1. Adaptive Power Control Algorithm for Ad Hoc Networks with Short and Long Term Packet Correlations Jun Zhang, Zuyuan Fang, and Brahim Bensaou Dept. of Computer Science, The Hong Kong University of Science and Technology IEEE Conference on Local Computer Networks (LCN), 2005

2. Outline Introduction Introduction Adaptive Power Control Based Packet Delivery Curve Adaptive Power Control Based Packet Delivery Curve Correlative Adaptive Power Control (CAPC) Correlative Adaptive Power Control (CAPC) Performance evaluation Performance evaluation Conclusion Conclusion

3. Introduction There are two broad ways to achieve less energy consumption There are two broad ways to achieve less energy consumption –Topology control by making the topology sparser, the ambient interference is reduced –Power control

4. Basic Concept of Power Control Protocols To adjust the transmission power (TP) level to the suitable level according to the network condition To adjust the transmission power (TP) level to the suitable level according to the network condition

5. Major Differences of Power Control Protocols The major difference between most power control algorithms lies in The major difference between most power control algorithms lies in 1.How to adapt the power to the network status 2.What metric to use to reflect the network status 3.How to track such status

6. Packet Delivery Curve Example # of transmitted packets # of packets received successfully

7. Feasible and Infeasible Packet Curves L: the longest streak of packet losses M: packet delivery curve count limit

8. Packet Delivery Curves (C2) Case 1 : curve drops below S = p * ( T – L ), it means either Case 1 : curve drops below S = p * ( T – L ), it means either –Avg. packet loss ratio is much higher than p –The longest streak of frame losses is much longer than L Infeasible power level, should increase power and reset the curve Infeasible power level, should increase power and reset the curve

9. Packet Delivery Curves (C1) Case 2: curve is always above S = p * ( T – L ), but is below S = p * T, when T = M Case 2: curve is always above S = p * ( T – L ), but is below S = p * T, when T = M –Current avg. packet loss ratio is higher than p –Longest streak of loses is less than L Infeasible power level, should increase power and reset the curve Infeasible power level, should increase power and reset the curve

10. Packet Delivery Curves (C0) Case 3: curve is always above S = p * T, exactly at S = p * T when T = M Case 3: curve is always above S = p * T, exactly at S = p * T when T = M Feasible power level Feasible power level

11. Summary The adaptive transmission power control The adaptive transmission power control –Reduces the TP while guaranteeing a similar throughput as when max power is used –Response quickly (C2) –Not to sensitive to long streaks of losses (C0)

12. Relations between the TP of RTS-CST-DATA-ACK When packet x is transmitted at a high TP (α), the successor packet y can be transmitted at a relative lower TP (β) When packet x is transmitted at a high TP (α), the successor packet y can be transmitted at a relative lower TP (β) (x, y) are in the set (x, y) are in the set –(RTS, CTS) –(CTS, DATA) –(DATA, ACK) The larger α, the smaller β and vice versa The larger α, the smaller β and vice versa

13. Relations between the TP of RTS-CST-DATA-ACK RTS frame may be discarded when the receiver is not idle RTS frame may be discarded when the receiver is not idle –The NAV of the receiver is not zero Increasing TP under this scenario is not suitable Increasing TP under this scenario is not suitable

14. Correlative Adaptive Power Control (CAPC) Each station maintains 2 curves for a neighbor Each station maintains 2 curves for a neighbor –RTS –DATA

15. Finite State Machine M: parameter measurement state M: parameter measurement state –Measure p and L R: RTS/CTS TP-lower bound state R: RTS/CTS TP-lower bound state –Do not allow to decrease RTS TP D: DATA/ACK TP-lower bound state D: DATA/ACK TP-lower bound state –Do not allow to decrease DATA TP

16. Algorithm Description

17. Simulation Scenario TCP congestion window is 32 TCP congestion window is 32 Packet size is 512 B Packet size is 512 B CBR for each UDP is 400 packets/s CBR for each UDP is 400 packets/s Max TP is 250 M Max TP is 250 M R1, R2 are 400, 4000 R1, R2 are 400, 4000 M is 25 M is 25

18. End-to-end Throughput

19. Throughput/Energy Consumption Ratio

20. Fairness Index

21. End-to-end Throughput with CBR Traffic

22. Throughput/Energy Consumption Ratio with CBR Traffic

23. Fairness Index with CBR Traffic

24. Conclusion This paper proposed a CAPC algorithm This paper proposed a CAPC algorithm –Relies on packet delivery curve CAPC tries to achieve the same throughput as IEEE 802.11 DCF but uses lowest TP as possible CAPC tries to achieve the same throughput as IEEE 802.11 DCF but uses lowest TP as possible

25. Thank you!!