Design of neural nonlinear PFC Controller to control speed of Autonomous Car

TL;DR

This paper presents a neural nonlinear predictive functional controller for autonomous cars that reduces fuel consumption by modeling engine behavior with neural networks and comparing its performance to a traditional PI controller.

Contribution

It introduces a neural network-based nonlinear PFC for autonomous cars, enhancing robustness and efficiency over existing PI controllers.

Findings

Reduced fuel consumption demonstrated in simulations

Controller robustness with constraint handling

Improved control performance compared to PI controller

Abstract

In this research, we are going to design a neural nonlinear predictive functional controller (PFC) to achieve a reduced fuel consumption for a chosen autonomous car walks according to a supplied speed trajectory on known roads. We used a fitting neural network as a simple tool for modelling the car's engine and control laws needed to calculate the suitable control commands passed to the brakes and gas pedals' actuators. Independent model method and constraints handling are used to provide controller robustness. We used MATLAB Simulink and IPG CarMaker to design and test our PFC controller. The performance of designed PFC controller is compared to the performance of a PI controller which exists within IPG CarMaker simulator. Keywords :- Predictive Functional Controller, Fuel Consumption, Neural Network, Independent Model, Constraint Handling, PI Controller.