Joint Load Balancing and Offloading Optimization in Multiple Parked Vehicle-Assisted Edge Computing

Algorithm 2

Workload allocation algorithm based on dynamic game.

 Input:The task , ; the node selection variable x_ij; the offloading task workload ω.

 Output:The final workload strategy , , , let j = 0 denotes local, j > 0 denotes edge node.

 Initialization:k_ij = 0, u_i =0.

Step 1. Workload allocation between nodes.

1: for each task D_i

2:  whileω_i < c_ido

3:   calculate the incremental cost of processing unit-sized tasks locally

4:   ifthen

5:    calculate

6:   end if

7:   compare

8:   allocate subtask to node j with the smallest cost increment，thenk_ij = k_ij + ω. u_i = u_i + Δu_ij

9:   setω_i = ω_i + ω

10:  end

11: setk_i = {k_i0, k_i1,k_i2, …, k_ij}

Step 2. Iterative to NE based on backward induction.

12: leti = i + 1 and repeat step 1

13: calculatek_i+1 = {k_i+1,0, k_i+1,1,k_i+1,2, …, k_i+1,j}, u_i+1

14: sendk_i+1 to user i and repeat step 1

15: calculatek^’_i, u^’_i

16: Ifu^’_i < u_ithen

17:  updatek_i = k^’_i

18:  Sendk_i to user i + 1 and repeat step 1

19:  n = n + 1

20:  calculatek^’_i+1, u^’_i+1

21:  repeat step 2 until and

22:  set and

23: end if

24: returnk_i, , u_i, u_i+1