In this chapter, we consider a two-dimensional maze navigation task. This task is easy to visualize, and it is relatively easy to write the simulator of the maze-navigating robot for a two-dimensional maze. The main goal of the robot is to navigate through a maze to the defined goal point in a specified number of time steps. The ANN that controls the robot is a product of the neuroevolution process.
The neuroevolution algorithm starts with a very basic initial ANN configuration that only has input nodes for sensors and output nodes for actuators, which gradually becomes more complex until a successful maze solver is found. This task is complicated by a peculiar configuration of the maze that has several cul-de-sacs
