Hi Cafa, I've completed engineering computing in uni last year,...

Hi Cafa,
I've completed engineering computing in uni last year, where for the major assignment we had to program a dot to self learn move across an x-y grid with obstacles taking the fastest route possible to its target.

I used Matlab as my program(not sure if you're familiar with that quite basic language) it took me and my partner one month and about 500lines of code to teach it. It took about 5mins to reach its target. Not only that but our dot was able to use the "undo" feature so it can take a step back and try another direction. In M1 you see the car starting from scratch. If I had to program that it would take days(maybe a month) to reach its target. I wouldn't even program it tbh I don't even know where to begin trying to think about it on the train now lol.

It's really not as easy as you make it sound, considering its self learning so you can't really teach an algorithm based on road width. The car picks a random direction on the x-y plane and then continues picking another x-y direction it continues this until it hits a wall. It then learns which x-y combination leads to it hitting at a wall and learns not to use that combination(via SNAP) relative to its position. As you can imagine there is literally millions of combinations it can move to hit a wall and it scans through all these combinations via trial and error.

This is how I imagine M1 to work so scouring through a million combinations in 2seconds is impressive if you have another suggestion let me know

M2 does it in 0.89seconds.
If you can beat that or even simulate an x-y grid and manoeuvre a dot to the other end that fast passed obstacles I'll put my money on you instead