AbstractControl systems where sensors and controllers communicate over most standard wireless networks (e.g., Wi-Fi, ZigBee, Bluetooth, 4G) are subjected to random delays. While the probability distribution of these delays is time-varying as it depends on factors such as the number of nodes or the distance between nodes, this is rarely taken into account in the networked control literature. In this paper, we tackle the problem of adaptively identifying a probabilistic delay model
for the network. We consider both independent and identically distributed delays and correlated delays following a Markov chain model. For both delay models, we propose a method relying on a parameterization in terms of a (multivariate)
Gaussian Mixture Model (GMM). Given that the process itself might be unknown or slowly varying we combine this with a method to adaptively identify a fully-parameterized state-space model of a continuous-time, linear time-invariant (LTI) dynamic system using input-output data and observed network-induced
delays. We validate our method using a simulated LTI system communicating over a Wi-Fi, ZigBee and Bluetooth networks.
|Date of Award||10 Jan 2020|
|Supervisor||Duarte J. Guerreiro Tomé Antunes (Supervisor 1)|