https://www.cs.cmu.edu/~motionplanning/lecture/Chap8-Kalman-Mapping_howie.pdf
d D. Fox Finding the correction (no noise!) y = Hx Ω = {x | Hx = y} HH Kν =ν T substituting yields x H Kν T Δ = Ω = {x | Hx = y} Δx we require x ˆ(k +1| k) • ⇒ HΔx = y − Hxˆ(k +1| k) = H(x − xˆ(k +1| k)) =ν H(xˆ(k +1| k) + Δx) = y is the "best" estimate of . G