Neural model of the vehicle control system in a racing game. Part 2. Research experiments

Abstract

This article, which is a continuation of the article under the same main title and subtitle: part 1 Design and its implementation, includes the obtained results of research experiments with the use of a designed and implemented racing game. It uses a neural model of the vehicle motion control system on the racetrack in the form of a Perceptron Artificial Neural Network (ANN). In designing the movement of vehicles on the racetrack, the following were used, inter alia, Godot Engine and MATLAB and Simulink programming environment. The numerical data (14 input quantities and two output quantities) for ANN training were prepared with the use of semi-automatic measurement of the race track control points. This article shows, among others, the results of 10 selected research experiments, testing and simulation, confirming the correct functioning of both the computer game and the model of the neural control system. As a result of simulation tests, it turned out that the longest lap of the track in the conducted experiments lasted 4 minutes and 55 seconds, and the shortest - 10.47 seconds. In five minutes, the highest number of laps was 34, while the lowest numbers of laps were 1 and 5. In the course of the experiments it was noticed that under the same conditions the ANN learning outcomes are sometimes different.