Core Concepts
The author proposes quantization schemes for stabilizing infinite-dimensional discrete-time systems subject to packet loss in the sensor-to-controller channel. The closed-loop system achieves exponential convergence under suitable conditions on the quantization error bounds and the duration bound of packet loss.
Abstract
The key highlights and insights of the content are:
The author considers abstract infinite-dimensional linear systems, especially regular linear systems, with finite-dimensional input and output spaces. The continuous-time plant is connected to a zero-order hold and a generalized sampler.
The author proposes quantization schemes for the discretized system with packet loss in the sensor-to-controller channel. The quantizers dynamically adjust the zoom parameters to ensure that the plant input and output are contained in the quantization regions and the quantization errors converge to zero.
The author shows that the closed-loop state starting in a given region exponentially converges to zero if the bounds of quantization errors and packet-loss duration satisfy suitable conditions. This is achieved by a discretization-based approach.
The author develops methods for approximately computing the norms of the operators representing the system dynamics, which are used in the quantizer design. The author focuses on the case where the plant is diagonalizable with finite-dimensional input and output spaces.
For sampled-data regular linear systems, the author shows that the exponential convergence at sampling times can be extended to the continuous-time interval [0, ∞).