Kalman filtering without Bayesians and Gaussians (Ⅱ)

摘要

This is the second of a multi-part post that introduces Kalman filtering in an accessible way to computer systems researchers. In the previous post, we described how two noisy estimates of a scalar quantity such as temperature can be fused using the optimal linear unbiased estimator shown in Figure 1.The only assumptions are that (ⅰ) we know the variance σ_i of each estimate x_i and (ⅱ) the estimates are uncorrelated.