For our science project in our last year of high school we were free to make whatever we wanted. We decided to make a small, autonomous, robot-car.

The first goal was to make a car that would find it's way out of a maze. This was far too easy after realizing the best method to get out of a maze is to always turn right or left. We had a car that was able to do this in about 10 hours of work.

This is what it looked like, just a very simple Arduino car with only one sensor.

The guideline for this project was 120 hours (per person) of work which meant we had to upscale our project. We did this by changing our subject to machine learning. We decided to make a car with image recognition. Our new goal was a car that would recognize arrows and follow the direction of the arrow.

Needing more space for parts like a Raspberry Pi, a camera, a distance sensor and many more we had to upgrade our platform. I made a 3D design of our new platform and had it 3D-printed at my sidejob.

These are the renders. I made a system to make attaching components easy by allowing many different mounting options like using rubber bands but also designated mounts for each component. In every crossing there's a hole to route all the cables. 

This is the final product with all the components. Because the frame looked like Wall-E we made a housing for it. The car works very well and recognizes the arrows correctly every time.

Here it is in the showcase at our school after the final presentation. In total we spent 290 hours making this car (together with a classmate, total of 2 persons)

0:40 - the car starts driving
1:15 - the communication between the Arduino and Raspberry Pi went wrong, the blinking LED indicates a second attempt.
3:00 - the car wants to go right, as the arrow indicated, but has a wall next to it. The car looks around and sees that going left is an option and turns this way.