Designing a Neural Network to Train a Bot to Traverse through an Arbitrary Course
In path planning operations, one of the fundamental issues is obstacle avoidance. If the machine can avert all
static or dynamic obstacles effectively, path planning will become easier and more accurate. Genetic and neuroevolutionary
algorithms have demonstrated to be effective for solving this optimization dilemma. These models mimic the
concept of natural evolution and have the aptitude to search progressive spaces and make choices in the most optimal way.
One direct application of these techniques is the development of automotive vehicles. In this paper we describe a neuro
evolutionary approach that proposes the evolution of chromosome attitudes that helps us generate a neural network which in
turn efficiently controls a simulated bot to avoid obstacles. The proposed project is a game that covers the interaction
between entities namely bots, terrain, static obstacles and other such conflicts. In order to maximize efficiency, we have used
multiple test cases enabling us to evaluate their results in a real‐ time scenario.
Keywords - Neural Network, Genetic Algorithm, Neuro Evolution, Obstacle Avoidance.