Hands-On Neuroevolution with Python

Sensor data generation

The maze-solver agent is controlled by an ANN that needs to receive sensor data as input to produce the corresponding control signals as output. As we have already mentioned, the navigator agent is equipped with an array of two types of sensors:

• Six rangefinder sensors for the detection of collisions with the maze walls, which indicate the distance to the closest obstacle in a particular direction.
• Four pie-slice radar sensors, which indicate the direction to the maze's exit point from any place in the maze.

The sensor values need to be updated at each simulation step, and the MazeEnvironment class provides two designated methods that update both types of sensor.

The array of rangefinder sensors is updated as follows ...