Page 26 - ITUJournal Future and evolving technologies Volume 2 (2021), Issue 1
P. 26

ITU Journal on Future and Evolving Technologies, Volume 2 (2021), Issue 1




          Taking the conditional expectations of (29), we have the  neglecting the negative term and taking t → ∞, we have
          result as                                                      t−1  (                         )
                                                                        1  ∑  ∑             ∑
                             ∑
                  ∗
                                               ∗
            V E{f (t)|Θ(t)} +   E{X i |Θ(t)}(t)(p (t) − γ i )       lim           E {X i (t)} +  E {Z u (t)}  ≤
                                               i                   t→∞ t
                                                                         τ=0
                             i∈N                                              i∈N           u∈U
                                                                           ∗
               ∑                                                   B + V (f (ϵ) − f(t))
             +    E{Z u (t)|Θ(t)}(δ u − µ (t)) ≤                                     .                      (37)
                                     ∗
                                     u
                                                                            ϵ
               u∈U
                            ∑                                                          opt , we obtain the  inal result
            V E{f(t)|Θ(t)} +   E{X i |Θ(t)}(t)(p i (t) − γ i )  Consider that E{f(t)} ≥ f
                                                               as
                            i∈N
               ∑                                                         t−1  (                         )
             +    E{Z u (t)|Θ(t)}(δ u − µ u (t)).                       1  ∑  ∑             ∑
                                                                    lim           E {X i (t)} +  E {Z u (t)}  ≤
               u∈U                                                 t→∞ t
                                                                         τ=0  i∈N           u∈U
                                                                           ∗
                                                                   B + V (f (ϵ) − f opt )
                                                                                     .                      (38)
                                                                            ϵ
          D. PROOF OF THEOREM 2
                                                               This shows that the queues are strongly stable for ϵ > 0.
          Proof. Suppose that a feasible policy ω exists, i.e., con‑
          straints (12)b and (12)c are satis ied. Furthermore, as‑
          sume that, for the ω policy, the following holds     REFERENCES
                         E{p i (t) − γ i } ≤ −ϵ,      (30)     [1] F. Alriksson, L. Bostrom, J. Sachs, E. Wang, and A. Zaidi,
                                                                                    ̈
                        E{δ u − µ u (t)} ≤ −ϵ,        (31)         “Critical IoT connectivity,” Ericsson review, vol. 102,
                            E{f (ϵ)} = f (ϵ),         (32)         pp. 52–64, 2020.
                                ∗
                                        ∗
          where f (ϵ) is a sub‑optimal solution. Applying (30) and  [2] I. Hou and P. R. Kumar, “Packets with deadlines: A
                 ∗
          (31) into (18), we obtain                                framework for real‑time wireless networks,” Synthe‑
                                                                   sis Lectures on Communication Networks, vol. 6, no.
            E{L(Θ(t + 1))} − E{L(Θ(t))} + V E{f(t)} ≤              1, pp. 1–116, 2013.
                  (                         )
                   ∑             ∑                             [3] I. Hou, V. Borkar, and P. R. Kumar, “A theory of QoS for
                                                   ∗
            B − ϵ      E{X i (t)} +  E{Z u (t)}  + V f (ϵ),
                                                                   wireless,” in Proc. IEEE INFOCOM, pp. 486–494, 2009.
                   i∈N           u∈U
          taking ϵ → 0 and the sum over τ = 0, . . . , t − 1 we obtain  [4] S. Lashgari and A. S. Avestimehr, “Timely throughput
                                                                   of heterogeneous wireless networks: Fundamental
             t−1
           1  ∑           −E{L(Θ(t))} + E{L(Θ(0))} + Bt            limits and algorithms,” IEEE Trans. Inf. Theory, vol.
                E{f(τ)} ≤
            t                           V t                        59, no. 12, pp. 8414–8433, 2013.
             τ=0
                        + f opt ,                     (33)     [5] O. Holland, E. Steinbach, R. V. Prasad, Q. Liu, Z. Dawy,
                                                                   A. Aijaz, N. Pappas, K. Chandra, V. S. Rao, S. Oteafy
          taking t → ∞, we obtain
                                                                   et al., “The IEEE 1918.1 ”tactile internet” standards
                            t−1                                    working group and its standards,” Proc. of the IEEE,
                          1  ∑                 B
                   lim sup     E{f(τ)} ≤ f opt  +  .  (34)         vol. 107, no. 2, pp. 256–279, 2019.
                  t→∞     t                    V
                           τ=0
                                                               [6] Y. Cui, V. K. N. Lau, R. Wang, H. Huang, and M. Shunqing
          That concludes the second part of Theorem 2. In order to
          prove the stability of the queues, we manipulate (33)    Zhang, “A survey on delay‑aware resource control for
                                                                   wireless systems–large deviation theory, stochastic
          (                          )
                                                                   lyapunov drift, and distributed stochastic learning,”
            ∑             ∑
               E {X i (t)} +  E {Z u (t)}  ≤                       IEEE Trans. Inf. Theory, vol. 58, no. 3, pp. 1677–1701,
           i∈N            u∈U                                      Mar. 2012.
          B   E{L(Θ(t + 1))} − E{L(Θ(t))}   V (f (ϵ) − f(t))
                                               ∗
            −                            −               .     [7] N. Master and N. Bambos, “Power control for packet
          ϵ                ϵ                      ϵ
                                                      (35)         streaming with head‑of‑line deadlines,” Performance
                                                                   Evaluation, vol. 106, pp. 1–18, 2016.
          By taking the sum over τ = 0, . . . , t − 1 and divide by t,
          we obtain                                            [8] N. Nomikos, N. Pappas, T. Charalambous, and Y.‑A.
                (                          )                       Pignolet, “Deadline‑constrained bursty traf ic in ran‑
             t−1
           1  ∑   ∑             ∑                                  dom access wireless networks,” in IEEE Proc. SPAWC,
                     E {X i (t)} +  E {Z u (t)}  ≤
           t                                                       pp. 1–5, 2018.
             τ=0  i∈N           u∈U
                                              ∗
           B    E{L(Θ(t))} − E{L(Θ(0))}   V (f (ϵ) − f(t))     [9] A. Faridi and A. Ephremides, “Distortion control for
             −                          +               ,
            ϵ              tϵ                   ϵ                  delay‑sensitive sources,” IEEE Trans. Inf. Theory, vol.
                                                      (36)         54, no. 8, pp. 3399–3411, 2008.
          10                                 © International Telecommunication Union, 2021
   21   22   23   24   25   26   27   28   29   30   31