"Stability of Digitally Interconnected Linear Systems"

Date and Time:

January 27, 2012 - 2:50pm - 3:10pm

Presentation Abstract:

A sufficient condition for stability of linear subsystems interconnected by digitized signals is presented.  There is a digitizer for each linear subsystem that periodically samples an input signal and produces an output that is quantized and saturated.  The output of the digitizer is then fed as an input (in the usual sense) to the linear subsystem.  Due to digitization, each subsystem behaves as a switched affine system, where state-dependent switches are induced by the digitizer.  For each quantization region, a storage function is computed for each subsystem by solving appropriate linear matrix inequalities (LMIs), and the sum of these storage functions is a Lyapunov function for the interconnected system.  Finally, using a condition on the sampling period, we specify a subset of the unsaturated state space from which all executions of the interconnected system reach a neighborhood of the quantization region containing the origin.  The sampling period proves to be pivotal---if it is too small, then a dwell-time argument cannot be used to establish convergence, while if it is too large, an unstable subsystem may not receive timely-enough inputs to avoid diverging. The method is computational and uses an overapproximation of the reachable set of states of a hybrid system (computed for instance with PHAVer or SpaceEx) as a constraint in the LMIs.