Research

My Master’s thesis currently in-the-works lies in the area of wireless sensor networks – small computing devices with sensors and wireless communications capabilities that are embedded into the physical world we live in. The applications are endless – pervasive computing for smart spaces and intelligent building energy management systems, structural health monitoring, wildlife habitat and forest fire monitoring, and indoor and outdoor localization and tracking, just to name a few.

Under significant resource constraints, the research challenge here is to design novel algorithms for data gathering and fusion to improve application performance, and wireless communication protocols to route the collected and aggregated data to the base station. The notion of application-level performance metrics such as tracking accuracy, considered with network-level QoS parameters such as delay and network lifetime, pose challenging trade-off issues when designing appropriate algorithms and protocols. Needless to say, the challenges in implementing them introduce further constraints and run-time adaptations to ensure reliable performance.

A broader research interest of mine lies in decision-support systems, particularly decision-making under uncertainty. For example, the state of wireless links in a network may be probabilistic depending on recent transmission success, and some nodes may be dormant in their sleep modes to conserve energy. Hence, decision making over an extended horizon to maximize expected rewards (or minimize expected costs) is more valuable.

The framework of Markov Decision Processes captures uncertainty quite suitably, and where computational and modeling complexities rule out tedious optimal computations, incremental learning and simulation-based optimization approaches, such as reinforcement learning, can greatly reduce complexity while converging to a near-optimal solution. In addition, I am interested to learn more about machine learning in general, applicable to application domains such as data and text mining for social media campaigns.