Analysis and Design of a Non-linear MPC Algorithm for Vehicle Trajectory Tracking and Obstacle Avoidance

This paper presents a non-linear Model Predictive Control (MPC) algorithm developed using GRAMPC library (GRadient-based Augmented-lagrangian framework for embedded non-linear MPC). Trajectory tracking and obstacle avoidance capabilities for vehicle equipped Advanced Driver Assistance Systems are becoming more and more important. These functions give better comfort and enhance safety for drivers and passengers. In this work, the vehicle has been modelled using six states (XY coordinates of the mass centre, yaw angle, velocity, reference yaw angle and Y errors) and two controls (front steer angle and acceleration). In real applications, the desired trajectory and the constraints are provided by the navigation system and sensors, while here a sinusoidal testbench has been chosen. GRAMPC library gives the opportunity of controlling many options, of managing real-time features and accuracy performances. Due to complexity of the non-linear MPC algorithm, classic Electronic Control Units (ECUs) based on low-cost microcontroller units (MCU), often do not have sufficient computing capabilities to meet the accuracy specifications. This explain the reasons of using high-end MCUs or, if necessary, HW-accelerated systems (MCU + FPGA), in order to guarantee performances and safety.