Finite-Horizon Optimal Tracking Guidance for Aircraft Based on Approximate Dynamic Programming

Algorithm 1

Actor-critic learning procedure of tracking guidance.

Input:

Perturbation equations at every downrange step.

Cost function along the trajectory and at final state.

Output:

Optimal control weights for tracking reference trajectory.

(1) Randomly select sets of , calculate

and , obtain by (23).

(2) for to do

(3) Initialize , , and actor training

step .

(4) repeat

(5) Randomly select , apply previous control to calculate .

(6) Substitute to (32) gives to .

(7) Get the error of actor network from and .

(8) Calculate the gradient of weights and update , and by (35).

(9) Push training step .

(10) until

(11) Randomly select sets of , apply

actor network to get .

(12) Calculate , and according to Eqs. (26) and (27).

(13) Apply least square estimate to get .

(14) end for