Effects of Sensing & Control Errors on Quality of Adaptation in Networked Systems

The adaptation logic of a networked system “S” is anchored on two functional elements: I) Sensing of observable system state and outputs by direct measurements; ii) Inference of hard-to-observe system state and environment parameters using model-based estimates. These observed system data is then used in a model-aided computation of the control actions to steer the output of “S” towards a desired operating level. Observational errors in the measurement and estimation processes (i) and (ii) impact the quality of adaptation attained by “S”. In a rate-adaptive video transport system (BAVT) for e.g., bandwidth estimation errors lead to incorrect allocation of video send rates to the sources, which can affect the rate stability & convergence during congestion recovery. Given that sensing & control errors arise due to imprecise and/or incomplete knowledge about the state and environment of “S”, a model-based treatment of the errors allows assessing their impact on the system behavior. We describe a software cybernetics framework to deal with the sensing & control errors during system operations (with a case-study of BAVT).