A Neural Network Approximation of Nonlinear Car Model Using Adams Simulation Results

A neural network model of a full car has been developed here on the basis of ADAMS simulation results. The model basically intended for roll control studies, is a completely non-liner model and has 104 degrees of freedom. ADAMS software has been used to determine the model behavior to specific steering inputs. The out put of the simulation program was then used to train a neural network constructed to approximate the model for controller design and real time studies of control action. Specific time delayed feedback inputs to the neural network resulted an efficient approximate model with good accuracy for control tasks.