Thursday Lunchtime Talk: Driverless Cars 😎
Thursday Lunchtime Talks (TLTs) are a long standing tradition at Osper. These are one-hour talks given by an Osper employee on a subject of their choosing. This can be anything. Previous subjects have included: PC games economies; classical music composition; someone’s adventure, involving a plane-crash, in Indonesia; or even a Price is Right game with Osper-related figures.
The aim is manifold. Everyone gets to learn about something new from someone in the team and also has access to that person to ask more questions later if they’re interested. Additionally, the person giving the talk can show everyone else something they’re interested in which the others might not know.
Shahriar, one of the software developers at Osper, gave a talk about driver-less cars. This was prompted by his attendance to an Udacity Course on the subject. It was a very engaging topic since most people have heard about self-driving cars, but might not know that much about what they actually are or how they work.
The presentation started with some pictures of different self-driving car designs, some built for research and some which are actually on the street today. Shahriar showed and explained the different parts that go on a car which needs to be able to sense its environment such as: GPS, radar, lidar, cameras and a lot of computing power.
The most exciting part of the presentation was a live demo and explanation of a simple algorithm for detecting lanes. One of the basic functions of a self-driving cars is for it to be able to know which lane it is traveling on and maintain course. Shahriar showed us how to do this with a few lines of code and we could also see the intermediate steps: separating the images from the car’s camera, making them grayscale, identifying the edges in the image, cropping the image to the area in front of the car, detecting lines and finally detecting which of the lines delineate the road lanes.
The most challenging part of the presentation was when Shahriar tried to explain, on a whiteboard, the part of the algorithm which detects lines in an image. This really tested everyone’s geometry knowledge. We got to looking at two graphs of dots in the picture represented in different dimensions which, when interposed would show the probability of the dots forming a line in the picture. It took us around 15 minutes to get our head around some simple geometry, but which conveyed some pretty advanced concepts.
The end was a short discussion by the audience on which companies are better positioned to roll out driverless cars and what the legal and moral implications would be for driver-less cars with issues such as: regulation, moral and legal responsibility, unemployment.
If you have any questions about this post or any Osper related queries at all, get in touch via the usual channels, for subject specific enquires email firstname.lastname@example.org 😃
Have a great day!
Ionuț and the Osper team