Coexistence Mechanism Using Dynamic Fragmentation for Interference Mitigation between Wi-Fi and Bluetooth David S. L. Wei Joint Work with Alex Chia-Chun Hsu and C.-C. Jay Kuo

Outline Overview of Wi-Fi and Bluetooth Previous Work Dynamic Fragmentation Algorithm Results Conclusion and Future Work

Content Overview of Wi-Fi and Bluetooth /Wi-Fi /Bluetooth Coexistence in the UL band Previous Work Dynamic Fragmentation Algorithm Results Conclusion and Future Work

Overview of Wi-Fi and Bluetooth Wi-Fi A dominating WLAN standard Medium to high date rate, medium range < 100m Use the ISM band Large selection of commodities a b g n Standard OFDM DSSS (CCK) CCK/OFDM OFDM Modulation 5GHz 2.4GHz Frequency 6 ~ 54 Mbps 2~11 Mbps 20 ~ 54 Mbps >100 Mbps Data Rate 60 ft. 300 ft. Max. Distance Industrial, Scientific, medical

Overview of Wi-Fi and Bluetooth Medium Access Control CSMA/CA Carrier Sense Multiple Access / Collision Avoidance Virtual Carrier Sensing Request-to-send/Clear-to-send RTS/CTS Network Allocation Vector NAV

Distributed Inter Frame Space Overview of Wi-Fi and Bluetooth Example: DCF mode Distributed Coordination Function Time DIFS Backoff Window BW RTS set NAV Short Inter Frame Space SIFS CTS DATA ACK BW* DIFS RTS

Overview of Wi-Fi and Bluetooth Bluetooth Popular WPAN standard Low data rate, low cost, short range < 10m use frequency hopping to avoid collision 1600 hops/s ~ 625 μs in every frequency channel SCO and ACL link Synchronous Connection-Oriented link Real-time application: voice stream HV-3 link: a packet is generated every 6 time slots Asynchronous ConnectionLess link Non-time-critical application: data traffic DH-1/3/5 link: one packet occupied 1/3/5 time slot

Overview of Wi-Fi and Bluetooth Coexistence in the UL band Time Frequency 2.4 GHz MHZ 1 MHZ 625 μs BT Channel 6 22 MHZ Wi-Fi

Overview of Wi-Fi and Bluetooth Coexistence in the UL band Packet loss caused by interference Overlap both in time and in frequency Over the SNR threshold

Content Overview of Wi-Fi and Bluetooth Previous works Adaptive Frequency Hopping D-OLA and V-OLA Fragmentation Dynamic Fragmentation Algorithm Results Conclusion and Future work

Previous Work Coexistence Working Group Many suggestions on improving coexistence Collaborative solutions Devices could exchange information Collocated and under a central controller Non-collaborative solutions No information exchange Most common scenario

Previous Work Adaptive Frequency Hopping Enhancement on BT, many variations Non-collaborative solution Distinguish good channels from bad ones Keep the hopping sequence on good channels more frequently

Previous Work Adaptive Frequency Hopping Frequency 2.4 GHz BT Wi-Fi

Previous Work D-OLA and V-OLA Proposed by Chiasserini and Rao, Infocom 2004 Data OverLap Avoidance Use different BT packet length to avoid bad channels Voice OverLap Avoidance Wi-Fi estimate the interference pattern of real- time packet Shorten Transmission or Postpone Transmission  Increase delay  Not a pure non-collaborative solution

Previous Work D-OLA 2.4 GHz BT Wi-Fi

Previous Work V-OLA

Previous Work Fragmentation DIFS BWDATA hdr ACK DIFS BW DATA1 ACK1 No fragmentation 2 fragments DATA2 ACK2

Previous Work Fragmentation Adaptive Fragmentation from Adjust fragmentation according to Packet Error Rate PER Many rounds before reach optimal length Optimal Fragmentation by Howitt 2005 Complexity is too high to determine the optimal fragment length at run time  No resolution on collision and interference  Need simple run time solution

Content Overview of Wi-Fi and Bluetooth Previous works Dynamic Fragmentation Algorithm Interference model State diagram of DFA Determine threshold Optimization Results Conclusion and Future Work

Dynamic Fragmentation Algorithm Interference model DATA hdr ACK 625 μs 366 μs P f : Probability of BT hops on Wi-Fi frequency N : # of BT time slot overlapped by Wi-Fi packet τ BT : Traffic load of BT σ : utilization of BT time slot N is crucial

Dynamic Fragmentation Algorithm State Diagram of DFA 2 states State 1, no fragmentation State 2, DATA → n fragments PER greater than P2, one state up if possible, further fragmentation PER lower than P1 → one state down if possible  How to choose P1, P2? 1 2 PER≤P2 PER>P2 PER≥P1 PER<P1

Dynamic Fragmentation Algorithm Determine Threshold DIFS BWDATA hdr ACK DIFS BW DATA1 ACK1 DATA2 ACK2 DIFS BW DATA1 ACK1 DATA1 ACK1 Retransmission DIFS BW Double the backoff window

Dynamic Fragmentation Algorithm Determine Threshold Time to transfer a packet with n fragments and suffer R retransmissions Compare the transmission time before and after state transition If true, then state transition is beneficial

Dynamic Fragmentation Algorithm Determine Threshold Case 1: less or equal to BW upper-bound Case 2: greater than BW upper-bound Before state transition P is the current PER Assume geometric distribution

Dynamic Fragmentation Algorithm Determine Threshold How to find PER after state transition? Same as previous slide Now we have all the parameters to calculate a theoretically correct threshold

Dynamic Fragmentation Algorithm Optimization Timing Cause Solution Wi-Fi From the beginning of a transmission Traffic Jam CSMA/CA Collision Most likely not Coexistence (BT) Coexistence Mechanism Interference  Transmission failure on following fragments is due to Interference DIFS BW DATA1 ACK1 DATA2 ACK2 BW DATA1 ACK1 DATA2 ACK2 DATA2 ACK2 DIFS BW With optimization DIFS DATA2 ACK2 DIFS

Dynamic Fragmentation Algorithm Optimization If true, then state transition is beneficial Only first fragment needs backoff window when retransmission DFAm : with node mobility DFAs : static network (throughput performance can be optimized)

Content Overview of Wi-Fi and Bluetooth Previous work Dynamic Fragmentation Algorithm Results Conclusion and Future work

Result Simulation PER equation validation

Result Simulation Threshold equation validationThroughput improvement

Result Simulation: ACL link Throughput of the Wi-Fi and BT in the presence of BT ACL link

Result Simulation: 2 SCO links Throughput of the Wi-Fi and BT in the presence of 2 SCO links between BT master/slave

Result Simulation: delay Average Wi-Fi delay vs. BT traffic load

Content Overview of Wi-Fi and Bluetooth Previous works Dynamic Fragmentation Algorithm Results Conclusion and Future work

Simple non-collaborative mechanism Increase collision/interference resolution Improve throughput and delay Built a reliable, powerful model Conclusion and Future work Conclusion If PER > 0.6, DFAs 56%, DFAm 30%

Conclusion and Future Work Cognitive Radio Existing policy fragmented the spectrum Bandwidth is scarce and expensive Good frequencies are taken Recent measurements by FCC shows 70% of the allocated spectrum is not utilized (US) Improve spectrum efficiency Unlicensed bands Need new solution for upcoming wireless service

Conclusion and Future work Cognitive Radio Paradigm shift – Cognitive Radio by Mitola 1999 “ radio or system that senses its operational electromagnetic environment and can dynamically and autonomously adjust its radio operating parameters to modify system operation, such as maximize throughput, mitigate interference, facilitate interoperability,…” IEEE , FCC, DARPA XG, OverDRiVE, SWRF, WWRF, …

Conclusion and Future work Cognitive Radio Cognitive radio requirements coexist with legacy wireless systems use their spectrum resources do not interfere with them Cognitive radio properties RF technology that "listens" to huge swaths of spectrum Knowledge of primary users’ spectrum usage Rules of sharing the available resources Embedded intelligence to determine optimal transmission based on primary users’ behavior

Conclusion and Future work Cognitive Radio: spectrum hole F1 F2 F3 F4 F5 Exclude Spectrum Gray spaceWhite space Spectrum hole Time Frequency

Thanks ^_^