Machine Learning based Self Driving Car Simulator along with Graphic User Interface (GUI)

This paper demonstrates a self-driving car simulator using machine learning along with Graphic User Interface (GUI). The car components with their corresponding physics, road structuring, traffic patterns, and neural network for predicting the next movement stage are realized using JavaScript programming. The GUI of the developed car simulator looks like the drone is filming it from the top of the car. The car has several sensors that can cast rays to detect objects. The distance metric obtained from the sensors is then fed to a neural network that will control the car's movement. The collision detection algorithm, a polygon intersection, uses segment intersection and linear interpolation to get the coordinates of the nearby objects. The segment intersection method works on any shape of objects and orientation in o(n2) time complexity that works in the developed simulation consistently. By enabling the testing of various scenarios and control improvements, the simulation model of a self-driving car is created that replicates the operation of a physical car driver using randomly generated weights and biases of neural networks. Here, the neural network provides evidence for decision-making so that the car can drive itself reliably. The developed model works on the traffic patterns it was trained in and can be further meshed together in the future with the help of different traffic pattern models to achieve higher safety requirements.