University of Arkansas (Jan 2016 – May 2020)
Low-latency Anomaly Detection:
- Main focus of my Ph.D. dissertation
- Developed a real-time algorithm for detecting false data injection attacks and state estimation in smart grid with dynamic models and evaluated the analytical performance of the algorithm using Markov-chain.
- Formulated a low-latency algorithm for detecing bearing faults of direct-drive wind turbines utilizing the statistical distribution of stator currents at a given frequency.
- Proposed a sequential algorithm for quick change point detection in a system with multiple post-change models under both bayesian and non-bayesian setting.
Optimized Scheduling:
- Formulated a scheduling strategy for information pushing system based on optimal stopping time theory to optimize the delay and energy efficiency.
- Designed Markov decision process (MDP) based multicast scheduling scheme in delay-constrained content-centric wireless networks while optimizing overall system cost.
- Proposed a periodic MDP-based online policy of battery charge scheduling for grid-connected photo-voltaic systems with the objective of minimizing the long-term energy cost purchased from the grid.
Mobile Edge Computing:
- Presented a Deep Reinforcement Learning- based approach for optimal dynamic computation offloading and resource allocation in multi-user mobile edge computing (MEC) systems using Deep Deterministic Policy Gradient (DDPG) algorithm.
Bangladesh University of Engineering and Technology (Mar 2013 – Jul 2014)
Image Processing and Pattern Recognition:
- Developed algorithms for human action recognition based on spatio-temporal variations of human silhouette while applying classification methods such as kNN and SVM.
- Designed schemes for lip contour extraction using morphological reconstruction based segmentation approach with k-means clustering.
