RILYBOT 4: What's the Point?
The purpose of the RILYBOT 4 project is to demonstrate a simple and intuitive method of training robots, with the emphasis on showing the principle and less emphasis on precision or consistency.
This training technique allows the robot's user to push the robot around in the way he or she wants it to behave. The robot responds to each training with slightly different behavior, reflecting differences in speed, timing and direction.
For example, if the trainer pushes the robot quickly until it is in a dark spot, then leaves the robot still for a while, the robot will mimic that behavior by moving around quickly in the light and slowly (or perhaps even backwards) in the dark. It has become a dark-seeker.
The trainer can then retrain the robot to seek light by moving it slowly while it is in a dark area and quickly when in light.
Because linear approximations are used to model the trained movements, the resulting behavior is only a crude approximation of what was trained. For example, if it is trained to react to hitting a wall by moving backwards briefly, then stopping for a while and then going forward, the resulting behavior will be a more gradual type of motion: moving backwards, slowing gradually to a stop and then gradually accelerating forwards.
This page was written in the "embarrassingly readable" markup language RHTF, and was last updated on 2014 Dec 11. s.27