The proposed research thesis will study coordination over networks. Specifically, it will undertake a systematic study of the various aspects and trade-offs involved in coordination among multiple agents connected by a communication network. This includes the amount of communication needed to achieve coordination, the amount and form of shared/correlated randomness available, the topology of the network used, and the level of coordination desired. As a starting point, the research will generalize Wyner's common information problem to the case where a server has access to independent sources of randomness shared with different agents, in order to minimize the rate of common message sent and study the trade-off with shared randomness.
The proposed research thesis will study coordination over networks. Specifically, it will undertake a systematic study of the various aspects and trade-offs involved in coordination among multiple agents connected by a communication network. This includes the amount of communication needed to achieve coordination, the amount and form of shared/correlated randomness available, the topology of the network used, and the level of coordination desired. As a starting point, the research will generalize Wyner's common information problem to the case where a server has access to independent sources of randomness shared with different agents, in order to minimize the rate of common message sent and study the trade-off with shared randomness.
The proposed research thesis will study coordination over networks. Specifically, it will undertake a systematic study of the various aspects and trade-offs involved in coordination among multiple agents connected by a communication network. This includes the amount of communication needed to achieve coordination, the amount and form of shared/correlated randomness available, the topology of the network used, and the level of coordination desired. As a starting point, the research will generalize Wyner's common information problem to the case where a server has access to independent sources of randomness shared with different agents, in order to minimize the rate of common message sent and study the trade-off with shared randomness.
Title: Coordination over networks Proposal: In a coordination problem, a group of agents collaborate with each other to produce a set of desired correlated actions. Such problems arise in several diverse areas such as power grids, financial markets, perimeter defense scenario, Emergency response management, Emergency in a Hospital, Manufacturing of a device etc An early work of this kind in Information theory literature is due to Wyner(1975) who characterized the minimum rate of common randomness required for two agents to sample(approximately) from the joint distribution p(x,y) of correlated random variables X,Y; this rate is known as the Wyners common information of the random variables X,Y. Some later works have focused on how coordination can be achieved by multiple agents connected by a communication network. The focus of much of this work has been on the amount of communication needed to achieve coordination. In practice, several other aspects are also of interest such as, the amount and form of shared/correlated randomness available, topology of the network used, the level of coordination desired and security. Depending on the application, the goals may differ. For instance, at a nuclear power plant, it may be important to build a robust protocols which can operate even in the presence of active adversaries in the network who may try to prevent coordination (denial-of-service) or manoeuvre the plant into an unsafe state. Whereas in a financial market network, where the actions of the players are subject to future scrutiny, it may be enough to ensure that agents trying to coordinate their actions do not inadvertently reveal their strategies through their actions and/or communication messages which can be observed by other agents. In this thesis, we propose to undertake a systematic study of various aspects and trade-offs involved in coordination over networks. Broadly, we consider various settings of problems and characterize the trade-offs with a focus on the other aspects of interest also. The first problem we plan to study is a generalisation of Wyners common information problem. As in Wyners original problem, there are two agents who want to produce samples from a desired joint distribution. A server can send a common message to both the agents. In Wyners problem, the agents and the server have access to independent(i.e., private) sources of randomness. In the variant which we study, in addition to this, the server has access to two independent sources of randomness each of which is shared with a different agent/processor. The goal is to minimise the rate of common message. It turns out that Wyners scheme cannot be directly used to make optimal use of the shared randomness. We study the trade-off between the rates of the message and the amounts of shared randomness available.