A Rate-Adaptive MAC Protocol for Multi-hop Wireless Networks

1. A Rate-Adaptive MAC Protocol for Multi-hop Wireless Networks 황 태 호 taeo@keti.re.kr

2. Gavin Holland • Texas A&M University • Nitin Vaidya • Texas A&M University • Department of Electrical and Computer Engineering Co-Director, Illinois Center for Wireless Systems Research Professor • Paramvir Bahl • Microsoft Research • ACM SIGMOBILE July 2001, Rome, Italy

3. Introduction - 1 • in WLAN (IEEE 802.11) • Devices can transmit at 11 Mbps, with 54 Mbps • Number of encoded bits per symbol  Data rate • Modulation in mobile wireless networks • path loss, fading, interference  SNR, BER variations • Support Multi-Modulation scheme • BPSK • QPSK • QAM16 • QAM64 • QAM256 • Tradeoff emerges between modulation schemes. • The higher the data rate, the higher the BER • Figure 1, Figure 2

4. Introduction - 2

5. Rate Adaptation • Dynamically switching data rates to match the channel conditions • Two Aspect • Channel quality estimation • Measuring Signal Strength, Symbol error rate, etc • Prediction of future quality • Rate selection • Channel Quality Prediction  Threshold selection • Minimize the delay between prediction and selection

6. Previous Work on rate adaption • Ref. [19]. Dual Channel Slotted ALOHA • Separate control channel • Receiver feedback to sender • Ref. [15]. Auto Rate Fallback(ARF, 802.11) • Lucent’s WaveLAN II • The sender selects the best rate based on previous tx data. • Ref [9]. Adaptive Transmission Protocol • Selects based on cached per-link information • Separate transmit receive tables • Maintained by exchanging control packet(RTS/CTS) • Cellular network • Channel quality estimation by the receiver • Rate selection by the sender using the feedback • Reside at the physical layer (symbol-by-symbol)  Improper to MAC based on contention access

7. Motivation • ARF Protocol • Receiver • Channel Quality Estimation • Rate Selection

8. Overview of IEEE 802.11 • Src sends a data packet to Dst • Transmission using one of basic rate set • All node can demodulatethe RTS/CTS packets • Virtual carrier sense • RTS includes DRTS • CTS includes DCTS • NAV • Network Allocation Vector • The aggregate durationof time that medium is presumed to be busy

9. Receiver-Based Autorate (RBAR) Protocol • The receiver selects the appropriate rate for the data packet during the RTS/CTS exchange • More accurate rate selection • Smaller overhead for the channel quality estimation • In control packet • Instead of DRTS ,DCTS  modulation rate and packet size • Src • chooses a data rate based on some heuristic method • Send RTS • Dst • Estimate the channel condition • Send CTS • Node A, B • Calculates the duration • Update NAV • Reservation SubHeader (RSH) in the MAC header of the data packet

10. Incorporation of RBAR into 802.11 • Data Packet • Header Check Sequence • RTS/CTS • Rate and Length • PLCP header • RSH rate

11. Simulation Environment • NS-2 • Extensions from the CMU Monarch project for modeling mobile ad hoc networks • Number of traffic generators • PHY/MAC/Networking stacks • Addition • Detailed MAC and PHY models • Modulation and rate adaption • Rayleigh fading simulator • Interfaces Intersil Prism II chipset • IEEE 802.11, DSSS radio, • Observation • Hot the individual rate adaption protocols reacted to the changing channel conditions

12. Simulation – ARF model • Rate selection • If no ACKs for two consecutive data packets, DOWN Rate • If received ACKs for ten consecutive data packets, UP Rate and timer cancelled • If timer expired, UP Rate • Relatively insensitive tochoice of timeout

13. Simulation – RBAR • Rate selection • Simple threshold based technique • Estimate : SNR of RTS • Select : (BER) ≤ 1E-5 , highest data rate

14. Simulation – Error Model 1 • Jake’s method • Simulation of Rayleigh fading • A finite number of oscillators with Doppler shifted frequencies • Instantaneous gain

15. Simulation – Error Model 2 • Log-distance path loss model • Friis free space propagation model • Noise model n : path loss exponent k : Boltzmann’s constant T : temperature (in Kelvin) BT : bandwidth

16. Simulation – Error Model 3 • Computed Bit Error Rate • BPSK , QPSK • M-ary QAM • Eb/N0 : bit energy to noise ratio • For gain, Coherence time • For noise, • Adjusting SNR

17. Simulation – Network Configuration • Configuration 1 • Two node • One of the nodes was fixed position, the other traveled along a direct-line path (300m) • Configuration 2 • 20 nodes • Random waypoint mobility • Random speed : 2, 4, 6, 8, 10 m/s • 1500 x 300 m2 • DSR (Dynamic source routing) Protocol • Average of 30 times

18. Performance Evaluation • Overhead of RSH

19. Slow Changing Channel Conditions • Configuration 1 • 0 ~ 300m, by 5m • 60s, Tx UDP packets(1460 bytes)

20. Fast Changing Channel Conditions • Experiment 1 • Configuration 1 • Mean node speed : 2, 4, 6, 8, 10 m/s • Single UDP Connection • Performance improvementfrom 6% (10m/s) to 20% (2m/s) • Experiment 2 • Single TCP Connection

21. Fast Changing Channel Conditions

22. Impact of Variable Traffic Sources • Configuration 1 • Bursty data sources • Pareto distribution

23. Multi-hop Performance • Configuration 2

24. Future work & Conclusion • Basic Access mode in 802.11 • Not used the RTS/CTS protocol • Hybrid scheme conditional RTS/CTS • When ACKs are lost • When Long packet size • Proposed RBAR • Optimizing performance WLAN