Machine Learning-Based Two-Stage Task Offloading Optimization for Power Distribution Internet of Things

Algorithm 1

ML-based two-stage task offloading optimization algorithm.

1: Input:, , .

2: Output: and

3: Phase 1. Initialization

4:  Initialize .

5: Fordo

6:  Phase 2. Large-Timescale First-Stage Server Selection

7:  Step 1:

8:  Initialize , , and .

9:  Step 2:

10:   and calculate the preference values and

   based on (11) and (12) and establish the preference lists and .

11:  Step 3:

12:  While and do

13:    proposes to its most preferred server based on .

14:   Fordo

15:    If the sum of temporary matches and new proposals for is less than quota then

16:     Temporarily match with the devices, update

  , and remove the matched devices from .

17:    else

18:     Temporarily match with its most preferred

   devices and update . Remove matched devices

  from and add unmatched devices into . Unmatched

  devices remove from .

19:    End if

20:    If the sum of matches for is equal to then

21:     Remove from .

22:    End if

23:   End for

24:  End while

25:  Fordo

26:   Phase 3. Small-Timescale Second-Stage Channel

  Selection

27:    makes the action decision based on (18).

28:    calculates and based on (14) and

  (15)

29:   Update and based on (16) and (17).

30:  End for

31: End for