Project Initial Plan

The world of autonomous racing presents a challenging control problem that requires accurate kinematic modelling, real-time decision making, and strict adherence to physical constraints. In the Formula Student AI (FS-AI) competition, autonomous race cars must adapt to a selection of circuits that simulate challenging conditions while maintaining high speeds, stability, and precision. Model Predictive Control (MPC) has emerged as a powerful approach in this application. MPC is a control strategy that uses a model of the system to predict its future behaviour and determine control actions that optimise performance over time. Rather than reacting only to a current state, MPC evaluates predicted future states to enable more accurate control decisions. MPC is also widely used to solve problems in the broader field of robotics, as well as in the automotive sector and aerospace.

The project will focus on the design and evaluation of an MPC-based trajectory controller for a Formula Student AI racing car. The controller will be developed using a kinematic vehicle model and integrated into a ROS 2 simulation and control framework, which assumes that localisation and trajectory inputs are already handled/provided. The MPC controller will then compute steering and velocity commands for trajectory control within a Gazebo or Unreal Engine-based Formula Student autonomous simulator, favouring sim-to-real transfer through realistic constraints and interfaces.