A critical evaluation of Pure Pursuit, MPC and MPCC: Balancing simplicity, performance and constraints

Stellenbosch University, Electrical and Electronic Engineering Department, Western Cape, South Africa

^* Corresponding author: 22902716@sun.ac.za

Abstract

Autonomous driving technology has advanced significantly in the past decade. In the case of autonomous racing, the crucial challenge lies in guiding the vehicle along the desired path while ensuring safety and efficiency. Several control systems have been developed for fast and accurate path tracking, offering their own unique advantages and limitations. Here we present an investigation into three prominent control strategies in a critical comparative analysis. These control algorithms include Pure Pursuit (PP), Model Predictive Control (MPC) and Model Predictive Contouring Control (MPCC). These control algorithms were chosen due to their difference in complexities starting from a simple PP to a much more complex MPCC. These algorithms are then experimentally validated on a physical F1Tenth vehicle. The simulation results show that PP exhibited the fastest computation time, its performance in the presence of noise and delay was inferior to MPC and MPCC. While MPC demonstrated strong performance in its robustness and resilience to noise and delay compared to PP and MPCC. MPCC, despite producing the fastest lap time, faced challenges in handling noise and delay although not as severe as PP, making MPC the best overall controller for unwanted disturbances. In contrast to the simulated findings, PP demonstrated superior performance in physical implementations compared to MPC and MPCC. The computational demands of optimization-based algorithms result in a computation delay where the algorithm calculates for the vehicle's position one iteration behind its actual movement. This shows that shortening computational delay is critical in a physical system.