Cross-Layer Schemes for Antenna Array Based Wireless Ad Hoc Networks – Design and Analysis Jayakrishnan Mundarath Jointly Advised by : Prof. Parmesh Ramanathan Prof. Barry Van Veen Preliminary Examination Talk

Outline Introduction – Ad Hoc networks and MIMO Design & Analysis Perspective Research Proposal – Unified Analysis Model Conclusion

Wireless Ad Hoc Networks No infrastructure support Nodes may rely on other nodes to forward packets on their behalf Example: IEEE Wide range of applications Need higher bandwidths at lower energy expense

Popular Standard - IEEE RTS-CTS-DATA-ACK framework Single antenna – Single Spatial Reuse When node A is communicating with node B all nodes in the neighborhood of A and B must remain idle Limits aggregate network throughput A B C D E F G RTS CTS DATA ACK

Multi-Antenna Systems (MIMO) Each node has N > 1 antenna Can “beamform” transmissions (favorably or unfavorably) towards receivers Can spatially multiplex multiple data streams Can exploit array gain to lower energy consumption Solution for ad hoc networks? AB CD

Outline Introduction – Ad Hoc networks and MIMO Design & Analysis Perspective Research Proposal – Unified Analysis Model Conclusion

Preliminary Work NULLHOC – MAC/PHY protocol that increases spatial reuse using nulling (IEEE Globecom’04, revision submitted to ACM Journal of Wireless Networks (WINET)) HYB – MAC/PHY protocol that exploits both spatial reuse and multiplexing (submitted to IEEE Trans. on Wireless Comm.) QSAP – Allocates spatial reuse to satisfy QoS (submitted to IEEE/ACM Trans. on Networking) DTNS Model – Markov chain model to predict protocol performance (submitted to IEEE/ACM Trans. on Networking)

Discrete Time Network State (DTNS) Model Goal Analytical characterization of effects of cross-layer designs on performance of multi-antenna wireless ad hoc networks Accomplished Cross-layer analytical model to assess network throughput for a class of ad hoc networks Future direction Application to a wider class of networks Exploring wider range of performance metrics, e.g., energy consumption, Quality of Service (QoS).

HYB – An illustrative example for DTNS Spatial reuse + spatial multiplexing Orthogonal control and data channels (CC and DC) Single spatial reuse control channel Multiple spatial reuse data channel AB CD DE DATA CCDC CONTROL DATA CONTROL DATA

HYB : Network Evolution Control Data Time

DTNS Considerations Medium Access Framework : RTS-CTS-DATA-ACK Channel knowledge at transmitter and receiver assumed (e.g. using two-way pilot sequence exchange) Orthogonal Control and Data channels Proportion of bandwidth assigned to CC = α

DTNS Specifics (1) Maximum spatial reuse d r Maximum spatial multiplexing d m Maximum EDB = k max,α d m (1- α) k max,α = maximum spatial reuse achievable with CC bandwidth α k max,α =

DTNS Specifics (2) Actual EDB < Maximum EDB due to MAC effects – e.g. collisions Physical layer effects – e.g. transmit power, poor SNR Possibly network/higher layer effects – packet availability, QoS constraints etc. To obtain actual EDB, model network time evolution using Markov chain Given d r choose optimal α opt as solution to: Then discretize time with one time slot = one control length

DTNS : Network Evolution Model Data (3,2)(2,2)(1,2) (3,2)(2,2)(1,2) (3,2)(2,2)(1,2) (3,1)(2,1)(1,1) (3,2)(2,2)(1,2) (0,0) (3,2)(2,2)(1,2) Time Control

DTNS : Network State Representation Data (3,2)(2,2)(1,2) (3,2)(2,2)(1,2) (3,2)(2,2)(1,2) (3,1)(2,1)(1,1) (3,2)(2,2)(1,2) (0,0) (3,2)(2,2)(1,2) (0,0) (3,2) (0,0) (3,2) (2,2) (0,0) (3,2) (2,2) (1,2) (3,1) (2,2) (1,2) (3,2) (2,1) (1,2) (2,2) (1,1) (0,0) (3,2) (1,2) (0,0) (3,2) (2,2) (0,0) Time Control

DTNS Markov Chain Transition probabilities derived from model of Channel statistics and physical layer scheme Bound on transmit power of each node MAC constraints such as collision Can accommodate other constraints (1,2) (2,2) (3,2) (0,0) (1,2) (2,2) (2,1) (0,0) (1,2)

DTNS : Network Analysis Network EDB given by k av (1-α) k av is the average number of streams – obtained from steady state analysis of the DTNS Markov chain Changing constraints amounts to modifying the transition probabilities

Ex.1 : Spatial Multiplexing on Eigen channels : MRATE N antennas – transmit up to N data streams Simple extension to IEEE RTS-CTS used for channel estimation Inverse water filling – allocate available power among spatial channels to achieve equal SNR Fill from best to worst DTNS chain has N states Rayleigh flat-fading channel model

Ex.1 : Spatial Multiplexing on Eigen channels MRATE – Results with adjusted back- offs N = 8 Different total available transmit powers

Ex.2 : HYB – Hybrid Protocol N antennas – allocated for spatial reuse and spatial multiplexing Maximum spatial reuse d r and maximum spatial multiplexing d m such that d m d r < N Rayleigh flat fading channels used

Ex.2 : HYB – Hybrid Protocol d m = 1, d r = 8d m = 2, d r = 4

Ex.2 : HYB – Hybrid Protocol d m = 4, d r = 2d m = 8, d r = 1

Ex.2 : HYB – Results Model captures trends accurately Discrepancies in absolute value a consequence of some specific characteristics of the protocol Sequence of different control messages have consequence on protocol performance A coarse model for such effects accounts for ~70-80% of the discrepancies Not included here since it requires elaborate description of HYB

Outline Introduction – Ad Hoc networks and MIMO Design & Analysis Perspective Research Proposal – Unified Analysis Model Conclusion

Research Proposal Multi-rate capable ad hoc networks – e.g IEEE a/b Different rate adaptation strategies Optimal MAC? Multi-hop topology – can DTNS model performance in multi-hop topologies? Quality of Service (QoS) – increasingly important in next generation ad hoc networks – best strategy?

P1 : Multi-rate protocols IEEE a/b – supports transmissions at multiple rates Strategies for rate adaptation exist in literature and practice First goal is to assess schemes with practical physical layer models Model network as Markov chain – transitions depend on Channel model and physical layer scheme Access strategy State representation and exact nature of transitions? Second goal is to analytically design an efficient rate adaptation strategy Is there an optimal rate adaptation strategy for a given channel model? What is “optimal” in this context?

P2 : DTNS Model for Distributed Topology Current DTNS models single hop topology Multi-hop topology is more challenging Performance metric –throughput per node Use flow contention graph to represent topology State representation – requires investigation Generalize to statistical topology models

P3 : QoS in Ad Hoc Networks QoS increasingly important in Ad Hoc Networks Analytical model for QoS in MIMO networks can Provide insights for more efficient resource allocation Enable to take cross-layer effects into account Possible approaches Represent network state at time k of N nodes as a N-vector of deviations Vector u(k) represents allocation strategy at time k Model cost function and derive optimal “strategy” to Minimize deviations Drive deviations to desired value

Outline Introduction – Ad Hoc networks and MIMO Design & Analysis Perspective Research Proposal – Unified Analysis Model Conclusion

Analytical models important for next generation ad hoc networks Research aims at achieving Deeper insights into performance limitations Identifying effects of cross-layer interactions Identifying optimal provisioning strategies Finding efficient designs

Thank you