Distributed estimation techniques for cyber-physical systems

Resumen

Nowadays, with the increasing use of wireless networks, embedded devices and agents with processing and sensing capabilities, the development of distributed estimation techniques has become vital to monitor important variables of the system that are not directly available. Numerous distributed estimation techniques have been proposed in the literature according to the model of the system, noises and disturbances. One of the main objectives of this thesis is to search all those works that deal with distributed estimation techniques applied to cyber-physical systems, system of systems and heterogeneous systems, through using systematic review methodology. Even though systematic reviews are not the common way to survey a topic in the control community, they provide a rigorous, robust and objective formula that should not be ignored. The presented systematic review incorporates and adapts the guidelines recommended in other disciplines to the field of automation and control and presents a brief description of the different phases that constitute a systematic review. Undertaking the systematic review many gaps were discovered: it deserves to be remarked that some estimators are not applied to cyber-physical systems, such as sliding mode observers or set-membership observers. Subsequently, one of these particular techniques was chosen, set-membership estimator, to develop new applications for cyber-physical systems. This introduces the other objectives of the thesis, i.e. to present two novel formulations of distributed set-membership estimators. Both estimators use a multi-hop decomposition, so the dynamics of the system is rewritten to present a cascaded implementation of the distributed set-membership observer, decoupling the influence of the non-observable modes to the observable ones. So each agent must find a different set for each sub-space, instead of a unique set for all the states. Two different approaches have been used to address the same problem, that is, to design a guaranteed distributed estimation method for linear full-coupled systems affected by bounded disturbances, to be implemented in a set of distributed agents that need to communicate and collaborate to achieve this goal. Under these conditions, the first technique uses sets that are mathematically described by zonotopes and it intends to achieve the minimization of the estimation uncertainty computing adequate local and neighbour gains. The observer can be designed in independent distributed steps by means of a simple algebraic equation. An important benefit of the proposed structure is the reduction of the computational requirements concerning existing solutions. Under the same conditions, the second technique uses sets that are mathematically described by constrained zonotopes. Periodic, multi-rate, event-based or fully asynchronous communication schemes are shown to be easily integrated with the proposed estimation structure. Particular attention was paid to the impact of transmission, depending on the different communication schemes. The nature of all the contributions of this thesis is theoretical. However, the solutions adopted could be applied to a wide variety of distributed systems. Nevertheless, simulations and numerical results are considered to compare the proposed solutions with existing ones in the field.