Collision Avoidance

The new rules for DIY Robocars (starting 9/21/19) call for all races to be head-to-head, i.e. two cars on the track at a time. Plus, under the new rules, and orange cone will be put onto the track in a random location, different for each race. I’ve been reasonably successful using OpenMV to find an orange blob and to avoid the cone. But what baffles me is how to see and avoid smacking into another car.

I’ve toyed with the idea of looking for floor colored blobs using invert=Yes so it would treat anything other than the gray carpet as an obstacle. But the floors I’ve tried are always mottled and don’t register as an identifiable color. What I get is a ton of false positives. Ideas?

I don’t think this is easy to do without some type of deep learning approach. I’d recommend instead to add some Ultra Sonic or 1d range sensors around the car and just avoid anything they see.