--- Could giant cell phone towers soon be relics?
Engineers are working on a new telecommunications model that could improve mobile service and reward those willing to forward data for others.
Device-to-device communications aim to enhance service coverage, particularly at cell edges or in black spots within a neighborhood through relaying and forwarding. In this context, each device is free to move independently, and will therefore change its links to other devices frequently due to connectivity issues. During these movements, a device may be requested to forward traffic unrelated to its own use, and therefore be a temporary router or a relay. This work analyses the forwarding dilemma from different perspective using psychological game theory. It also provides different approaches that try to solve the forwarding problem in multi-hop relay-enabled networks by means of empathy, mutual support and partial altruism.
1. Device-to-Device (D2D) Communication Networks
Device-to-Infrastructure networks --> D2D mobile users
Each device can be a temporary router and transmit data for others
Benefits: Increase connectivity and coverage in the network
2. Problem formulation: Define Data Forwarding as a Game
Two types of players: Regular node (selfish) & Relay node (altruistic)
Decision domain: Forward (F) or not forward (nF)?
Forwarding Dilemma
Only one contributor (the source node) enjoys the benefit, while all intermediary devices have to bear the energy cost. If everyone is selfish, there will be no D2D communication.
Table 1: Strategic form representation for 2 source-destination pairs (Payoff for successful end-to-end transmission:1; Energy cost: c > 0)
(nF, nF) is the unique Nash equilibrium
3. Learning Scheme: Imitative CODIPAS
We propose an imitative combined distributed payoff-and-strategy learning algorithm to find the Nash solution. The idea is:
Behaviors influence the payoffs as in Table 1
Consequences influence the behavior of the wireless node
Simulation: Convergence of Imitative CODIPAS to equilibrium payoffs (0, 0)
Convergence of Imitative CODIPAS to optimal strategies (nF, nF)
4. How to solve the Forwarding dilemma?
Add altruistic players
Consider two types of wireless nodes: regular nodes (S) and altruistic nodes (A). An altruistic wireless node will forward the data of the others whenever the channel state allows. The selfish wireless node can decide to forward or not. Now a strategy is a mapping from the type to the choice space. A wireless node does not know the type of the other node. This means that the interaction of incomplete information, where we have four different configurations.
In this case, there are four Nash equilibria:
(1-c, 1-c), (-c, 1), (1, -c), (0, 0).
Forwarding Game with incomplete information
Introduce reputation level for each user
The reputation level is used in order to help peers decide with whom to cooperate or not and to decide forwarding or not the data. The reputation level is used in order to help peers decide with whom to cooperate or not and to decide forwarding or not the data. Peers with good reputation are favored and their data are forwarding with high probability.
Design a rewarding scheme: give incentive r to each forwarding action
Data forwarding is crucial for the viability of mobile ad-hoc networks. It extends the coverage and improve connectivity by mobility and multi-hop forwarding. Thus, it is public good. However, there is critical need to incentive the users to participate in data forwarding. When a wireless device forwards successfully the data of another, he will be rewarded with a certain value r.
When r - c is positive, (F, F) is a Nash equilibrium
5. Other Types of Game: Empathy and Berge Solution
Empathy in forwarding dilemma
Deviation to the material payoff outcomes: What if participants in a one-shot forwarding dilemma game know before making their decision that the other device has already decided not to forward (defected)? From the perspective of classic game theory with material payoff, a dilemma no longer exists because of dominating strategy. It is clearly in their best interest to defect too. The empathy-based test predicts, however, that if some of them feel empathy for the other, then a forwarding dilemma remains: self-interest counsels not to forward (defection); empathy-induced behavior may counsel not. The several previous test revealed that empathy seems far more effective than most other techniques that have been proposed to increase cooperation in one-shot forwarding dilemma games. We introduce a psychological payoff that is not only self-interested but also other-regarding through the two random variables λ12 and λ21.
Berge Solution and Mutual Support The strategy profile a∗ is a Berge solution if
Berge strategy yields the best payoffs to the others players who also play Berge strategies. In forwarding dilemma game one has
Therefore, (F, F) is the unique Berge solution in this game. Berge solution provides a better outcome as (F, F), which means a mutual support between the relay nodes.
6. Conclusion
Model D2D communication as a forwarding game.
Propose a learning scheme to find the Nash solution
To solve the forwarding dilemma in multi-hop networks: 1. directly add relay nodes 2. reward users to forward data
Discussed other types of games: Empathy and Berge Solution
This research has won the Best Paper Award at the International Conference on Wireless Networks and Mobile Communications (WINCOM17) and reported by NYUAD research news.