aanirudh07 / self-driving-2d-neat

This repository contains the implementation of a 2D self driving agent using NEAT to navigate a racetrack autonomously.

self-driving-2D-NEAT

This repository contains the Python code for creating a self-driving AI based on the NeuroEvolution of Augmenting Topologies algorithm. NEAT is used to evolve Artificial Neural Networks using genetic algorithm techniques.

• Each generation has 20 instances.
• Each instance is associated with a feed-forward neural network which learns how to navigate the track.
• Each instance is given a fitness value according to the how long it remained on the track.
• New generations are created until at least one car is able to finish a lap without leaving the track.

Packages Required

1. pygame

      pip install pygame
   

2. neat-python

      pip install neat-python==0.92
   

The Car

The car has a total of 3 sensors, positioned at -45, 0 and 45 degrees respectively. Each sensor has a maximum range of 200 pixels. The blue circles are used to detect if the car left the track (by detecting the green color of grass around the track).

The Network

A simple feed forward neural network with 3 inputs and 2 outputs is used. There are no hidden layers. The 3 inputs correspond to the sensor values. There were orignally four outputs: Left, Right, Speed up, Slow down. However, after some experimentation it was found that the later two increased the complexity of the network unnecessarily. The final network has only two outputs: Left and Right.

Tunable Parameters

The following parameters can be modified:

1. config.txt

      fitness_criterion
      fitness_threshold
      reset_on_extinction
      pop_size
      activation_default
      num_hidden
      num_inputs 
      num_outputs
   

2. main.py

      angle
      rotation
      velocity
      radar_angles
      translation constant in drive() method of Car class