Intelligent software controller for simulated car racing

Abstract

The primary objective of this project is to develop intelligent software for a simulation car in a car racing platform, in order to enhance the performance of the car. This project mainly divides into four development stages: exploration, extraction, transformation and application. A sensor system is used to receive the environment data from the simulation result of the platform and a tailor made tool called "Road Detector" is used to retrieve the overview information of the racing track. Some computations and interpretations are made to transform the data into meaningful information. This information is involved for the process of decision making. In this project, significant improvements are made in different aspects, for example, top speed on straight track, turning performance and turning strategies. On average, the intelligent agent takes about 40% less time to complete lots of racing tracks compared with the agent provided by the platform.