Replace computer vision system with two pigeons.

[TomFraser]

I propose a change to replace the complex and unnecessary machine learning based computer vision system for cone detection with a much more simplistic two pigeon approach. (See attached image)

Advantages of a two pigeon approach:

• Training a pigeon is easier than an ML model
• Pigeons are a readily available resource (can collect local pigeons at track and train before comp)
• Pigeons are much more flexible than a model, they don't care about cone size or lighting conditions
• Bird feed is cheaper than a ZED camera
• Who doesn't like birds???
• The bird stack is

The high level tasks to complete are as follows:

1. Acquire two pigeons
2. Train each pigeon to expect a food reward when it pecks either a blue or yellow cone respectively
3. Build a cage-like structure to house each pigeon slightly angled away from the center line of the car
4. Place a clear touch sensor in front of the pigeons and track their pecks as cone locations
5. Develop an algorithm to determine the best RVWP from the pigeon pecks

Inspired by: https://en.wikipedia.org/wiki/Project_Pigeon