Article Preview
TopIntroduction
With the rapid development of computing, communication and sensing technology, recent years have seen an ever-increasing research interests in applying wireless sensor networks (WSNs) to monitor the operation of large-scale multiagent systems, such as electric power systems, water management systems, gas-pipeline systems, weather forecast systems, and to name but a few (Kar & Moura, 2009; Khan & Moura, 2008; Li & Guo, 2015; “Distributed kalman filtering for sensor networks”, 2007; Scutari & Barbarossa, 2008; Wang, Ahn, Lu, & Staskevich).
For such types of large-scale multiagent systems, it is normal that individual sensors may only cover a limited number of agents due to their physical limitations. Thus, to detect the overall behavior of multiagent systems, sensor fusion problem has to be addressed so that the measurements by individual sensor components can be effectively processed and utilized. The standard solution is to employ a centralized fusion center for information collection and processing. However, such a solution lacks scalability, fault tolerance, and robustness to limited communication. In this paper, we propose a new distributed least-squares estimation algorithm for sensor fusion in WSNs with limited communication.
The proposed solution in this paper follows the line of research on cooperative control and distributed estimation problems in multiagent systems (Francis & Maggiore, 2016; Qu, 2009; Qu, Wang & Hull, 2008; Saber, Fax & Murray, 2007; Wang, 2016; Wang et al., 2013). Plenty of results are available in this regard (Lewis et al., 2014). For instance, the consensus control was solved for first-order linear systems in (Lin, Brouchke & Francis, 2004; Saber & Murray, 2004), for second-order linear systems in Tanner, Jadbabaie and Pappas (2007), for high-order linear systems (Qu, Wang & Hull, 2008; Wang et al., 2006), and for nonlinear systems (Dong et al., 2016; Li, Rang & Xiao, 2016; Lin, Francis & Maggiore, 2007; Moreau, 2005; Wang, 2016). There are also many results solving the distributed estimation problems. For example, average consensus algorithm was designed in Spanos, Olfati-Saber and Murray (2005) to find the average of stationary signals. Distributed Kalman filtering algorithms were proposed in (Kar & Moura, 2009; Khan & Moura, 2008; Olfati-Saber, 2005; Scurati & Barbarossa, 2008). Two dynamic average consensus algorithms, a proportional algorithm and a proportional-integral algorithm, were proposed to solve the distributed estimation problem in Freeman, Yang, and Lynch (2006). Distributed Kalman filtering algorithm was used in Lynch et al., 2008) for environmental modeling. The nonlinear protocol for consensus estimation was proposed in Nosrati, Shafiee, and Menhaj (2012), and it was proved that the tracking errors were upper bounded.