UNIVERSITY OF PADOVA Cooperative Transmission Techniques on Ad Hoc, Multi-Hop Wireless Networks Student: Cristiano Tapparello Master of Science in Computer Engineering Advisor: Michele Rossi
Bio Born in Vicenza on 16/05/1982 Master s Degree in Computer Engineering on October 2008 IEEE Student Member (since 2004) and Postmaster/Webmaster of IEEESB Padova (since 2005) Active member (since 2004) and treasurer (since 2005) of JUG Padova Interest: computer programming... Department of Information Engineering - Sweetnet Meeting - Signet Group - 2009 2
Summary Introduction Problem Formulation Matematical Model DP resolver FRTDP resolver Results Conclusion Department of Information Engineering - Sweetnet Meeting - Signet Group - 2009 3
Introduction In the last years ad hoc wireless networks are acquiring increasing interest Transmission inside this networks is characterized by failure probability that increase with the distance between two nodes Cooperation is a method proposed in literature to reduce the failure probability of a message delivery Cooperation can be used in different ways In this work, cooperation is considered as simultaneous multiple transmission from different terminals of the network Department of Information Engineering - Sweetnet Meeting - Signet Group - 2009 4
Problem Formulation Network with N terminals At a particular time, a source node S wants to send a message to a destination node D Two situations: S transmits and D directly receives S transmits and D doesn t receive (but probably some other nodes of the network do!) Department of Information Engineering - Sweetnet Meeting - Signet Group - 2009 5
Problem Formulation 2 If D doesn t receive: It s possible to use other nodes of the network as repeaters It s possible that more terminals transmit simultaneously to increase the probability of message propagation, hoping to reach the destination node. Department of Information Engineering - Sweetnet Meeting - Signet Group - 2009 65
Matematical Model Set of nodes on the scenario, X Finite state space, S = {s 0,1,2,, t} Transition probability, p(x,u,y) Department of Information Engineering - Sweetnet Meeting - Signet Group - 2009 7
Matematical Model 2 Cost of each transmission, c(x,u,y) The function cost determine the policy we want to achieve (energy consumption, delay, ) Set of goal state G x S if D x x G Every state t G is absorbent: p(t,u, t') = 1, u U(t),t' G Department of Information Engineering - Sweetnet Meeting - Signet Group - 2009 87
DP resolver Based on Dynamic Programming Optimality equation: Maximum number of transmission round, n J(n + 1,x n ) = 0, x n : D x n J(n + 1,x n ) = C MAX, otherwise Department of Information Engineering - Sweetnet Meeting - Signet Group - 2009 9
DP resolver 2 Initialization: Resolution: Finalization: 0 1 1 n Department of Information Engineering - Sweetnet Meeting - Signet Group - 2009 109
FRTDP resolver Asynchronous Real Time Dynamic Programming, execute repeated walks on the state space Optimality equation: J * (x) = min u U(x) y N(x) ( ) p(x,u,y) c(x,u,y) + γ J * (y) Keep an interval on the value of the optimal solution: Procedure to limit the number of states to explore and the number of actions to consider Department of Information Engineering - Sweetnet Meeting - Signet Group - 2009 11
FRTDP resolver 2 Department of Information Engineering - Sweetnet Meeting - Signet Group - 2009 12 11
Results Generation of simulation scenario: Spatial Density of the nodes: N ( d ) 2 S,D Cost: c(x,u,y) = u Department of Information Engineering - Sweetnet Meeting - Signet Group - 2009 13
Results 2 (DP) 10 nodes Varying spatial density 11 nodes, 550 m Same density but different maximum number of hop we can use to reach the destination Department of Information Engineering - Sweetnet Meeting - Signet Group - 2009 14
Results 3 (FRTDP) Same Spatial Density 550 m Different Spatial Density Department of Information Engineering - Sweetnet Meeting - Signet Group - 2009 15
Conclusion Multiple simultaneous transmission, despite augmented cost required, is chosen in an optimal transmission policy DP resolver is applicable only to limited size problems (max 14 nodes) but gives optimal solution The developed FRTDP algorithm solve efficiently problems with more than 30 nodes Department of Information Engineering - Sweetnet Meeting - Signet Group - 2009 16
Future Works Optimize FRTDP resolver to solve big networks (up to 50 nodes) Study of the optimal policy varying: channel rappresentation cost function maximum number of terminals that can transmit simultaneously Implementation in a network s protocol Department of Information Engineering - Sweetnet Meeting - Signet Group - 2009 17 16
UNIVERSITY OF PADOVA AND NOW... SOMETHING NEW More about FRTDP Resolver New results
Consideration Master s Thesis work tells cooperation is used in an optimal transmission policy but DP approach is limited FRTDP algorithm promised good performance After graduation I worked hard on the FRTDP resolver Some problems comes up... :-( but... I fixed them all and now the resolver works fine! :-) Department of Information Engineering - Sweetnet Meeting - Signet Group - 2009 19
New insight The network topology proposed on the Master s Thesis doesn t represent effectively cooperation s advantage when we increase N Lots of study have been done in this direction Moreover we study the impact of the procedure proposed to limit the number of states Changed the cost function: ( ( )) c( x,u,y) = α + ( 1 α ) u + ω y x Department of Information Engineering - Sweetnet Meeting - Signet Group - 2009 20
Network topology S x x x x D Various value of inter-column distance x Fixed distance between two nodes in a column of 2 m Probability assumes a single-link outage probability of 0.2 at 30 m (path loss + flat Rayleigh fading) Department of Information Engineering - Sweetnet Meeting - Signet Group - 2009 21
Energy and Delay VS x 70 60 c x,u,y Ce (Cd) Ce (Ce) Cd (Cd) Cd (Ce) ( ) ( ) = α + ( 1 α ) u 50 40 30 20 10 0 45 46,25 47,5 48,75 50 51,25 52,5 53,75 55 56,25 57,5 58,75 60 61,25 62,5 63,75 65 66,25 67,5 68,75 70 71,25 72,5 73,75 75 76,25 77,5 78,75 80 Department of Information Engineering - Sweetnet Meeting - Signet Group - 2009 22
Cooperation 1,0 3,5 Cooperation Mean Cooperation 0,9 0,8 3,0 0,7 0,6 2,5 Cooperation 0,5 0,4 2,0 Mean Cooperation 0,3 0,2 1,5 0,1 0 45 46,25 47,5 48,75 50 51,25 52,5 53,75 55 56,25 57,5 58,75 60 61,25 62,5 63,75 65 66,25 67,5 68,75 70 71,25 72,5 73,75 75 76,25 77,5 78,75 80 1,0 x [m] Department of Information Engineering - Sweetnet Meeting - Signet Group - 2009 23
Trade-Off Energy Delay (x= 55m) 21 19 max trans = 1 max trans = 2 max trans = 3 max trans = 4 max trans = 5 c x,u,y ( ) ( ) = α + ( 1 α ) u 17 Energy 15 13 11 9 4,5 5,0 5,5 6,0 6,5 7,0 7,5 8,0 8,5 Department of Information Engineering - Sweetnet Meeting - Signet Group - 2009 Delay 24
Impact of the procedure to limit states Δ 0 0,1 1 10 15 20 30 60 STATE FAILURE % COST % ENERGY % DELAY 1,3E+09 0 0,00 0,00 0,00 2,6E+06 2 0,00 0,08 0,05 2,6E+06 3 0,00 0,08 0,05 2,3E+06 9 1,55 0,12 0,08 1,7E+06 26 5,13 0,12 0,08 1,4E+06 81 6,78 0,07 0,04 9,5E+05 142 12,14 0,10 0,07 2,1E+05 242 50,54 23,14 6,42 Energy and Delay determined by 10 6 simulations of the transmission policy Department of Information Engineering - Sweetnet Meeting - Signet Group - 2009 25
Conclusion Cooperation it s an effective solution to the problem of sending messages from a source to a far destination The FRTDP algorithm s implementation works fine, fast and with limited resources utilization The algorithm can solve any kind of Stochastic Shortest Path problem (changing the cost function and the transition probability) Department of Information Engineering - Sweetnet Meeting - Signet Group - 2009 26