undergraduate university; he was in the process of establishing a new robotics program there and was required to teach one class a semester.
After I finished my final exam, he asked me to contact him if I was interested in working on any projects dealing with robotics. So there I was in the Fall of 2016, telling him I was ready to some research for him and in return he would write me a letter. I did not imagine that he would assign me (some 20 year old undergrad who was just entering his 3rd year of post-high school education) to working on his autonomous car project. To be honest, I did not even know he had acquired an autonomous car until he told me (I think the university was waiting for a grand reveal the following semester). I felt nervous taking up a project with so much and many dire consequences if it failed. However, I agreed to work on this project.
My first major assignment was to set up some GPS communication systems for the autonomous car using Real-Time Kinematic satellite navigation. It might not sound glorious, but this was a difficult task. This kind of satellite communication relies on a base and receiver pairing with strong signal strength between satellites and base as well as satellites and rover. In a wide open area, there is little to no issue with this kind of communication. In the crowded streets of Boston, however, there are tons of issues getting this to work functionally.
One of the major problems is that large buildings and buildings that are made of brick/have large glass windows introduce all kinds of noise to these communication systems. It's very difficult to get a high signal strength and therefore difficult to pinpoint GPS coordinates over a drive. However, there are mechanisms that will do the job as well. Simpler GPS and IMU devices that can get the job done with some correction applied with directional vector calculation and interpolation of GPS coordinates.
From that point on, I worked on a horizon image localization project that essentially allowed us to map GPS coordinates on one run to horizon images and then regenerate those GPS coordinates using the horizon lines of the first run and comparing them. This process is a little complicated and it would require another post to accurately describe. While driving around the car, we were restricted in our ability to drive in an autonomous fashion.
While collecting data, the school did not want the car to be caught doing so because the city of Boston would have had a problem with it and the university would face penalties they did not want to deal with. That's beyond the fact that there are obvious safety risks with an autonomous car. As a society, we still struggle with coming up with minimal risk driving strategies if harm is inevitable. It's honestly one of the things that is holding us back from having many of these autonomous systems everywhere.
Thanks for reading and if you enjoyed this blog, feel free to upvote, resteem, and follow me. I post about everything from sports to my life experiences and technology.
