CRII: CSR: Federated Resource Management in Mobile Edge Computing

This project seeks to develop a collection of resource management techniques that quantify and improve the trade-off between QoS and cost in mobile edge computing systems for computationally expensive services. The proposed work is the first systematic study of concurrent QoS satisfaction and cost effectiveness of multi-layer mobile edge computing systems. Activities will focus on i) developing integer and mixed integer linear programming optimization and Markov decision process models of joint service placement and request scheduling problems that integrate both QoS and cost requirements in the model, ii) introducing efficient algorithms and heuristics to solve the problems given both static and dynamic users, iii) using a federated learning and deep reinforcement learning approach to find the optimal solution in dynamic environments, iv) understanding and exploiting the benefits of coding approaches, namely, random linear network coding and minimum bandwidth codes for such frameworks by evaluating different coding strategies, and v) implementing and evaluating the proposed algorithms on an a testbed built from Raspberry pis, computers and other smart devices. Mobile edge computing is a key technology that enables many of the IoT applications, thus the proposed work will have a significant impact on bringing such applications closer to reality. PI has a unique collection of research and technical experience in large scale network management, mathematical modeling and implementation, and network coding that will be used to conduct the goals of the project.