Algorithm 1. Process of Data Accumulation in Sensor Nodes

Input: Distance measurements from the sensor to the target; Target information from other nodes

Output: Target position and velocity estimation $X\_est$; State estimation covariance $P\_est$; Current motion model probability $u\_current$

1: Measure the distance from the target to the sensor node

2: if other nodes sent likelihood function and related information then

3:          Receive information from other nodes

4:          if length of accumulated information ≥ time window length then

5:                $X\_est, P\_est, u\_current$← LSTM-based motion model classification for target tracking

6:          else

7:                $X\_est$,$P\_est$,$u\_current$← Traditional IMM target tracking

8:          end if

9: end if

10: Send target information to other nodes in the target’s moving direction

11: return $X\_est$, $P\_est$, $u\_current$