(a) Initialization | Conditioned on each hypothesis \(H^{j}\), initialize the state \({\hat{X}}_{0}^{(j)}\) and its MSE \(P_{0}^{(j)}\) |
Initialize the hypothesis probability \(P\{H^{j} \}=1/N\). | |
(b) Update | At time k, with data \(Z_{k}\) available: |
Update the state estimate \({\hat{X}}_{k}^{(j)}\) and its MSE \(P_{k}^{(j)}\) under each hypothesis \(H^{j}\); | |
Update the posterior probability of each \(P\{H^{j}|Z^{k}\}\) according to (6) | |
(c) Further | Compute the expected estimation cost \(\varepsilon _{ij}^{k}\) according to (3); |
computation | For each candidate \(D^{i}\), compute the estimate \({\check{X}}_{k}^{(i)}\) according to (5) |
(d) Output | The optimal CCJDE-MDT detection result is \(D_{k}^{i}\): \(C^{i}(Z^{k})\le C^{l}(Z^{k}),\forall l\); |
The corresponding CCJDE-MDT tracking result is \({\check{X}}_{k}^{(i)}\) |