Virtual Development and Validation of a Function for an Automated Lateral Control using Artificial Neural Networks and Genetic Algorithms

In this paper the development of a function for an automated lateral guidance by means of a consistently model-based and verification-oriented methodology of development will be presented. With a virtual test bench being used, the development process starts out with modelling the vehicle dynamics, the sensor system, and the environment. It is followed by a detailed layout of a self-learning driving function for the vehicle to stay in its lane on any single-lane road using artificial neural networks (ANN) and genetic algorithms (GA). The process is completed by a systematic analysis and validation of the thus developed driving function for lateral guidance.