An analysis of a riding control algorithm for two wheeled vehicles with a neural network modeling

This research deals with an analysis of the riding control algorithm for the two wheeled vehicles as the first step towards the modeling of the riders. The rider tasks are divided into two parts. One is for recognition of the environment around the vehicle, and the other is for the decision of the output of the rider. For the recognition part, the relative recognition of the environment based on the rider's position with a radar sight description is used. For the decision about the steering torque, the neural network modeling which describes a dynamic system with feedback and feed forward is used. It is shown that the results of the model agree with the experimental results. Next, the algorithm with sensitivity analysis of the rider model is analyzed. In the first step of this analysis, the method of analyzing the algorithm is checked against the Kondo's driver model of which the control algorithm is well known. The neural network model is constructed with the results of the Kondo's model, and the algorithm analyzed with the model agrees with that of the Kondo's model. Next, the output sensitivities to each input for the rider models are studied. From the results, it is found that the rider always changes his/her main information to decide the control vector depending on their position in the course.