Distributed sensor fusion estimation algorithms based on Kalman Filtering

Abstract

Efficient distributed sensor fusion is critical for reliable state estimation in applications such as autonomous vehicles, robotics, and environmental monitoring. This review examines four main distributed Kalman filtering approaches: matrix-weighted fusion, covariance intersection, feedback based optimal fusion, and machine learning–augmented schemes. Core equations for each method are outlined. Communication requirements, computational complexity, and estimation accuracy are systematically compared across diverse network conditions, including synchronous, asynchronous, and lossy environments with packet loss. Practical challenges addressed encompass scalability in large-scale, high-dimensional systems, numerical stability under limited computational precision, and inherent trade- offs between estimation performance and resource consumption. Case studies and extensive simulations demonstrate the real-world efficacy of each method. Finally, key future research directions are highlighted, focusing on edge- optimized architectures, robust algorithms tolerant to significant delays and asynchronous updates, and the integration of essential security and privacy features. This synthesis provides a roadmap for advancing distributed Kalman filters within resource-constrained sensor networks.