MIMO-CAST: A CROSS-LAYER AD HOC MULTICAST PROTOCOL USING MIMO RADIOS Soon Y. Oh*, Mario Gerla*, Pengkai Zhao**, Babak Daneshrad** *Computer Science Dept., **Electrical Engineering Dept. University of California, Los Angeles Guangyu Pei, Jae H. Kim Boeing Phantom Works
Outline Multicast ODMRP Operation –Hidden terminal problem Multi Point Relay – partial solution of the Hidden Terminal Problem The MIMO Solution MIMO-CAST Simulation Results
S R R R R Join Query Join Reply Forwarding node Link Multicast Mesh Forwarding Mesh ODMRP Mesh Structure
ODMRP Multicast Mesh On-demand Mesh Creation –A source initiates Join Query flooding only when it has data to send –Multicast group members send Join Reply messages following backward pointers to the source –Upon receiving a Join Reply, a node set itself as Forwarding group Data Forwarding –Forwarding group nodes relay data packets Route Maintenance –Periodic Join Query/Join Reply exchanging
A C S B Hidden Terminals in ODMRP Source S broadcasts to A and B In turn A and B re-broadcast A and B hidden Node C experiences a collision
Multi Points Relay to alleviate the Hidden Terminal Problem MPR is a feature of the OLSR routing protocol - used in Link State update dissemination Each node chooses a set of nodes (MPR Selectors) in the neighborhood, which will retransmit its packets. MPR alleviates, but does not eliminate hidden terminals
24 retransmissions to diffuse a message up to 3 hops Retransmission node 11 retransmission to diffuse a message up to 3 hops Retransmission node
Enter the MIMO Solution MIMO benefits A B C D Interference range of A Interference range of B Reduced interference range of A A B C D E F –Beamforming Achieving space reuse –Multiple streams Increasing point-to-point capacity We exploit the Beamforming feature of MIMO
MIMO Channel: Linear System where n + + +
Parallel Decomposition of the MIMO Channel Singular Vector Decomposition U and V are Unitary matirces –UU* = I m VV* = I n Λ is a diagonal matrix of singular values of H MIMO channel is transformed r (≤min(m,n)) parallel non-interfering single channels V*VUU* Pre-processingPost-processing Channel x1x1 xnxn y1y1 ymym
MIMO Beamforming r = w R H Hw T s where w T = [w T1 w T2 w T3 ] T : tx weights, w R = [w R1 w R2 w R3 ] T : rx weights, H: 3x3 channel matrix, w T1 w T2 w T3 w R1 w R2 w R3 H Transmitter Receiver s r w R1 w R2 w R3 h T = Hw T Transmitter Receiver s r = w R H h T s where h T = Hw T : equivalent to 3x1 vector channel Estimate h T on reception of Channel Learning Preamble
Nullifying On reception of Channel Learning Preamble from A, Node D learns h A = Hw A r = w D H h A s To null A, D finds w D such that w D H h A = 0 n-1 interferers are nullable with n antennas AB D C
MIMO MAC Protocol MIMO-CAST uses Beamforming Technique Channel Learning –Before transmitting a data packet, the upstream node sends a CL (Channel Learning) Preamble including weight vector –Upon receiving CL Preamble, the intended receivers adjusts their weight vectors (to optimize reception) Blocking Interference –If a node receives a non-intended CL Preamble, it recalculates own weight vector to null the signal
Simulation Settings Simulation Environments –Qualnet –200 nodes in 1000m x 1000m –1500 bytes/packet –802.11b 15dBm transmission power –2Mbps channel capacity and 370m radio range Metrics –Packet Deliver Ratio : the fraction of packets received averaged over all receivers –Throughput : total received byte of data packet divided by the total simulation time –Average End-to-End delay : the averaged time taken for a packet to be transmitted across the network form a source to a receiver Compare MIMO-CAST with MPR-Multicast with SISO and ODMRP
Hidden Terminal Scenario Source Forwarder Receiver Build a topology that emphasizes the hidden terminal problem Node 1,2,3, and 4 receive duplicated packets
Packet Deliver Ratio in Hidden Terminal Scenario Packet Deliver Ratio –the fraction of packets received averaged over all receivers
Normalized Overhead in Hidden Terminal Scenario Normalized Packet Overhead –total number of packet transmissions divided by the total number of data received
Throughput in Variable Traffic Scenario Data sending rate increases from 10packets/s to 200packets/s 1 source, 100 members among 200 nodes
Throughput vs Increasing Membership Number of receiver increases from 20 members to 100 members 1 source and 200 packets/s with 1500 byes/packet
Average End-to-End Delay vs Increasing Membership Number of receiver increases from 20 members to 100 members 1 source and 200 packets/s with 1500 byes/packet
Conclusions MIMO-CAST is a cross-layer multicast protocol –Selective reception feature of the MIMO system is used to improve multicast (at the network layer) Significantly reduces duplicated retransmission Reduces delays Simulation results confirm MIMO-CAST performs far better than conventional multicast protocols with IEEE and a SISO system Future work: –Use MIMO to combine (in phase) instead of nullifying competing broadcasts (cooperative radio approach) –Testbed Implementation This will allow verification of the simplifying assumptions used in simulation