Using artificial neural networks to predict vehicle acceleration and yaw angles

The effectiveness of artificial neural networks in vehicle parameter prediction, using a variety of sensor data and artificial neural network (ANN) architectures, is outlined in this study in an attempt to determine its practicality for use in various controllers. To this end, the parameters to be used for ANN training and testing were chosen with regard to vehicle dynamics controllers, and the testing conditions representative of relevant driving conditions were also selected. The study also includes a brief description of the differences between the two ANNs used within the investigation, namely backpropagation and general regression neural networks. These ANN architectures were then used to gain predictions of lateral and longitudinal acceleration and yaw angle. Provided that these predictions showed sufficient accuracy, they could then be used in vehicle dynamics control systems, at a later date, to control parameters such as brake force and engine power to ensure vehicle stability.