Controladores multivariables para un vehículo autónomo terrestre: Comparación basada en la fiabilidad del software

In this paper, three multivariable speed controllers (linear quadratic regulator-LQR, proportional integral derivative - PID, and Fuzzy) were compared with each other to find which one has the best software reliability. The reliability tests were conducted on perturbed controllers with injected faults, simulating typical programmer errors. These controllers were designed to operate in an autonomous ground vehicle, and they were tuned by using a genetic algorithm. Given the large number of tests to be performed it was decided to build a multi-computer simulator in which they were carried out more than 90000 essays. In each of the trials, the perturbed controllers were subjected to a tour of approximately 20 minutes on a slightly wavy ground. With the obtained data, the reliability curves were elaborated by means of the Kaplan-Meier procedure, and this allowed their comparison which was the aim of this research. Under the observed experimental conditions, the LQR controller provides the best behavior, the second position belongs to the PID controller, and the third one to the fuzzy controller.