Partial Diffusion Kalman Filtering for Distributed State Estimation in Multiagent Networks.

Area of Science:

• Control Systems Engineering
• Networked Systems
• Distributed Computing

Background:

• State estimation in multiagent networks is crucial for distributed learning.
• Limited communication resources pose a significant challenge in these networks.
• Existing methods may not be optimal under communication constraints.

Purpose of the Study:

• To develop a fully distributed state estimation algorithm for multiagent networks with limited communication.
• To introduce the partial-diffusion Kalman filtering (PDKF) algorithm.
• To analyze the performance and stability of the proposed PDKF algorithm.

Main Methods:

• Development of the partial-diffusion Kalman filtering (PDKF) algorithm.
• Agents share only a subset of intermediate estimate vectors.
• Theoretical analysis of algorithm stability and convergence (mean and mean-square).
• Derivation of a closed-form expression for steady-state mean-square deviation.

Main Results:

• The PDKF algorithm is proven to be stable and convergent in both mean and mean-square senses.
• A closed-form expression for the steady-state mean-square deviation is derived.
• Numerical examples demonstrate the algorithm's effectiveness.
• The PDKF algorithm offers a beneficial trade-off between estimation performance and communication cost.

Conclusions:

• The partial-diffusion Kalman filtering (PDKF) algorithm provides an effective solution for distributed state estimation in communication-constrained multiagent networks.
• The algorithm achieves stability and convergence while significantly reducing communication overhead.
• PDKF presents a valuable approach for optimizing resource utilization in networked systems.