Deep Reinforcement Learning-Based UAV Path Planning for Energy-Efficient Multitier Cooperative Computing in Wireless Sensor Networks

Algorithm 1

DQN algorithm for UAV path planning.

1: Randomly initialize predict network parameter and target network parameter .

2: forepisode in range (Max_Iteration) do

3: Initialize position and energy of UAV, position and energy of nodes, and task parameters.

4: whiledo

5:  Get the current state from environment.

6:  Select UAV action with (27).

7:  Get reward with (26) and then the system transits to the next state .

8:  Store the record to the experience pool.

9:  Randomly sample records from the experience pool to form the minibatch.

10: Compute the target value with (29).

11: Update the parameters of the predicted Q network by minimizing the loss function (28).

12: Update parameters by after a preset number of steps.

13: end while

14: end for