Recent advances in the field of dynamical systems have led to the development of discrete sequential systems with applications in numerous areas including automation processes, digital computer mechanization, and series systems. Currently, designed centralized state estimation methods are used to monitor the operation of discrete sequential systems. This, however, depends on the measured state of the available individual local subsystems. Unfortunately, this approach involves high discrete sequential systems dimensions and collecting measurements form each and every local subsystem which is differently located so as to estimate their states.
This leads to a complex system that is generally time-consuming and expensive in terms of costs. As such, researchers have been looking for alternatives to address this challenge and have identified distributed state estimation approach as a promising solution. On the other hand, state estimation of the discrete sequential systems has not been fully explored.
To this note, Dr. Bo Chen, Dr. Guoqiang Hu at Nanyang Technological University in collaboration with Professor Daniel Ho at the City University of Hong Kong and Professor Li Yu from the Zhejiang University of Technology investigated distributed state estimation for time-varying discrete sequential systems. Fundamentally, they purposed to effectively enabled decoupling of large-scale interconnected systems so as to enhance their efficiency and applications. This included improving state estimation accuracy for the individual subsystems. Their research work is currently published in the research journal, Automatica.
In brief, the research team initiated their research work by cross-examining time-varying discrete sequential systems taking into consideration the Gaussian white noises. Next, they designed a local optimal distributed Kalman filter in minimum linear variance sense. In addition, with a bounded mean square error of the distributed estimator, a stable condition was effectively derived. This was performed while taking into consideration the distributed Kalman filter structure and the noise statistical properties. Furthermore, the possibility of using only one sensor to monitor an individual subsystem in the discrete sequential systems under distributed structure was also examined. So as to validate the effectiveness of the proposed method, the authors developed an illustrative method. This was realized by bounding the distributed Kalman filter mean square error with time-varying gains to infinity.
On the other hand, it was necessary to employ a target tracking system so as to demonstrate the feasibility of the system. From the experimental results, the authors observed that the estimation performance of the accuracy could be enhanced by developing certain fusion criteria. This however required design of multiple sensors fusion method for monitoring the operations of all the subsystems forming a discrete sequential system.
In summary, the research team successfully investigated state estimation problems in time-varying discrete sequential systems. From the demonstrated example, the authors noted that rather than using one sensor to monitor an individual subsystem, a multisensory fusion can significantly improve the estimation performance of the system. In general, the study provides vital information that will further advance future distributed Kalman filter based research work and especially in cases where multiple sensors are used to monitor individual subsystems.
Chen, B., Hu, G., Ho, D., & Yu, L. (2019). Distributed Kalman filtering for time-varying discrete sequential systems. Automatica, 99, 228-236.Go To Automatica