Algorithm 2: The target recovery mechanism.

Step 1: The MN detects and tracks the target:

1:Reduce its sampling time interval to $\mathsf{\Delta }/N_{\mathsf{\Delta }}$, and implement AUKF to estimate the position of target at each timestep. 2:After $N_{\mathsf{\Delta }}$ samples, predict the next position of target. 3:Activate all static nodes whose sensing range covers the predicted position. 4:If there are no appropriate nodes to form a recovery task cluster, then 5: continue to implement the step 1. 6:end if.

Step 2: The recovery cluster detects the target:

7:If the cluster could track the target well according to the Equation (34), then 8: Execute the location process two times to obtain the target position and velocity. 9: Recover the sampling time interval to $\mathsf{\Delta }$ 10:else 11:  the current CH informs the MN that the target is lost and skip to step 1. 12:end if.

Step 3: The downstream cluster tracks the target:

13:Initialize the noise covariance with ${\mathbf{Q}}_0$ and ${\mathbf{R}}_0$, target state with the position and velocity of target. 14:Perform the standard UKF to predict the next target state, and select the downstream cluster nodes. 15:The recovery cluster broadcasts a target recovery message and activate the downstream task cluster to work.