Many problems in multiagent networks can be solved through distributed learning (state estimation) of linear dynamical systems. In this paper, we develop a partial-diffusion Kalman filtering (PDKF) algorithm, as a fully distributed solution for state estimation in the multiagent networks with limited communication resources. In the PDKF algorithm, every agent (node) is allowed to share only a subset of its intermediate estimate vectors with its neighbors at each iteration, reducing the amount of internode communications. We analyze the stability of the PDKF algorithm and show that the algorithm is stable and convergent in both mean and mean-square senses. We also derive a closed-form expression for the steady-state mean-square deviation criterion. Furthermore, we show theoretically and by numerical examples that the PDKF algorithm provides a trade-off between the estimation performance and the communication cost that is extremely profitable.

The performance of partial diffusion Kalman filtering (PDKF) algorithm for the networks with noisy links is studied here. A closed-form expression for the steady-state mean square deviation is then derived and theoretically shown that when the links are noisy, the communication-performance tradeoff, reported for the PDKF algorithm, does not hold. Additionally, optimal selection of combination weights is investigated and a combination rule along with an adaptive implementation is motivated. The results confirm the theoretical outcome.

Increasing the energy efficiency of an Internet of Things (IoT) system is a major challenge for its successful implementation. To reduce the computation and storage burden and enhance the efficiency of traditional IoT, an energy-efficient diffusion-based algorithm for state estimation in multiagent networks is proposed in this article. In the proposed algorithm [referred to as the reduced-link diffusion Kalman filter (RL-diffKF)] the nodes (agents) can communicate only with a fraction of their neighbors and each node runs a local Kalman filter to estimate the state of a linear dynamic system. This algorithm results in a significant reduction in communication cost during both adaptation and aggregation processes albeit at the expense of possible degradation in the network performance. To justify the stability and convergence of the RL- diffKF algorithm, an in-depth analysis of the performance is reported. We also consider the problem of optimal selection of combination weights and use the idea of minimum variance estimation to analytically derive the adaptive combiners. The theoretical findings are verified through numerical simulations.

Naturally complex-valued information or those presented in complex domain are effectively processed by an augmented complex least-mean-square (ACLMS) algorithm. In some applications, the ACLMS algorithm may be too computationally- and memory-intensive to implement. In this paper, a new algorithm, termed partial-update ACLMS (PU-ACLMS) algorithm is proposed, where only a fraction of the coefficient set is selected to update at each iteration. Doing so, two types of partial-update schemes are presented referred to as the sequential and stochastic partial-updates, to reduce computational load and power consumption in the corresponding adaptive filter. The computational cost for full-update PU-ACLMS and its partial-update implementations are discussed. Next, the steady-state mean and mean-square performance of PU-ACLMS for non-circular complex signals are analyzed and closed-form expressions of the steady-state excess mean-square error (EMSE) and mean-square deviation (MSD) are given. Then, employing the weighted energy-conservation relation, the EMSE and MSD learning curves are derived. The simulation results are verified and compared with those of theoretical predictions through numerical examples.