Skip to main content
. 2026 May 17;28(5):561. doi: 10.3390/e28050561
Algorithm 1 PA-BAC for optimal user association at slot t.
  •   1:

    Input: Given UAV deployment positions {pbm∗(t)}m=1M; ship positions {pun(t)}n=1N; priority weights {w(un)}n=1N.

  •   2:

    Output: Optimal association α∗(t).

  •   3:

    Compute link feasibility and rates for all UAV–ship pairs using the channel model; obtain γbm,un(t) and Rm,n(t).

  •   4:

    Formulate the BILP (objective (29) with constraints C1–C5).

  •   5:

    Perform greedy association from higher to lower priority to obtain a feasible initial association α0(t), and compute Hw0(t).

  •   6:

    Initialize the active problem queue Q and push the initial BILP into Q. Set the current best as α∗(t)←α0(t),Hw∗(t)←Hw0(t).

  •   7:

    while Q is not empty do

  •   8:

          Pop and remove a subproblem I from the front of Q.

  •   9:

          Solve the LP relaxation of I, obtaining the relaxed solution α′(t) and the objective value Hw′(t).

  • 10:

          if α′(t) is infeasible or Hw′(t)≤Hw∗(t) then

  • 11:

                Prune I and return to Step 7.

  • 12:

          else if α′(t) is an integer (integral) solution then

  • 13:

                Update: α∗(t)←α′(t),Hw∗(t)←Hw′(t), and return to Step 7.

  • 14:

          else

  • 15:

                Search for cutting planes violated by α′(t) and collect them into the cut set L.

  • 16:

                if L is not empty then

  • 17:

                      Add L into subproblem I, and return to Step 9.

  • 18:

                else

  • 19:

                      Decompose the original problem into restricted feasible subproblems via branching; perform priority-guided branching by preferentially selecting variables corresponding to higher-priority users as branching variables; push the newly generated subproblems into queue Q, and return to Step 7.

  • 20:

                end if

  • 21:

          end if

  • 22:

    end while

  • 23:

    return  α∗(t)