Algorithm 2. Target Tracking Process for Motion Model Classification Using Long Short-Term Memory

Input: Distance measurement from the sensor to the target ($dist\_mea$); State estimates of each model ($Xn\_est$); Covariance estimates of each model ($Pn\_est$); Motion model likelihood value matrix over multiple time steps ($u\_history$)

Output: Estimated target position and velocity ($X\_es$t); Covariance of target state estimation ($P\_est$); Current target motion model probability ($u\_current$)

1: for i = 1 to 3 do

2:          $u\_current\left(i\right) \leftarrow move\_model\_likelihood\_func (dist\_mea, Xn\_est(i), Pn\_est(i\left)\right)$

3: end for

4: $u\_current \leftarrow softmax (u\_current)$

5: $u\_current \leftarrow LSTMNet (u\_history, u\_current)$

6: for i = 1 to 3 do

7:          $X\_est \leftarrow X\_est + Xn\_est\left(i\right) \ast u\_current\left(i\right)$

8:          $P\_est \leftarrow P\_est + Pn\_est\left(i\right) \ast u\_current(i$)

9: end for

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