Year: 2011
Type: dissertation
Abstract: Cyber-physical Systems (CPS's) are networks of embedded systems which operate under tight power and timing constraints and with varying levels of precision.The primary goal is to meet the above requirements for an application that is distributed over a network.Adaptive Classification is introduced in the context of a chemical gas cloud localization application running over the CPS.Standard machine learning binary classifiers are employed and optimized for implementation in reconfigurable hardware at varying precisions, power requirements and speeds.An exemplar set of classifiers contains one chemical species classifier and four individual chemical classifiers such as, Alcohol versus the-rest (of the species) and methyl alcohol versus the-rest(of the individual alcohols).Other sets classify additional species and chemicals.The network is populated with these classifiers by assigning one set of chemical classifiers per node and apply techniques from the field of evolutionary game theory to allow the nodes to self-configure over time.iii
Author: Michael Gilberti
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2793284280