Estimation and control with limited information and unreliable feedback

Quantized systems;Stability of hybrid systems;Input-to-state stability (ISS);Output feedback and Observers;Disturbances;Control with communication constraints;Time delays;Robust Estimation;Compressed Sensing;Vision Based Control

Sharon, Yoav

关键词: Quantized systems;    Stability of hybrid systems;    Input-to-state stability (ISS);    Output feedback and Observers;    Disturbances;    Control with communication constraints;    Time delays;    Robust Estimation;    Compressed Sensing;    Vision Based Control;   

Others  :  https://www.ideals.illinois.edu/bitstream/handle/2142/18507/Sharon_Yoav.pdf?sequence=1&isAllowed=y

美国|英语

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

【 摘 要 】

Advancement in sensing technology is introducing new sensors that can provide information that was notavailable before. This creates many opportunities for the development of new control systems. However,the measurements provided by these sensors may not follow the classical assumptions from the controlliterature. As a result, standard control tools fail to maximize the performance in control systems utilizingthese new sensors. In this work we formulate new assumptions on the measurements applicable to newsensing capabilities, and develop and analyze control tools that perform better than the standard toolsunder these assumptions. Specifically, we make the assumption that the measurements are quantized. Thisassumption is applicable, for example, to low resolution sensors, remote sensing using limited bandwidthcommunication links, and vision-based control. We also make the assumption that some of the measurementsmay be faulty. This assumption is applicable to advanced sensors such as GPS and video surveillance, aswell as to remote sensing using unreliable communication links.The first tool that we develop is a dynamic quantization scheme that makes a control system stableto any bounded disturbance using the minimum number of quantization regions. Both full state feedbackand output feedback are considered, as well as nonlinear systems. We further show that our approachremains stable under modeling errors and delays. The main analysis tool we use for proving these resultsis the nonlinear input-to-state stability property. The second tool that we analyze is the Minimum Sumof Distances estimator that is robust to faulty measurements. We prove that this robustness is maintainedwhen the measurements are also corrupted by noise, and that the estimate is stable with respect to suchnoise. We also develop an algorithm to compute the maximum number of faulty measurements that thisestimator is robust to. The last tool we consider is motivated by vision-based control systems. We use anonlinear optimization that is taking place over both the model parameters and the state of the plant inorder to estimate these quantities. Using the example of an automatic landing controller, we demonstratethe improvement in performance attainable with such a tool.