Quantification of human driving skill for human adaptive mechatronics applications by using neural network system

Abstract

Human Adaptive Mechatronic (HAM) is a two way relationship between human and machine. In HAM, not only human need to adapt the characteristics of machine but the machine also has to learn on human characteristic. HAM is an enhance system for Human Machine System (HMS) which consists only one way relationship between human and machine. As a part of mechatronics system, HAM has an ability to adapt with human skill improve the performance of machine. The example of application where HAM can be applied is driving a car. One of the important elements in HAM is the quantification of human skill. Thus, this project proposed a method to quantify the driving skill by using Artificial Neural Network (ANN) system. Feedforward neural network is used to create a multilayer neural network and five models of network were designed and tested using MATLAB Simulink software. Then, the best model from five models is chosen and compared with other method of quantification skill for verification. All simulation data are taken from M. Hafis Izran’s in his PhD thesis experiment. Based on results, the critical stage in designing the networks is to set the number of neurons in the hidden layer that affects an accuracy of the outputs.