An Optimal Distributed Call Admission control for Adaptive Multimedia in Wireless/Mobile Networks

1. An Optimal Distributed Call Admission control for Adaptive Multimedia in Wireless/Mobile Networks Reporter:電機所 692415088 鄭志川

2. Outline • Abstract • Introduction • Adaptive multimedia and the Idea of CAC-BRA scheme • Modeling CAC-BRA • Linear programming and Simplex method • Numerical results

3. Abstract • There is a great demand on multimedia application with Quality of Service (QoS) in wireless/mobile networks. • propose an optimalcall admission control framework with bandwidth reallocation algorithm(CAC-BRA). • We adopt semi-Markov Decision Process (SMDP) approach to model call admission control and bandwidth reallocation algorithm at the same time. • Simplex method in linear programming is used to solve the optimal decision problem.

4. Introduction • Non-adaptive situation • Adaptive situation, where the bandwidth of an ongoing call is time varying during its lifetime. • References give only a sub-optimal or near optimal solution, but we adopt semi-Markov decision process approach (SMDP), which gives us an optimal solution.

5. Adaptive multimedia and the Idea of CAC-BRA scheme The layered coding approach • In order to simplify the problem, we assume only one class of users. Any customer uses one bandwidth among {b1,b2,…,bK} where bj<bj+1 for j = 1,2,…,K-1, If K=1, we have a non-layered coding traffic. Example: For a video stream coded with H.263, MPEG-2 I frames, and MPEG-2 whole (I,B and P) frames. Here K=3, b1:H.263, b2:MPEG-2 I frames, b3:MPEG-2 whole frames.

6. Adaptive multimedia and the Idea of CAC-BRA scheme The idea of CAC-BRA • CAC-BRA decides on not only whether an arrival will be accepted or not, but also which call will be changed to how much bandwidth. • We don’t allow the forced-termination of existing calls in the cell. The arrival calls, new arrival calls and handoff arrival calls, could be blocked. • Our goal is to maximize the revenue, and at the same time to satisfy the QoS parameters, which are an upper bounds on handoff blocking probabilities.

7. 1.Decision Epoch 2.State Space 3.Active Space 4.Reward Function 5.Transition Probability Maximum Revenue And QoS guarantee Constraints Linear Programming Uniformization In MDP + Discrete time Modeling CAC-BRA Consider a dynamic system

8. :the new call arrival rate for customers (poisson distribution) :the rate of handoff to other cells h :the handoff call arrival rate for customers µ :the service rate for customers (exponential distribution) CAC-BRA

9. CAC-BRA The state of the system is X=(x1,x2,…,xK) Where xi stands for the number of users who is using bandwidth bi in the system for all 1 ≤ i ≤ K Decision Epoch V=(X , a) where X is the current state, and the variable a represents both the event type and the action.

10. New call Departure from the cells with bg Handoff call Handoff call accepted to have bandwidth bf2 The action of bandwidth reallocation if it is a new call Arrival and the arrival is accepted to have bandwidth bf1 CAC-BRA The State Space The Action

11. CAC-BRA New call or handoff arrival Departure from the cells

12. The New State CAC-BRA

13. The New State CAC-BRA

14. :the new call arrival rate for customers (poisson distribution) :the rate of handoff to other cells h :the handoff call arrival rate for customers The expected sojourn time µ :the service rate for customers (exponential distribution) CAC-BRA The Action Space are defined and Constrained by (1)-(15);

15. CAC-BRA The Transition Probability The Revenue Rate

16. Linear programmingand Simplex method Constrain New state Dropping Probability

17. Numerical results

18. Numerical results

19. Numerical results

20. Conclusion and future work • SMDP • Mathematics • Matlab