Autonomous Driving: Baseline Autonomy

Autonomous driving;Vehicle Estimation;Vehicle Control

Dakibay, Assylbek

University of Waterloo

关键词: Autonomous driving;    Vehicle Estimation;    Vehicle Control;   

Others  :  https://uwspace.uwaterloo.ca/bitstream/10012/12065/1/Dakibay_Assylbek.pdf

瑞士|英语

来源: UWSPACE Waterloo Institutional Repository

【 摘 要 】

In near future Autonomous driving will affect every aspect of transportation and offer asignificant boost in mobility for everyone. Autonomous driving techniques and modulesmust be chosen according to the task the platform is developed for. Slow speed drivingon campus or highway driving in poor weather conditions, may require different sets ofsensors, vehicle models and as a result different software architecture. Some of the mainmodules that an autonomous driving system needs are the vehicle state estimator andvehicle controller. The development of these two modules relies heavily on the robustnessof the vehicle model chosen and the task at hand.University of Waterloo decided to join the Autonomous Driving research by partici-pating in the project, which required development and implementation of the autonomousdriving demo sequence for Consumer Electronics Show in 2017. Since the demo sequencewas to be performed at slow speeds and, because certain vehicle parameters were notavailable at the time, a kinematic vehicle model was used in implementation of the coreautonomous driving modules: state estimation and control. These modules are imple-mented on a full scale autonomous driving platform and were designed based on the needsand requirements of the demo sequence. The implementation results show that the cho-sen vehicle model enables the state estimator to fuse incoming sensor data and allows thecontroller to track the desired path and velocity profile.For further deployment of the autonomous driving platform for research in urban andhighway driving an aggressive driving framework was proposed that is based on dynamicvehicle model and incorporates the tire forces in the generation of the speed profile andkeeps the vehicle at the limits of adhesion. The aggressive driving controller can be utilizedfor emergency evasive maneuvers at low road friction conditions. The controller was testedon a high fidelity simulation software for a double lane change emergency maneuver. Theresults showed that the aggressive driving framework proposed can be successfully incor-porated into the autonomous driving architecture and can perform position and velocitytracking at maximum possible speed.