Since we are discussing the distributive Kalman filter based covariance intersection, we have to take a look at how it develops.
The paper, named "Diffusion Kalman filter based on Covariance Intersection," is of great importance. The real contribution of this paper is not in the algorithm that they proposed, in my opinion. One has to note that the algorithm they proposed still relies on the measurement from far agents to keep the estimation stability. However, this requirement puts severe burden on routing task, which is even more difficult in distributive agents.
I think the main contribution of this paper is that they showed diffusion step is merely a degenerative case of covariance intersection update. Both updates apply convex combination of the incoming information. However, diffusion step does not take care of the covariance.
In conclusion, this paper paves the way for further investigating this type of distributive estimation methods.