A Multiband MAC Protocol for Impulse-based UWB Ad Hoc Networks Ioannis Broustis, Srikanth V. Krishnamurthy, Michalis Faloutsos, Mart Molle and Jeffrey R. Foerster {broustis, krish, michalis, cs.ucr.edu intel.com Ioannis Broustis, Srikanth V. Krishnamurthy, Michalis Faloutsos, Mart Molle and Jeffrey R. Foerster {broustis, krish, michalis, cs.ucr.edu intel.com

2 The context Wireless needs High Speed networking Low cost, low power transport Home, enterprise environments Current wireless solutions Low data rates, high power consumption UWB pros High data rates Low-power operation and low cost Low probability of detection Low interference levels Picture from

3 Motivation & contribution A lot of work has been done in the PHY layer of UWB Only a few MAC proposals for UWB Most of them for master-slave deployments Many assumptions - some of them cannot be implemented in the real world Some do not take into account the PHY characteristics We design and evaluate a novel multiband MAC protocol for UWB ad hoc networks Utilizes efficiently the available bandwidth Achieves much better performance than other MAC protocols for Ad Hoc UWB Conforms with FCC regulations

4 Roadmap UWB OverviewThe problemOur MAC protocol Simulation ResultsConclusions

5 UWB definitions Any signal that occupies: At least 500 MHz of bandwidth, or More than 25% of a fractional bandwidth: Available bandwidth: 7500 MHz FCC has allocated the band from 3.1 GHz to 10.6 GHz for UWB communications Emission levels must fall under max limits (average dBm/MHz) Traditionally: pulse transmissions Range: 0 to 15m UWB Spectrum (7.5 GHz) a (0.1 GHz) Frequency (GHz) EIRP FCC limit: dBm/MHz PSD

6 Bandwidth utilization Single-Band Implementation One transmission occupies the whole BW at a time Multi-Band Implementation The 7.5 GHz are divided into multiple bands FCC regulations must be obeyed Benefits from multiband approach Low interference from/to systems that share a portion of the BW Parallel data transmissions in the different bands Similar H/W cost with single-band implementations a ( GHz) EIRP

7 Time Hopping, as per Time Hopping Sequences (THS) Binary Pulse Position Modulation Many pulses per bit, to increase reliability THS overlap  Pulse collisions Tx, Rx based THS PAM also possible Impulse-based UWB TfTf TcTc THS 1 : 0, 3, 2, 6 THS 2 : 4, 6, 3, time T c frame 01

8 Roadmap UWB OverviewThe problemOur MAC protocol Simulation ResultsConclusions

9 What is the problem? UWB pulses are subject to Multipath Delay Spread Multiple time-shifted pulse copies appear at the receiver Intersymbol Interference (ISI) Tens of nanoseconds (~ 25 to 30nsec for indoor environments) Collisions at the receiver, with subsequent pulse transmissions –From the same or different transmitter A B Power Delay Profile time obstacle A

10 Potential solutions Equalizers, CDMA + Rake receiver Add overhead and Hardware cost Pulse spacing at least equal to the delay spread duration The adoption of a multi-band mechanism does not reduce the data rate A set of carriers modulate the pulse in each band and determine the pulse shape Single-band Multi-band time T c frame Pulse width Delay Spread ~0.3nsec ~3nsec (10 bands)

11 Roadmap UWB OverviewThe problemOur MAC protocol Simulation ResultsConclusions

12 MAC overview We divide the available BW into B bands One band for requests and band information. The rest for data transmissions and ACKs Map of Band availability Superframes: Transmission of all control and data packets Availability frames: Declare intention to keep using a band time frequency Control (REQ) Data 1 Data 2 Data 3 Data 4 Data B-1 Superframe Availability frame k1k1 k3k3 k B-1 ….. k2k2..…..

13 Bandwidth: each of our bands is 500 MHz wide Emission limits : dBm/MHz For the received SNR we have: Attenuation for each band P T : Transmitter PSD ( dBm/MHz) N 0 : PSD of the thermal noise (-114 dBm/MHz) d: Tx-Rx distance SNR R = 3 dB f c for the upper band For the last band: f c = GHz  distance ~ 7 meters We set this distance as the maximum distance for all bands We use lower transmission powers for the other (lower) bands We conform with the average power and the pulse frequency is  1 MHz.  We conform with the peak power constraint as well :) Conformance with FCC regulations contribution

14 Nodes that intend to keep occupying a data band, transmit a short beacon during the availability frame The rest of the nodes “listen” to the whole availability frame Information about which bands will be occupied during the upcoming superframe MAC details: band selection Availability frame Superframe Data band k slot k REQ band

15 MAC details: request (REQ) initiation The REQ packet is transmitted in the Req-band It includes the selected band of the Tx The receiver’s THS is used Nodes are allowed to initiate a REQ transmission only at the beginning of a superframe Availability frame Superframe REQ (Receiver’s THS)REQ band Data band Free

16 4 possible cases 1. Everything goes fine The receiver decodes the request Both nodes switch to the selected band The receiver sends the RACK packet (consecutive pulses) The Data and DACK packet transmissions follow MAC details: REQ acknowledgment Availability frame Superframe REQ REQ band Data band RACKDATADACK

17 4 possible cases 2. Two or more requests towards the same receiver collide The receiver cannot decode the request The transmitters switch to their selected bands, waiting for the RACK After a specific time interval they will assume that their request did not reach the receiver Backoff timers are initiated (decreased by one per superframe) When backoff=0 the node retransmits the request MAC details: REQ acknowledgment REQ (same THS) Superframe … … … Superframe REQ (same THS) REQ band Data band Response not received Availability frame REQ (same THS) Back-off countdown Availability frame

18 4 possible cases 3. The intended receiver is currently busy The receiver will not hear the request The transmitter however will switch to its selected band The transmitter initiates a backoff timer and retransmits the request as soon as this timer becomes zero MAC details: REQ acknowledgment Availability frame Superframe … … … Superframe REQ towards node C REQ band Data band Data-band DATA chunk from C to D DACK REQ Back-off countdown Response not received

19 MAC details: REQ acknowledgment 4 possible cases 4. Two or more RACKs collide If two or more transmitters select the same band, a RACK collision is likely to occur in that data band Further actions are temporarily aborted, until the upcoming availability frame The requests are retransmitted after the end of the upcoming availability frame With our policy, Data packet collisions are avoided Availability frame REQ Superframe REQ REQ band Data band RACK Abort Temporarily

20 MAC details: DATA and DACK The RACK, DATA and DACK packets are transmitted with consecutive pulses After the end of the session, transmitter and receiver switch to the REQ band If they don’t have packets to send, they stay idle listening to their own THSs Availability frame Superframe REQ REQ band Data band RACKDATADACK

21 Roadmap UWB OverviewThe problemOur MAC protocol Simulation ResultsConclusions

22 Comparisons We compare our scheme with a single-band approach, in which: THSs are used for all kinds of packets. Each pair of nodes has a predetermined common - unique THS Steps: REQ RACK DATA DACK READY A B The Tx sends a request to the Rx as per the Rx’s THS Both Tx and Rx switch to their common THS. The Rx sends a reply back The Tx further transmits the data packet The Rx sends an ACK as soon as it receives the data packet Both Tx and Rx switch to their own THSs. They further transmit a short beacon to indicate their availability

23 Simulation set-up Simulator in C++ Nodes6 to 30 Bands15 Region30x30 m 2 square, multi-hop Range7 meters Node degree  3, Brownian motion Ratio T f / T c 6 Bit repetition2, with 1/3 conv. encoder T c chip60 nsec, (2 x delay spread) Superframe11200 chips Availability frame14 slots, 33 chips each Light trafficCBR, arrival every 40 msec Heavy trafficCBR, arrival every 1.4 msec Poisson, (lambda = 5.028) Data packet250 bytes Control packets15 bytes

24 Simulations: pulse collisions Decreased by an order of magnitude Data packets in our case are collision-free

25 The bit error rate is decreased by more than 4 times in our case Simulations: BER

26 Time from: packet arrival in the queue until completion of its transmission Decreased by a factor of 6 for low densities Time from: packet arrival in the queue until completion of its transmission Decreased by a factor of 6 for low densities Simulations: average packet delay

27 Higher as much as 16.7% in our case Light traffic  beneficial for the single-band case Would observe larger difference with heavier traffic  Higher as much as 16.7% in our case Light traffic  beneficial for the single-band case Would observe larger difference with heavier traffic  Simulations: average network throughput

28 High CBR arrival rate More than an order of magnitude better throughput in our case High CBR arrival rate More than an order of magnitude better throughput in our case Simulations: average throughput for high loads

29 Roadmap UWB OverviewThe problemOur MAC protocol Simulation ResultsConclusions

30 Conclusions We propose a novel multiband MAC protocol for UWB ad hoc networks Better network performance than previous impulse-UWB MAC No equalizer or CDMA required to address the delay spread effects Utilizes efficiently the 7.5 GHz bandwidth Adopts all the advantages of a multiband UWB approach Respects the FCC regulations Our ongoing work with UWB: 1. New multiband MAC that employs binary conflict resolution Applicable for home, office and wearable ad hoc networks Demonstrates much better performance in terms of throughput and delay

31 Questions? (References available upon request)