DoTVeX 1.0: Dynamic Observer Toolbox Using Convex Optimization Formulations

DoTVeX is an open source CVX-based Dynamic observer Toolbox on MATLAB which can be used to construct state estimators for wide classes of dynamical systems (e.g., networked systems, delayed systems, cyber-physical systems) by leveraging convex relaxations and solving disciplined

conVeX programs. That is, given a state space representation of a dynamical system, the toolbox can be employed to design the observer which generates estimates of internal system states, exogenous inputs and unknown sensor attacks/noise. There are two categories of dynamic

state estimation: stochastic estimation and deterministic observers. There are numerous toolbox implementations for Kalman filter estimation and it’s derivatives. However, no toolbox exists for deterministic observers. This gap is addressed in this thesis.

A major objective of the toolbox—the first of its kind—is to assist researchers and educators who are interested in dynamic state-estimation problems. The toolbox is used to (i) design cutting edge state and unknown input observers; (ii) simulate the estimation results post-implementation; and (iii) compute performance guarantees for the observer scheme or maintain pre-defined/user specified performance levels. The present thesis serves as a brief introduction to DoTVeX. The motivation behind the current toolbox is provided, and general classes of dynamical systems (for example: networked control systems, time-delayed systems, nonlinear systems) which can be handled by the toolbox are listed. Convex relaxations and disciplined convex programming for observer design using CVX is discussed, with accompanying implementation results. A comprehensive

overview of DoTVeX’s software architecture is presented.