Hey, why should I forward any packets to you? 
Kavé Salamatian 
Université de Savoie
We have to cooperate but
we don't want to waste resources/banwidth 
we are rational; ready to give more if we get more!
Everybody is selfish but nobody is malicious
nreliable nodes: cannot count on their permanent presence ; might be opportunistically off
So what is the problem ?
Cooperation ???
Full cooperation 
Do the best possible behavior to achieve a performance goal 
Is the goal achievable ?
How to achieve the goal ?
Non–cooperativeSelfish behavior 
Different rational goal 
How to mitigate conflicting rational
Non-cooperative
Malicious behavior : Harmful goal How to contain irrational objectives 
Cooperation assumes rationality Social intelligence  Machine learning 
Each Node implements a forwarding function The forwarding function implement the cooperation
What about selfishness ?
classical framework
General framework
Let’s define for each packet a set of attributes Ai
Destination address D(Pi)
Some Attributes are extracted from packet
some are coming from local context  
Utility ?
Let’s define a utility function U(Ai, D(Pi), ID, A)
The utility of forwarding message i directed  to D(Pi) to node ID with context A 
The utility function capture the selfishness of the node
Generic forwarding scheme
Calculate for each packet in buffer its utility
Forward the largest utility 
Classical routing 
 Assign the utility function 1 if the node ID is on the path to destination D(Pi) null otherwiseCommunity or content networking
Give a higher utility to some specific contents or community.Cooperative monitoring
Give higher utility to information that give better view of network stateCooperative learning
Give higher utility to information that will help in learning of neighbors. 
Our utility 
forward the data that will result in the lowest estimation error
Lowest MSE
crash courses
Each node maintains a state vector 
How to share useful information about states with interested nodesIn the simplest setting all the node are interested about all states Nodes should be able to define which variables they are interested in
Linear estimation crash course
Compression crash course
projection phase
Quantization phase
Local compression
Each node calculate locally its covariance
apply a KLT
apply the resulting projection do quantization forward to node c the compressed state vector The state vector is reconstructed at node c
Distributed compression
How to do ?
Iterative approach 
General case 
Reception processing
Covariance estimation 
Preference List
Forward variables that are correlated with preference list of neighborsincentive/punishment mechanismThe node implements the distributed  compression.Apply the optimal projectionForward the projections when they change
Forwarding scheme
incentive or punishment
The node adds estimated variables from its own preference list  to the forwarded variablesThis adds its estimation error as a noise in the neighbors estimation processNeighbors have an incentive to help it to reduce its estimation errorsCooperation by punishment rather than by incentive
Let's be bad ! 
The proposed punishment mechanism acts as a shadow priceIf mechanism for enforcing shadow price exists we can implements a Pareto-Optimal cooperation mechanismBeing social becomes helpful How to enforce shadow prices ?By entangling the performance of or neighbors estimation with our own estimation Linear projections provide an elegant solution Cooperation by punishment not by incentives
enforcing cooperation
performances 
being social is helpful 
We define a cooperative framework for networks Illustrated with distributed state sharingApplicable to a large set of scenarios
DTN, distributed compression/transmissionThe essence of distributed setting is social intelligenceNode selfishness is essential 
Conclusions
Thanks
this work have been developed in the context of the EU FP7 Ecode project
www.ecode-project.eu