Using hyperneat for robot maze navigation simulation
- Hi All,
Would HyperNeat work well for the following problem?
I want to do a robot simulation of maze navigation. The robot inputs will have two 'eyes', that are represented as 32 x 32 retinas. The eyes see in two different directions, the right and left walls of the maze. It would have a sensor that is true when the robot is touching a forward facing, non-visible wall. The 'colors' the robot can see are red, green, and blue. Not all of the gradations of these colors, each element is just one of these three colors, with green being about 50% of the field. There would be four outputs: right, left, forward and backward. The walls would have symbols, maybe arrows, that point to the way out of the maze. The arrows may be in different colors on different walls, but always visible to a human with normal vision.
The robots task would be to navigate out of the maze.
1. How would I encode the 3 colors, -1, 0, +1 ?
2. Would the c++ version be much faster or better than the java version?
Any other suggestions on how to encode and decode the environment?