Experimental Evaluation of a Distributed Kalman Filter Algorithm

This paper evaluates the performance of a distributed Kalman filter applied to an ultrasound based positioning application with seven sensor nodes. By distributed we mean that all nodes in the network desires an estimate of the full state of the observed system and there is no centralized computation center after deployment. Communication only takes place between neighbors and only once each sampling interval. The problem is solved by communicating estimates between neighbors and then forming a weighted average as the new estimate. The weights are optimized to yield a small estimation error covariance in stationarity. The minimization can be done off line thus allowing only estimates to be communicated. In the experimental setup the distributed solution performs almost as good as a centralized solution. The proposed algorithm also proved very robust against packet loss.