Vinod Ramaswamy

A Mean Field Game Approach to Scheduling in Cellular Systems

We study auction-theoretic scheduling in cellular networks using the idea of mean field equilibrium (MFE). Here, agents model their opponents through a distribution over their action spaces and play the best response. The system is at an MFE if this action is itself a sample drawn from the assumed distribution. In our setting, the agents are smart phone apps that generate service requests, experience waiting costs, and bid for service from base stations. We show that if we conduct a second-price auction at each base station, there exists an MFE that would schedule the app with the longest queue at each time. The result suggests that auctions can attain the same desirable results as queue-length-based scheduling. We present results on the asymptotic convergence of a system with a finite number of agents to the mean field case, and conclude with simulation results illustrating the simplicity of computation of the MFE.

A sensor coverage game with improved efficiency guarantees

Distributed sensor coverage problem deals with designing local policies which enable a collection of mobile sensors placed in a mission space to reach a stable allocation which maximizes the coverage of the mission space. Due to stringent local information restrictions, classical approaches to distributed coverage such as Lloyds algorithm cannot guarantee the worst case performance to be greater than 1/n of the optimal where n is the number of agents. Can we improve efficiency guarantees by providing additional, but limited information about the mission space to each agent? Modeling sensor coverage problem as a strategic game, we design an agents payoff as a combination of a local utility function, which computes the marginal contribution of the agent to the system objective, and a network utility function, which computes the potential gain in the system objective that is achievable by relocating any other agent to formers local neighborhood. We show that the worst case efficiency of the designed game is at least 50%: Finally, we prove that the game is weakly acyclic and hence local agent policies such as better/best reply dynamics are guaranteed to converge to an equilibrium.

Which protocol? mutual interaction of heterogeneous congestion controllers

A large number of congestion control protocols have been proposed in the last few years, with all having the same purpose-to divide available bandwidth resources among different flows in a fair manner. Each protocol operates on the paradigm of some conception of link price (such as packet losses or packet delays) that determines source transmission rates. Recent work on network utility maximization has brought forth the idea that the fundamental price or Lagrange multiplier for a link is proportional to the queue length at that link, and that different congestion metrics (such as delays or drops) are essentially ways of interpreting such a Lagrange multiplier. We thus ask the following question: Suppose that each flow has a number of congestion control protocols to choose from, which one (or combination) should it choose? We introduce a framework wherein each flow has a utility that depends on throughput and also has a disutility that is some function of the queue lengths encountered along the route taken. Flows must choose a combination of protocols that would maximize their payoffs. We study both the socially optimal, as well as the selfish cases to determine the loss of system-wide value incurred through selfish decision making, so characterizing the “price of heterogeneity.” We also propose tolling schemes that incentivize flows to choose one of several different virtual networks catering to particular needs and show that the total system value is greater, hence making a case for the adoption of such virtual networks.

Mutual Information based communication aware path planning: A game theoretic perspective

This paper examines the problem of distributed path planning for a mobile sensor network comprised of communication-aware robots performing general information gathering missions. Mutual information is derived for distributed sensing over packet erasure channels that model multi-hop communication. We model distributed path planning as a non-cooperative game and derive utility functions that are optimized locally by each robot. Each robot computes the control input in a distributed manner that results in a combined action that can be bounded by the optimal centralized result by utilizing sub-modularity in certain cases. It is shown that when the communication model includes multi-hop communication to expand the coverage of the sensor network, the property of sub-modularity is lost. We further show that the additional global knowledge required for the local computation of utility functions can be learned by simple consensus approaches. Finally, we discuss a sampling approach to approximate the proposed utility functions in order to reduce the associated computational requirements.

Co-optimization of communication, sensing, and computation for information gathering using cloud computing

This paper describes a co-optimization method for communication-aware information gathering in multi-agent systems using cloud computing. The cloud process complements the local estimate of each agent by sending more accurate information to the agent. However, using cloud computing to improve local information gathering can lead to asynchronous delays in processing estimation and sending its results to each agent. This paper describes how information flows and channel selection can be used to manage information between the cloud and the agent, as well as among agents. Since centralized optimization cannot be performed realistically due imperfect communication, this paper uses a game-theoretic distributed optimization approach when evaluating the co-optimization method. A set of simulations show that the estimates from the distributed, cloud-based approach converge to those from a single centralized estimator.

Incentives for P2P-assisted content distribution: If you can't beat 'em, join 'em

The rapid growth of content distribution on the Internet has brought with it proportional increases in the costs of distributing content. Adding to distribution costs is the fact that digital content is easily duplicable, and hence can be shared in an illicit peer-to-peer (P2P) manner that generates no revenue for the content provider. In this paper, we study whether the content provider can recover lost revenue through a more innovative approach to distribution. In particular, we evaluate the benefits of a hybrid revenue-sharing system that combines a legitimate P2P swarm and a centralized client-server approach. We show how the revenue recovered by the content provider using a server-supported legitimate P2P swarm can exceed that of the monopolistic scheme by an order of magnitude. Our analytical results are obtained in a fluid model, and supported by stochastic simulations.

Which protocol? Mutual interaction of heterogeneous congestion controllers

A large number of congestion control protocols have been proposed in the last few years, with all having the same purpose—to divide available bandwidth resources among different flows in a fair manner. Each protocol operates on the paradigm of some conception of link price (such as packet losses or packet delays) that determines source transmission rates. Recent work on network utility maximization has brought forth idea that the fundamental price or Lagrange multiplier for a link is proportional the queue length at that link, and that different congestion metrics (such as delays or drops) are essentially ways of interpreting such a Lagrange multiplier. We thus ask the following question: Suppose that each flow has a number of congestion control protocols to choose from, which one (or combination) should it choose? We introduce a framework wherein each flow has a utility that depends on throughput, and also has a disutility that is some function of the queue lengths encountered along the route taken. Flows must choose a combination of protocols that would maximize their payoffs. We study both the socially optimal, as well as the selfish cases to determine the loss of system-wide value incurred through selfish decision making, so characterizing the “price of heterogeneity”. We also propose tolling schemes that incentivize flows to choose one of several different virtual networks catering to particular needs, and show that the total system value is greater, hence making a case for the adoption of such virtual networks.