I have started building out a "robot lab" in which to test several of my theories on basic capabilities that I want to give ARTUR's initial AI before putting it into the transmutation routine. The lab consists of a robot with four "bumper" sensors, four "thruster" motors, and a maze which can be defined in a simple text array.
So far I've got the maze and robot functioning (just a slight hit detection bug I need to work out). I've gotten the neurons hooked up for the bumpers, thrusters, and position. The HTM layers still need to be hooked up.
For now, the robot is simply controlled by the arrow keys. Obviously the goal is to hook up HTM layers and implement reinforcement learning to have it learn and drive itself to complete the maze. Lots of potential tests that can be done with this app, so should get pretty interesting.
The WIP app can be accessed HERE
. I'll put up the source code on GitHub once I have finished writing all the remaining basic components.