Eugenio Alcala

Automobile racing presents a formidable challenge, demanding drivers to operate their vehicles at the limits of friction while maintaining control. Achieving optimal performance requires extensive coaching for car owners and demonstrative methodologies in race-driving coaching exhibit considerable potential to enhance the learning process. However, since the learner should be in the driver’s seat, this methodologies can greatly benefit from autonomous and consistent vehicle operation capabilities at performance levels comparable to or exceeding those achieved by professional human drivers. This study presents a real-world comparison of an in-house developed autonomous racing system against five highly skilled race-car drivers in a five-lap time trial race under identical conditions. Temporal, trajectory and dynamic performance metrics describing the driving performance of each participant were obtained and discussed to gain insights into the human-machine performance gap. Our findings demonstrate that the autonomous driving system can perform at a very high level, outperforming highly skilled racing drivers. Notably, the autonomous system excelled in areas such as consistency and smoothness, dynamically generating precise trajectories with lower dispersion than the human participants. The system also maintained optimal slip angles and required fewer significant steering corrections, indicating smother driving. Our analysis showcases where the autonomous system falls short of the best human drivers, providing clear directions for further development to close the performance gap on the racetrack. These results underscore the potential of autonomous racing systems as real-life racing demonstrators and therefore their utility for enhancing race-driving coaching.

This article presents an approach to address the problem of localisation within the autonomous driving framework. In particular, this work takes advantage of the properties of polytopic Linear Parameter Varying (LPV) systems and set-based methodologies applied to Kalman filters to precisely locate both a set of landmarks and the vehicle itself. Using these techniques, we present an alternative approach to localisation algorithms that relies on the use of zonotopes to provide a guaranteed estimation of the states of the vehicle and its surroundings, which does not depend on any assumption of the noise nature other than its limits. LPV theory is used to model the dynamics of the vehicle and implement both an LPV-model predictive controller and a Zonotopic Kalman filter that allow localisation and navigation of the robot. The control and estimation scheme is validated in simulation using the Robotic Operating System (ROS) framework, where its effectiveness is demonstrated.

This article presents an approach to address the problem of designing advanced control and estimation techniques for the autonomous driving guidance. In particular, this work takes advantage of the properties of polytopic LPV systems and predictive optimal control to guide the vehicle along a planned trajectory. Linear Parameter Varying (LPV) theory is used to model the dynamics of the vehicle and implement an LPV-Model Predictive Controller (LPV-MPC) that can be computed online with reduced computational cost. Furthermore, the LPV framework is used to design an optimal observer that estimates vehicle variables that cannot be measured. The control and estimation scheme is validated in simulation using the Robotic Operating System (ROS) framework where its effectiveness is demonstrated.

In this paper, the Model Predictive Control (MPC) and Moving Horizon Estimator (MHE) strategies using a data-driven approach to learn a Takagi-Sugeno (TS) representation of the vehicle dynamics are proposed to solve autonomous driving control problems in real-time. To address the TS modeling, we use the Adaptive Neuro-Fuzzy Inference System (ANFIS) approach to obtain a set of polytopic-based linear representations as well as a set of membership functions relating in a non-linear way the different linear subsystems. The proposed control approach is provided by racing-based references of an external planner and estimations from the MHE offering a high driving performance in racing mode. The control-estimation scheme is tested in a simulated racing environment to show the potential of the presented approaches.

In this work, a novel approach is presented to solve the trajectory tracking problem for autonomous vehicles. This approach is based on the use of a cascade control where the external loop solves the position control using a novel Takagi Sugeno - Model Predictive Control (TS-MPC) approach and the internal loop is in charge of the dynamic control of the vehicle using a Takagi Sugeno - Linear Quadratic Regulator technique designed via Linear Matrix Inequalities (TS-LMI-LQR). Both techniques use a TS representation of the kinematic and dynamic models of the vehicle. In addition, a novel Takagi Sugeno estimator - Moving Horizon Estimator - Unknown Input Observer (TS-MHE-UIO) is presented. This method estimates the dynamic states of the vehicle optimally as well as the force of friction acting on the vehicle that is used to reduce the control efforts. The innovative contribution of the TS-MPC and TS-MHE-UIO techniques is that using the TS model formulation of the vehicle allows us to solve the nonlinear problem as if it were linear, reducing computation times by 10-20 times. To demonstrate the potential of the TS-MPC, we propose a comparison between three methods of solving the kinematic control problem: using the non-linear MPC formulation (NL-MPC) with compensated friction force, using the TS-MPC approach with compensated friction force and using TS-MPC without compensated friction force.

This study presents a solution for the integrated longitudinal and lateral control problem of urban autonomous vehicles. It is based on a gain‐scheduling linear parameter‐varying (LPV) control approach combined with the use of an Unknown Input Observer (UIO) for estimating the vehicle states and friction force. Two gain‐scheduling LPV controllers are used in cascade configuration that use the kinematic and dynamic vehicle models and the friction and observed states provided by the Unknown Input Observer (UIO). The LPV–UIO is designed in an optimal manner by solving a set of linear matrix inequalities (LMIs). On the other hand, the design of the kinematic and dynamic controllers lead to solve separately two LPV–Linear Quadratic Regulator problems formulated also in LMI form. The UIO allows to improve the control response in disturbance affected scenarios by estimating and compensating the friction force. The proposed scheme has been integrated with a trajectory generation module and tested in a simulated scenario. A comparative study is also presented considering the cases that the friction force estimation is used or not to show its usefulness.

This work proposes a solution for the longitudinal and lateral control problem of urban autonomous vehicles using a gain scheduling LPV control approach. Using the kinematic and dynamic vehicle models, a linear parameter varying (LPV) representation is adopted and a cascade control methodology is proposed for controlling both vehicle behaviours. In particular, for the control design, the use of both models separately lead to solve two LPV LMI-LQR problems. Furthermore, to achieve the desired levels of performance, an approach based on cascade design of the the kinematic and dynamic controllers has been proposed. This cascade control scheme is based on the idea that the dynamic closed loop behaviour is designed to be faster than the kinematic closed loop one. The obtained gain scheduling LPV control approach, jointly with a trajectory generation module, has presented suitable results in a simulated city driving scenario.