State Estimation for Formula Student Driverless Application

The ”Formula Student Driverless” is a specific category within the Formula Student com- petition in which students design, build, and race single-seat vehicles without drivers. This category focuses on the development of autonomous or driverless vehicles, using advanced technologies such as autonomous driving systems, sensors, artificial intelligence, and other emerging technologies to enable the vehicles to operate without human intervention. In this thesis work carried out in collaboration with the ”Squadra Corse Driverless Polito” team, a method for estimating the longitudinal and lateral velocity of the vehicle in the absence of sensors such as a Ground Speed Sensor is introduced. Knowledge of these state variables is crucial for managing the controls of an autonomous vehicle. In fact, in addition to enabling basic controls such as Traction Control and Torque Vectoring, knowledge of these variables is essential for improving the mapping of the vehicle’s position on the track, which cannot rely solely on GPS sensors. This work builds upon some previous work by the team, such as the estimation of longi- tudinal velocity, and then implements new techniques for estimating lateral velocity. After an analysis of the most commonly used techniques in literature and traditional vehicles, the work done by a Formula Sae team for the estimation of these variables is described. In particular, two different types of Kalman Filters were designed (one kinematic and one dynamic). This paper describes their implementations and operations, and also presents an innovative method for combining the results of the two filters. The entire work was first validated through numerous simulations using Vi-Grade soft- ware and was subsequently validated on the track with experimental data. Finally, an analysis of the influence of the accuracy of state estimation on the vehicle’s odometry was conducted by comparing results obtained on the track using a Ground Speed Sensor with results obtained through estimation.