|ROSELLI FEDERICO||Cycle: XXX |
Section: Systems and Control
Tutor: BOLZERN PAOLO GIUSEPPE EMILIO Major Research topic
:Vehicle dynamics planning and control for safety and comfort in autonomous cars
Advisor: SAVARESI SERGIO MATTEOAbstract:
Autonomous driving is going to replace many of today’s forms of transportation and also change the life style of most of the people. The rapid rate of technological advance is making this scenario much closer to reality; nevertheless some critical issues, mainly legal and technological, must be faced.
This dissertation deals with the two main topics related to vehicle dynamics control in self-driving cars: safety and comfort. The replacement of human drivers with intelligent algorithms will wipe out many control systems developed to improve the driving experience or help the diver in the most challenging maneuvers, but will amplify the need for a comfortable and safe operation.
To guarantee an acceptable level of safety, the vehicle must be able to handle critical maneuvers such as obstacle avoidance even at high speed. Two lateral controllers to track emergency maneuvers are presented: one combines the look-ahead error with the lateral shift of the CoG, the second one is based on the path curvature. Both control strategies exploit the preview of the reference trajectory, crucial in case of quick maneuvers. Differently from most approaches in the literature, on-line optimization is not required by these controllers, simplifying real-time implementation. The algorithms are validated experimentally on a Dodge Dart equipped with by-wire interfaces and high precision GPS and on a modiﬁed RC car; good tracking performance are obtained up to the friction limits also on evasive maneuvers. A reactive planner for emergency lane change is also developed: the algorithm considers the handling limits due to road/tire friction and guarantees that the generated trajectory can be tracked by the path following controller.
The last part of the dissertation focus on the possible comfort improvement due to the presence of autonomous drive sensors. Stereo cameras can be used to detect the presence of a bump or a pothole in front of the vehicle and identify some important characteristics of the obstacle; this information is used to design a semi-active controller to optimize comfort and plan the longitudinal speed proﬁle that better reduces the excitation.