Exploitation of Path Diversity in Cooperative Multi-Hop Wireless Networks Dissertation Committee Department of Electrical and Computing Engineering University of Delaware Dr. Cimini Dr. Cotton Dr. Shen Dr. Morris (ECE Department) (CIS Department) (CERDEC) Candidate Chair : Jonghyun Kim : Dr. Bohacek
Introduction and challenges Aggressive path quality monitoring BSP Efficient path quality monitoring LBSP Opportunistic forwarding LBSP2, LOSP, LMOSP Conclusion and future work Outline
Introduction and challenges Mobility Modeling 2004 ~ papers Mobility Modeling 2004 ~ papers Cooperative Path Diversity 2005 ~ present 4 papers Cooperative Path Diversity 2005 ~ present 4 papers Channel Activity Analysis 2007 ~ paper Channel Activity Analysis 2007 ~ paper User Perceptual Quality Evaluation 2008 ~ paper User Perceptual Quality Evaluation 2008 ~ paper Application Traffic Identification & Modeling 2008 ~ paper Application Traffic Identification & Modeling 2008 ~ paper Research
Routing Technique Proactive (e.g., OLSR) Reactive (e.g., AODV) Introduction and challenges
: Routing control packet transmission : No transmission Proactive
Introduction and challenges : Routing control packet transmission : No transmission Reactive
Introduction and challenges : data packet from transport layer Reactive
Introduction and challenges Routing Technique Proactive (e.g., OLSR) Reactive (e.g., AODV) Single path (e.g., AODV) Multiple paths (e.g., AOMDV)
Introduction and challenges Single path B A
Introduction and challenges Multiple paths B A
Introduction and challenges Routing Technique Proactive (e.g., OLSR) Reactive (e.g., AODV) Single path (e.g., AODV) Multiple paths (e.g., AOMDV) Cooperative path diversity (BSP, LBSP, LOSP, LMOSP)
Cooperative path diversity BA Introduction and challenges
Cooperative path diversity BA One possible path Introduction and challenges
Cooperative path diversity BA Another possible path Introduction and challenges
Cooperative path diversity B Many possible paths A Introduction and challenges
Cooperative path diversity B Best path A Introduction and challenges
Cooperative path diversity Nodes are moving Link quality varies Best path varies Path quality varies Introduction and challenges
Challenges How to define the path quality based on channel conditions? How to monitor the time-varying path quality to determine the best path cooperatively?
Overview Cooperative path diversity (BSP, LBSP, LOSP, LMOSP) Aggressive path quality monitoring (BSP) Efficient path quality monitoring (LBSP) Introduction and challenges Opportunistic forwarding with path qualities (LOSP, LMOSP)
Introduction and challenges Aggressive path quality monitoring BSP Efficient path quality monitoring LBSP Opportunistic forwarding LBSP2, LOSP, LMOSP Conclusion and future work Outline
Aggressive path quality monitoring Objectives Define path quality Monitor path quality aggressively/ideally to investigate maximally possible benefits offered by path diversity routing Protocol proposed : BSP (best-select protocol)
Path quality Depends on channel conditions (e.g., channel loss, SNR, transmit power) Aggressive path quality monitoring Depends on protocol designer’s routing objectives Maximize the minimum SNR along the path (max-min SNR) Maximize delivery probability Maximize throughput Minimize end-to-end delay Minimize total power Minimize total energy
Dynamic programming Achieves routing objectives = cost-to-go from node (n,i) to destination Aggressive path quality monitoring
Dynamic programming Achieves routing objectives = cost-to-go from node (n,i) to destination Relay-set 3Relay-set 2Relay-set 1Relay-set 0 Aggressive path quality monitoring
Dynamic programming Achieves routing objectives 0,1 1,1 1,2 2,1 2,2 3,1 Relay-set 3Relay-set 2Relay-set 1Relay-set 0 src dst = cost-to-go from node (n,i) to destination Aggressive path quality monitoring
Dynamic programming Achieves routing objectives src dst 0,1 1,1 1,2 2,1 2,2 3,1 = cost-to-go from node (n,i) to destination Aggressive path quality monitoring
Dynamic programming Achieves routing objectives 0,1 1,1 1,2 2,1 2,2 3,1 = cost-to-go from node (n,i) to destination J (1,1) = 30 J (1,2) = 20 Aggressive path quality monitoring
Dynamic programming Achieves routing objectives 0,1 1,1 1,2 2,1 2,2 3,1 = cost-to-go from node (n,i) to destination J (1,1) = 30 J (1,2) = 20 J (2,1) = Aggressive path quality monitoring
Dynamic programming Achieves routing objectives 0,1 1,1 1,2 2,1 2,2 3,1 = cost-to-go from node (n,i) to destination J (1,1) = 30 J (1,2) = 20 J (2,1) = J (2,1) = Aggressive path quality monitoring
Dynamic programming Achieves routing objectives 0,1 1,1 1,2 2,1 2,2 3,1 = cost-to-go from node (n,i) to destination J (2,1) = Aggressive path quality monitoring
Dynamic programming Achieves routing objectives 0,1 1,1 1,2 2,1 2,2 3,1 = cost-to-go from node (n,i) to destination J (2,1) = Aggressive path quality monitoring
Dynamic programming Achieves routing objectives 0,1 1,1 1,2 2,1 2,2 3,1 = cost-to-go from node (n,i) to destination Aggressive path quality monitoring
Dynamic programming Achieves routing objectives = cost-to-go from node (n,i) to destination Previous step’s cost-to-go Stage information Aggressive path quality monitoring
Dynamic programming Achieves routing objectives = cost-to-go from node (n,i) to destination 0,1 1,1 1,2 2,1 2,2 3,1 J (1,1) = 30 J (1,2) = Aggressive path quality monitoring
Dynamic programming Achieves routing objectives = cost-to-go from node (n,i) to destination 0,1 1,1 1,2 2,1 2,2 3,1 J (1,1) = 30 J (1,2) = Aggressive path quality monitoring
Max-min SNR Aggressive path quality monitoring
Max delivery probability Aggressive path quality monitoring
Max delivery probability Aggressive path quality monitoring
Max delivery probability Aggressive path quality monitoring n,i n-1,I n-1 (1) n-1,I n-1 (2) n-1,I n-1 (3)
Max delivery probability Aggressive path quality monitoring n,i n-1,I n-1 (1) n-1,I n-1 (2) n-1,I n-1 (3)
Max delivery probability Aggressive path quality monitoring n-1,I n-1 (1) n-1,I n-1 (2) n-1,I n-1 (3) n,i
Max throughput Aggressive path quality monitoring
Min end-to-end delay Aggressive path quality monitoring
Min total power Aggressive path quality monitoring
Min total energy Aggressive path quality monitoring
Construction of relay-sets 0,1 1,1 1,2 2,1 2,2 3,1 AODV finds a traditional single path Aggressive path quality monitoring
Construction of relay-sets 0,1 1,1 1,2 2,1 2,2 3,1 Relay-set 3Relay-set 2Relay-set 1Relay-set 0 Aggressive path quality monitoring
Construction of relay-sets 0,1 1,1 1,2 2,1 2,2 3,1 Relay-set 3Relay-set 2Relay-set 1Relay-set 0 Aggressive path quality monitoring
Construction of relay-sets 0,1 1,1 1,2 2,1 2,2 3,1 Relay-set 3Relay-set 2Relay-set 1Relay-set 0 Aggressive path quality monitoring
Construction of relay-sets 0,1 1,1 1,2 2,1 2,2 3,1 Relay-set 3Relay-set 2Relay-set 1Relay-set 0 Aggressive path quality monitoring
Path quality monitoring 0,1 1,1 1,2 2,1 2,2 3,1 CIEREQ (channel info exchange request) CIEREP (channel info exchange reply) Relay-set 3Relay-set 2Relay-set 1Relay-set 0 Aggressive path quality monitoring
Path quality monitoring 0,1 1,1 1,2 2,1 2,2 3,1 CIEREQ : data frame Aggressive path quality monitoring
Path quality monitoring 0,1 1,1 1,2 2,1 2,2 3,1 CIEREP Path qualities between relay-set 3 and 2 are monitored Aggressive path quality monitoring
Path quality monitoring 0,1 1,1 1,2 2,1 2,2 3,1 data Aggressive path quality monitoring
Path quality monitoring 0,1 1,1 1,2 2,1 2,2 3,1 CIEREQ Aggressive path quality monitoring
Path quality monitoring 0,1 1,1 1,2 2,1 2,2 3,1 CIEREP Path qualities between relay-set 2 and 1 are monitored Aggressive path quality monitoring
Path quality monitoring 0,1 1,1 1,2 2,1 2,2 3,1 Assume that Aggressive path quality monitoring
Path quality monitoring 0,1 1,1 1,2 2,1 2,2 3,1 data Path qualities are monitored every packet transmission Aggressive path quality monitoring
Path quality monitoring 0,1 1,1 1,2 2,1 2,2 3,1 Aggressive path quality monitoring
Simulation UDelModels : Urban city, mobility, channel models Numerical analysis Ideally construct relay-sets and receive CIEREQ/CIEREP Packet level simulation QualNet network simulator CBR traffic (1024 bytes per second) Comparison between J single and J diversity J single : source’s J along the single path found initially J diversity : source’s J along the best path among all paths Aggressive path quality monitoring
Results : benefits of path diversity Aggressive path quality monitoring Sparse Dense Sparse Dense Sparse Dense Sparse Dense Sparse Dense Sparse Dense Max delivery prob.Max throughput J diversity / J single Max-min SNR J diversity — J single J diversity / J single Min powerMin energyMin delay J diversity / J single
Results : path selection differences Aggressive path quality monitoring Fraction of relays shared Minimum relay-set size max-min SNR max throughput min total powermin energy min end-to-end delay max delivery probability vs.
Introduction and challenges Aggressive path quality monitoring BSP Efficient path quality monitoring LBSP Opportunistic forwarding LBSP2, LOSP, LMOSP Conclusion and future work Outline
Efficient path quality monitoring Objectives Monitor path quality efficiently to reduce overhead J broadcast, J -test, power control Robust routing function Automatic path stretching and shrinking Protocol proposed : LBSP (local BSP)
Path quality monitoring : J broadcast 0,1 1,1 1,2 2,1 2,2 3,1 JBC ( J broadcast) Relay-set 3Relay-set 2Relay-set 1Relay-set 0 Efficient path quality monitoring
Path quality monitoring : J broadcast 0,1 1,1 1,2 2,1 2,2 3,1 Efficient path quality monitoring
Path quality monitoring : J broadcast 0,1 1,1 1,2 2,1 2,2 3,1 Efficient path quality monitoring JBC Path qualities between relay-set 1 and 0 are monitored
Path quality monitoring : J broadcast 0,1 1,1 1,2 2,1 2,2 3,1 Efficient path quality monitoring JBC Path qualities between relay-set 2 and 1 are monitored
Path quality monitoring : J broadcast 0,1 1,1 1,2 2,1 2,2 3,1 Efficient path quality monitoring JBC Path qualities between relay-set 3 and 2 are monitored
Path quality monitoring : J broadcast 0,1 1,1 1,2 2,1 2,2 3,1 Efficient path quality monitoring : best path Next hop best node
Path quality monitoring : J broadcast When this J -broadcast occurs? When current best path quality degradation is experienced. Efficient path quality monitoring 0,1 1,22,2 3,1 src dst
Path quality monitoring : J broadcast When this J -broadcast occurs? When current best path quality degradation is experienced. Efficient path quality monitoring Reference path quality for the first data frame Path quality for the subsequent data frame
Efficient path quality monitoring Path quality monitoring : J -test n-1,1 n-1,3 n-1,2 JBC n,i JBC
Efficient path quality monitoring Path quality monitoring : J -test n-1,1 n-1,3 n-1,2n,i If,
Efficient path quality monitoring Path quality monitoring : J -test n-1,1 n-1,3 n-1,2 broadcast JBC JBC relay-set ( n+1) Avoid broadcasting lower path quality than n,i
Efficient path quality monitoring Path quality monitoring : power control Efficient path quality advertisement Higher path quality Lower path quality Higher power Lower power n,1n,1 n,3n,3 n,2n,2n+1,i
Efficient path quality monitoring Path quality monitoring : power control Efficient path quality advertisement Higher path quality Lower path quality Higher power Lower power Exploit the “near-far” problem
Efficient path quality monitoring Path quality monitoring : power control
Efficient path quality monitoring Path quality monitoring : power control n,1n,1 n,3n,3 n,2n,2n+1,i 17dBm 12dBm 10dBm
Efficient path quality monitoring Path quality monitoring : power control n,1n,1 n,3n,3 n,2n,2n+1,i 17dBm 12dBm 10dBm JBC
Efficient path quality monitoring Automatic path stretching and shrinking Stretching 0,1 1,1 1,2 2,1 2,2 3,1 Relay-set 3Relay-set 2Relay-set 1Relay-set 0 A : Current active best path
Efficient path quality monitoring Automatic path stretching and shrinking Stretching 0,1 1,1 1,2 2,1 2,2 3,1 Relay-set 3Relay-set 2Relay-set 1Relay-set 0 A
Efficient path quality monitoring Automatic path stretching and shrinking Stretching 0,1 1,1 1,2 2,1 2,2 3,1 Relay-set 3Relay-set 2Relay-set 1Relay-set 0 A
Efficient path quality monitoring Automatic path stretching and shrinking Stretching 0,1 1,1 1,2 2,1 2,2 4,1 Relay-set 3Relay-set 2Relay-set 1Relay-set 0 3,1 Relay-set 4
Efficient path quality monitoring Automatic path stretching and shrinking Shrinking 0,1 1,1 1,2 2,1 2,2 3,1 Relay-set 3Relay-set 2Relay-set 1Relay-set 0 : Current active best path
Efficient path quality monitoring Automatic path stretching and shrinking Shrinking 0,1 1,1 1,2 2,1 2,2 3,1 Relay-set 3Relay-set 2Relay-set 1Relay-set 0
Efficient path quality monitoring Automatic path stretching and shrinking Shrinking 0,1 1,1 1,2 2,1 2,2 3,1 Relay-set 3Relay-set 2Relay-set 1Relay-set 0
Efficient path quality monitoring Automatic path stretching and shrinking Shrinking 0,1 1,1 1,2 2,1 2,2 2,3 Relay-set 3Relay-set 2Relay-set 1Relay-set 0
Efficient path quality monitoring Numerical analysis : setting sourcedestination 50m100m 50m relay-set 2relay-set 1relay-set 3relay-set 0
Efficient path quality monitoring Numerical analysis : results 25 nodes per relay-set 20 nodes per relay-set 15 nodes per relay-set 10 nodes per relay-set 5 nodes per relay-set Solid line Dashed line : optimal : LBSP Chips per symbol Improvement in SNR (dB) ( J diversity — J single )
Efficient path quality monitoring Numerical analysis : results Chips per symbol Improvement in SNR (dB) no power control and no J -test no power control but with J -test with power control but no J -test with power control and J -test
Efficient path quality monitoring Numerical analysis : results Chips per symbol Improvement in SNR (dB) MAX_POWER – TARGET_POWER 5 dB 7 dB 10 dB 15 dB* 20 dB
Efficient path quality monitoring Packet level simulation : setting UDelModels : Urban city, mobility, channel models Simulator : QualNet CBR traffic (1024 bytes per 50 ms for 5 min)
Efficient path quality monitoring Packet simulation : results Packet delivery ratio AODV AOMDV LBSP confidence interval # of new route searches AODV AOMDV LBSP End-to-end delay (ms) AODV AOMDV LBSP J diversity — J single AODV AOMDV
Efficient path quality monitoring Packet simulation : results Routing overhead ratio AODV AOMDV Efficiency AODV AOMDV LBSP # of new route searches Without automatic path stretching and shrinking With automatic path stretching and shrinking
Introduction and challenges Aggressive path quality monitoring BSP Efficient path quality monitoring LBSP Opportunistic forwarding LBSP2, LOSP, LMOSP Conclusion and future work Outline
Opportunistic forwarding Objectives Compare opportunistic forwarding (OF) and deterministic forwarding (DF) to see if path diversity is better exploited by OF or DF. Without node mobility With node mobility Protocol proposed LOSP (local opportunistic-select protocol) LMOSP (local monitoring-added OSP)
Opportunistic forwarding How it works? IN1 IN2T IN3 : data frame : transmitter : intended node T IN
Opportunistic forwarding How it works? IN1 IN2T IN3
Opportunistic forwarding How it works? IN1 IN2T IN3 Priority : IN1 > IN2 > IN3
Opportunistic forwarding Agreement IN1 IN2T IN3 ACK : overhearing
Opportunistic forwarding Agreement IN1 IN2T IN3 ACKACK : overhearing
Opportunistic forwarding Agreement IN1 IN2T IN3 ACK : overhearing obstacle
Opportunistic forwarding List of priority nodes pn tnT bn Preferred node Target node Backup node Priority : pn > tn > bn LPN = {pn, tn, bn}
Opportunistic forwarding List of priority nodes T pn tn bn Preferred node Backup node Target node
Opportunistic forwarding List of priority nodes Target node Make the most progress to the destination The node that achieves max-min SNR Preferred node Make better progress to the destination The node that has larger cost-to-go Backup node Make some progress to the destination The node that has smaller cost-to-go
Opportunistic forwarding Sequence of nodes Deterministic forwarding (src, tn, tn, tn, tn, dst) Opportunistic forwarding (src, tn, pn, pn, tn, dst) (src, tn, bn, pn, tn, dst) …
Opportunistic forwarding Bit-rate
Opportunistic forwarding Protocols to be compared Deterministic forwarding LBSP (local best-select protocol) Efficient path quality monitoring, automatic path stretching and shrinking, route recovery Opportunistic forwarding LOSP (local opportunistic-select protocol) One time J broadcast to construct LPN for each route failure, no route recovery Opportunistic forwarding with the path quality degradation detection LMOSP (local monitoring-added OSP) Path quality monitoring is added like LBSP, mixture of OF and DF
Opportunistic forwarding Sequence of nodes Deterministic forwarding (src, tn, tn, tn, tn, dst) Opportunistic forwarding (src, tn, pn, pn, tn, dst) (src, tn, bn, pn, tn, dst) …
Opportunistic forwarding Radio model Packet error probability nominal steep steepest shallowest shallower shallow SNR (dB) Mbps
Opportunistic forwarding Packet level simulation : setting UDelModels : Urban city, mobility, channel models Simulator : QualNet CBR traffic (512bytes per 50 ms for 5 min)
Opportunistic forwarding Results : performance of the first data packet LBSPv2LOSPLMOSP nominalsteepsteepest shallowershallow shallowest scenario number bit-rate (Mbps) (no node mobility)
Opportunistic forwarding Results : performance of the first data packet nominalsteepsteepest shallowershallow shallowest scenario number SNR (dB) LBSPv2LOSPLMOSP (no node mobility)
Opportunistic forwarding Results : performance before the first route failure (node mobility involved) nominalsteepsteepest shallowershallow shallowest scenario number LBSPv2LOSPLMOSP bit-rate (Mbps)
Opportunistic forwarding Results : performance before the first route failure (node mobility involved) nominalsteepsteepest shallowershallow shallowest scenario number LBSPv2LOSPLMOSP SNR (dB)
Opportunistic forwarding Results : performance during the connection lifetime nominalsteepsteepest shallowershallow shallowest scenario number LBSPv2LOSPLMOSP packet delivery ratio
Opportunistic forwarding Results : performance during the connection lifetime nominalsteepsteepest shallowershallow shallowest LBSPv2LOSPLMOSP Route failure rate scenario number
Opportunistic forwarding Results : performance during the connection lifetime nominalsteepsteepest shallowershallow shallowest LBSPv2LOSPLMOSP Efficiency scenario number duration that user data packets are transmitted duration that any packet including overhead is transmitted Efficiency =
Opportunistic forwarding Conclusion Without mobility (e.g., stationary mesh network) Opportunistic forwarding is preferred except for the overhead With mobility Deterministic forwarding is preferred Path diversity is better exploited by deterministic forwarding
Introduction and challenges Aggressive path quality monitoring BSP Efficient path quality monitoring LBSP Opportunistic forwarding LBSP2, LOSP, LMOSP Conclusion and future work Outline
Conclusions The significant benefits of path diversity are possible using aggressive path quality monitoring. Reducing overhead and advertising path quality efficiently are possible using the proposed novel techniques, still maintaining high benefits. Path diversity is better exploited by deterministic forwarding with node mobility. Conclusions and future work
Future work Estimate the dynamics of channel by observing ongoing channel activity. Achieve fast estimation of link/path qualities from the channel dynamic estimation. i.e., given the estimated current channel state, estimate link/path qualities. Develop models of channel evolution. Conclusions and future work
Schedule Conclusions and future work DateTask 11/11/2011 ∼ 11/20/2011 Packet level simulation 11/21/2011 ∼ 12/31/2011 Real channel measurement 12/16/2011 ∼ 12/31/2011 Develop models of channel evolution 01/01/2012 ∼ 01/15/2012 Writing up all findings 12/01/2011 ∼ 01/31/2012 Proofreading the whole thesis
Thanks