Control design for comfort optimization in assisted and autonomous driving

In the last years, the development of Advanced Driver Assistance Systems (ADAS) and self-driving vehicles is one of the most important and popular automotive topic. Vehicle automation is changing the experience of driving. It is predicted to have a beneficial impact on many aspects of driving, such as traffic flow, comfort, allocation of time resources, and safety. Since safety has special relevance, many studies in the field of automated driving to date have focused on it. Little focus has so far been placed on the aspect of comfort in autonomous vehicles. However, this aspect seems crucial to investigate because uncomfortable implementations will have a negative influence on the driver and passengers feeling. This thesis is focused on the control design for comfort optimization in assisted and autonomous vehicles. The first part of this work is devoted to understanding which of the existing comfort evaluation methods can be used for both assisted and autonomous ground vehicles. To consider peak and average motion, the Weighted Root Mean Square Acceleration (WRMSA) is proposed to assess the driving comfort. In particular, the adopted strategy refers to the procedure contained in the standard ISO 2631. The second part is related to the implementation of three different control strategies. Firstly, a combined lateral and longitudinal controller for autonomous driving based on Model Predictive Control (MPC) is analyzed. The proposed strategies allow lateral guidance and speed regulation by acting on the front wheel steering angle and on acceleration/deceleration to minimize the vehicle’s lateral deviation and relative yaw angle with respect to the reference trajectory. The significant parameters, as the weighting parameters, are varied by trial and error to get the best comfort evaluation. Secondly, an adaptive MPC control for the lateral control is linked to a PID control for the longitudinal control of the vehicle. In this case, once MPC parameters are fixed, the PID parameters are optimized to maximize passenger comfort. Finally, lateral control is performed through a non-linear control law, that allows to track the trajectory, while longitudinal control is, as in the second case, implemented using a PID. The proposed controllers and comfort evaluation method are tested in a proper scenario to check the performances and to prove the validity of the comfort evaluation method. In the very first stage, the dynamics of the vehicle is modelled using a three degree of freedom rigid vehicle model. Then, a complete vehicle model implemented in CarSim is used to collect more outputs and to improve the estimation of comfort. The overall system has been developed using MATLAB, Simulink, Model Predictive Control Toolbox, Automated Driving System Toolbox and CarSim. When the different control strategies (MPC, MPC and PID, PIDs) are tested, the comfort is evaluated using the ISO 2631. The standard provides indications of likely reactions to various magnitudes of overall vibration total values. It states that until 0.3 m/s2 the ride is not uncomfortable while the ride becomes slightly uncomfortable in the range between 0.3 m/s2 and 0.615 m/s2. The results obtained using the proposed control strategies show that the index ranges between 0.15 m/s2 and 0.6 m/s2.