Our project which is part of the PINNs research field, was about finding NN architectures that can predict physical phenomenon – in our case a turn of a driving car.
During our project we first create some datasets of a turning car, and second we trained MLP architectures to predict it.
Our MLP architectures divides into two sections: first, a MLP architecture which should predict the car maneuver without any prior knowledge of physical equations(PDEs). Second, a MLP architecture which predicts the maneuver in addition to some simplified physical equations.
Our project will examine the trade-off between large networks and small ones and the benefit of combining them with some simplified physical equations. The factors that would be examined are accuracy and running time.
We will show in this project that a simple NN combined with ackerman model have much better accuracy then ackerman model alone, and that the running time of this combined model is more or less similar to the running time of ackerman model alone.