Development of Model Predictive Motion Planning and Control for Autonomous Vehicles

The aim of this paper is to present a model predictive control strategy for autonomous vehicles with capabilities to handle trajectory planning (trajectory optimization) and trajectory tracking. In addition, with the content presented in 2019, current development of the trajectory planner has been updated and now is obtained in an online fashion using a linearized vehicle model. For the trajectory planning, a set of boundaries have been given in terms of road limits, reference path with lateral limits and speed limitations for each section. In relation to vehicle dynamics performance boundaries have been given such as maximum lateral acceleration and sideslip angle. In addition, the trajectory planner integrates an ACC logic in the MPC formulation to maintain an adequate distance to the vehicle in front. With this set up, a model predictive algorithm calculates a dynamically feasible trajectory which is the input to the second MPC algorithm working in series. This second MPC is the trajectory tracker that aims to follow closely the reference given. By using a dynamic model of the vehicle, predictions of the future states of the vehicle helps to anticipate the actions to be performed within a defined horizon. The work presented has been developed under the scope of a European research project in the field of autonomous driving acceptance: SUaaVE project which will be introduced in the 1st chapter. In the 2nd chapter a detailed description of the controllers is given. In 3rd chapter a description of the tests is disposed followed by the results of the motion comfort survey performed have been analyzed using advanced machine learning techniques with the aim of being able to predict passenger feelings and be able to influence on vehicle behavior accordingly.