Guidance, control and estimation of autonomous vehicle systems

autonomous system;unmanned system;distributed control;Adaptive Control;cooperative control;vision based guidance

Li, Zhiyuan

关键词: autonomous system;    unmanned system;    distributed control;    Adaptive Control;    cooperative control;    vision based guidance;   

Others  :  https://www.ideals.illinois.edu/bitstream/handle/2142/46871/Zhiyuan_Li.pdf?sequence=1&isAllowed=y

美国|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

This research focuses on designing practical guidance, control and estimation algorithms for autonomous vehicle systems. The objective is to provide robust controland estimation algorithms for both single and multiple autonomous vehicles under realistic motion, sensing, and communication conditions such as uncertaindynamics, passive sensor, limited communication range, and the complicated coupling among them.With that objective in mind, we start with designing vision-based guidance and estimation algorithms for small unmanned air vehicles (UAVs) to track a groundtarget. The tracking task is for the UAV to maintain a horizontal orbit around the target with a predefined radius, so as to provide continuous visual surveillanceof the target with a desired resolution. We design simple vision-based guidance laws for the cases of both static target and moving target by controlling only theturn rate of the UAV, and give rigorous proofs of the “almost global” asymptotic stability of the closed-loop systems. We extend the tracking algorithm for a singleUAV to the case of coordinated target tracking with multiple UAVs by controlling only the turn rates, where, in addition to orbiting about the target, each UAV isrequired to maintain given phase differences from others. In order to provide continuous estimates of the target’s motion, including its position, velocity, andheading angle, we formulate an estimation problem in a deterministic setup such that the recently developed fast estimator can be applied which yields guaranteedtransient performance. The second part of this thesis is dedicated to the topic of distributed control of a group of unmanned vehicles, in the presence of realistic dynamical, sensing andcommunication constraints. The objective is to drive a group of unmanned vehicles with uncertain dynamics from different initial conditions to aggregate towardsa moving target of interest and align their velocities with it, resulting in a moving flock. We develop a cascaded control framework to decouple the inter-agentcoordination from local uncertainty compensation for each single agent, such that existing algorithms in literature designed for simple ideal agent kinematics can beused as the outer-loop, while L1 adaptive controllers are used for the inner-loop. Two different ideal agent model are considered, namely, the double integrator andthe nonholonomic model.To better handle the uncertainty compensation under limited computation and sensing capability, which is a quite common case for cheap and small autonomousvehicles, we develop a new L1 adaptive controller in the third part of this thesis. It features a modified piecewise constant adaptive law that imposes significantlyless stringent requirements on the computation and sensing frequencies. The main idea is to more efficiently exploit the information of the uncertainties from thepast samples and use this information to compensate for the uncertainty in the next sample period. We compare the performance and robustness trade-off of thenew and the existing L1 adaptive controllers.