Self Steering Vehicles (Neuro Evolution)

Demo

Implemented by applying NeuroEvolutionary Algorithm (Neural Network + Genetic Algorithm)

Here the Entities are as follows :

1. Particle / Vehicle (The agents in the algorithm)

• Each particle is associated with a velocity and has rays emitting from it which is fed as input
• Each particle is associated with a brain (Neural Network) that steers the particle / vehicle.
• The next generation emerges by mutating the weights of the fittest parents in the previous generation.
• The fitness function is defined as the inverse square of the distance from the final point.

2. Boundary (Track)

• The Track is generated randomly using Perlin Noise
• After each generation of particles a new track is generated inorder to avoid memorization of track by agents.

Play with the variables in

public/sketch.js

var TOTAL = 100 // No of agents generated during each generation
var MUTATION_RATE = 0.05 //Mutation Rate (Probability of undergoing mutation)
var showRay = false //(true/false) Set to true in order to see the rays
var total = 30 // No of points used for random path generation using Perlin Noise .Greater the value of total the more complex the track is generated.

Libraries :

• Tensorflow.js
• p5.js

TODO:

• Improve the UI.Add User Input to adjust the variables.