Research
Research Assistant Stony
Brook University (2009 to present)
Advisor Prof. Alex Doboli [adoboli]
- Mixed-Signal Systems Design Automation
The demand for low-cost, high-quality mixed-signal systems has rapidly increased due to expanding applications in mobile communications, intelligent navigational systems, and smart grids. This trend requires productivity gains of up to 1000X which current Computer-aided Design environments cannot achieve. A critical research problem in this context is to find efficient solutions for analog specification, synthesis, and layout tools. While analog blocks will not become larger, modern mixed-signal systems will use more of these circuits. This pushes the need for precise, circuit specific models that can bridge the gap between simulator-in-the-loop approaches and analytical model-based methods. The shift of analog design to higher levels of abstraction is vital and opportunities exist in simultaneously optimizing the tightly-coupled circuit and layout, trade-off management, and template-based design with constraints extraction.
My research in this area is geared towards the development of systematic comparison models of analog integrated circuits. The work focuses on various design aspects, including topological and symbolic expression matching, performance constraint extraction, and trade-off characterization. These techniques are important as they allow systematic characterization and mining of the design space in search of efficient design patterns, templates, and directions that can produce high quality solutions. They are also relevant in terms of synthesis, enabling Computer-aided Design environments to use designer-like insight for challenging problems like topology selection, design reuse, and physical floor-planning based on identified similarities and differences between verified solutions. I am currently expanding these methods for layout-aware circuit classification schemes.
- Embedded Sensing Networks
Coupled with emerging applications, such as Cyber-Physical Systems, this field is shaping to offer challenging and important research problems. The trend is to integrate highly complex systems that combine physical sensing, computational, security, and communication capabilities. Novel reconfigurable system-on-chip architectures together with hardware-software co-design research is needed to ensure cost-effective development of such systems and to tackle decision making and data acquisition issues. Data mining techniques offer potential benefits by expanding current methods beyond optimization of the communication bandwidth.
The research I conduct in this field is focused on developing distributed data aggregation methods for networks of embedded sensors. Addressed topics include optimizing the aggregation scheme for sensed data from geographically distributed cyber-physical systems and efficiently mapping virtual communication paths onto partially-known networks. Using reconfigurable system-on-chip devices, my approach satisfies resource, timing, and precision constraints. The resulting network data models offer robust decision making strategies.