Neural-based model predictive control for tackling steering delays of autonomous cars

Abstract

We propose a Neural Based Model Predictive Control (N-MPC) approach to tackle delays in the steering plant of autonomous cars. We examined the N-MPC approach as an alternative for the implementation of the Intelligent and Autonomous Robotic Automobile (IARA) steering control subsystem. For that, we compared the standard solution, based on the Proportional Integral Derivative (PID) control approach, with a N-MPC approach. For speeds of up to 25 km/h, the IARA's steering plant delay is not a problem for the PID control approach. However, in higher speeds, it causes large steering oscillations, which prevent proper operation. For this, we modeled the IARA's steering plant using a neural network and employed the neural model in the N-MPC. Our experimental results showed N-MPC can drastically reduce the impact of IARA's steering plant delays, which allowed its autonomous operation at speeds of up to 37 km/h.