| Popis: |
This paper presents the realization of a new moving horizon estimator approach for nonlinear networked control systems and its experimental validation. Each sensor is equipped with an unsynchronized clock and transmits its time-stamped and non-uniformly sampled measurements via packets to the estimator site over a digital communication network. These packets are subject to random delay or might even be completely lost. The estimation problem is formulated as a suitable nonlinear program which can be efficiently implemented in real-time. The key features of the proposed method include the independence from the underlying communication protocols as well as from clock synchronization. Furthermore, the estimator concept allows the controller to be designed for the nominal case, i.e. without considering the network. The measurements are conducted on a networked pendulum test-rig. The swing-up and stabilization of the real pendulum represents a challenging benchmark for the estimator due to the unstable and non-minimum phase system dynamics along with the network-induced non-negligible time delays and packet drops. A two-degree-of-freedom control scheme is used throughout the experiments. Due to the good performance of the proposed estimator, a simple linear LQR controller is sufficient for the feedback part. |
