Event
Ph.D. Research Proposal Exam: Subhankar Banerjee
Thursday, January 16, 2025
1:00 p.m.-2:00 p.m.
https://umd.zoom.us/j/4793002414
Maria Hoo
301 405 3681
mch@umd.edu
ANNOUNCEMENT: Ph.D. Research Proposal Exam
Name: Subhankar Banerjee
Committee:
Professor Sennur Ulukus (Chair)
Professor Richard J. La
Professor Kaiqing Zhang
Date/time: Thursday, January 16, 2025 at 1:00-2:00 pm
Location: https://umd.zoom.us/j/4793002414
Title: Optimal Control in Status Update Systems
Abstract: We study information freshness for different network models. In our first completed work, we consider an adversarial communication network, where a base station aims to sample fresh update packets from a source and transmit them to a monitor, such that the age of the monitor gets minimized. However, there is a power-constrained adversary present in the system which jams the communication channel between the monitor and the base station. We formulate this problem in a game theoretical setting and find various equilibrium points. In our second completed work, we consider a similar model as our first completed work with a power-constrained base station. We again formulate the problem in a game theoretical setting and find different equilibrium points. In our third completed work, we consider an age minimization problem with a transmitter and a receiver, where the update packets arrive at the transmitter in a stochastic manner. The transmitter has a storage option to store the incoming packets for later transmission. There is a storage cost associated with the storing process. Thus, there is a trade-off between the age of information and the storage cost. We find an optimal policy that minimizes the storage cost plus the average age of information. In our fourth completed work, we consider an age-minimization problem where the transmitter employs a hybrid automatic repeat request protocol. We formulate this problem as a Markov decision process and find several structural properties of an optimal transmission policy. In our fifth completed work, we consider an age-minimization problem in a preemptive server. The optimal preemption policy depends on the service time statistics. We provide a necessary and sufficient condition on the service time statistics, such that the always preemption policy is an optimal policy. We also propose a double threshold-based policy, which is optimal for several service time statistics. With numerical results, we show that the performance of the proposed double threshold policy is very close to an optimal policy for arbitrary service time statistics. In our sixth completed work, we consider an age minimization problem where the transmitter is equipped with an energy harvester and employs a rateless code to combat channel erasure. We formulate this problem as a Markov decision process and propose a low-complexity optimal transmission policy. In our seventh completed work, we consider the age minimization problem in a multi-source network. We formulate this problem as a Markov decision process and find several structures of an optimal policy. We also propose a low-complexity policy by leveraging the structures of an optimal policy.
Next, we present three proposed works. In our first proposed work, we consider a budget-constrained age minimization problem. We formulate this problem as a constrained Markov decision process. We aim to find a non-iterative low computationally complex policy to solve the constrained Markov decision process. In our second proposed work, we consider a remote Markov source tracking problem with an energy-harvester transmitter. We employ the metric age of incorrect information to quantify the tracking error. We aim to find an optimal sampling policy that should be employed by the transmitter to minimize the age of incorrect information. For many problems, we cannot use the Markov decision process approach due to the curse of dimensionality. For these problems, index policies are known to perform well. However, proving the indexability of a problem is challenging. We address this issue in our third proposed work, where we aim to develop an easy-to-verify condition for the Whittle indexability.
