***Please contact me for information on new and current research projects.***
Wireless Distributed Embedded Systems
Embedded systems can be defined as application specific devices that employ a subset of computing, communication, storage, sensing and actuating components to address a specialized set of tasks. They range in their complexity from simple and inexpensive electronic toys and appliances (e.g. toasters) to complex and multi-billion dollar radar and sonar arrays. The history of application specific systems is as long as general purpose computing systems, but recently has experienced fast quantitative growth and qualitative change. For example, the addition of sensors and wireless ad hoc multi-hop low power communication capabilities created new classes of embedded systems such as wireless sensor networks and pervasive computing devices. These new classes of embedded systems open numerous profound and qualitatively new capabilities, but also pose demanding technical challenges due to their cost and power sensitivity and reliability requirements related to their in field deployment. My goal is to develop techniques for efficient design, analysis, deployment, and operation of wireless distributed systems.
- Low-Power Wireless Radio Analysis, Modeling, & Simulation
- Force-directed Models of Low Power Wireless Ad-hoc Systems with Lossy Links
- Statistical Wireless Link Generators
- A. Cerpa, J. L. Wong, M. Potkonjak, and D. Estrin, "Temporal Properties of Low-power Wireless Links: Modeling and Implications on Multi-hop Routing," in ACM International Symposium on Mobile Ad Hoc Networking and Computing, 2005, pp. 414-425. [PDF]
- A. Cerpa, J. L. Wong, L. Kuang, M. Potkonjak and Deborah Estrin, "Statistical Model of Lossy Links in Wireless Sensor Networks," in ACM/IEEE Fourth International Conference on Information Processing in Sensor Networks (IPSN'05), 2005. [PDF] A previous longer version could be found as a technical report. CENS Technical Report 0041, April 2004.
Sensor Prediction for Energy Efficiency
Modeling is a problem that inherently permeates many tasks in both sensor networks and computational sensing. In terms of complexity, capability to be rigorously defined, conceptual difficulty, and required creativity it spans a wide spectrum from modeling faults and errors in sensor readings to modeling dynamics of the instrumented environment and properties of events of interest. In my current research, the goal is to develop a systematic non-parametric data-driven approach for intersensor modeling. Intersensor modeling aims to predict a reading at a particular sensor at a particular time using one or more readings from other sensors at the same or other times and/or readings from the same sensor at other time moments. Intersensor modeling is probably the single most pervasive prediction task in sensor networks and an enabler for numerous sensor network tasks such as faulty data detection, missing data recovery, and compression.
- Non-Parametric Statistical Techniques for Modeling Sensing Phenomena
- Energy Efficient and Accurate Data Collection
- J. L. Wong, S. Megerian, and M. Potkonjak, "Inter-sensor Modeling Beyond Prediction from a Single Sample/Single Sensor: Techniques and Applications," UCLA Computer Science Technical Report #060005. [PDF]
- J. L. Wong, S. Megerian, and M. Potkonjak, "Staggered Sampling for Energy Efficient Data Collection," UCLA Computer Science Technical Report #060006. [PDF]
Statistical Techniques for Optimizing Designs in the Presence of Manufacturing Variability
Manufacturing variability (MV) of deep sub-micron technologies is in particularly important for embedded and application-based systems in communications, DSP, control and sensing. These systems are inherently subject to strict timing and power constraints and unless they are satisfied the manufacturing of IC is rendered irrelevant. In addition, embedded and application-based systems are very cost sensitive and rarely last several generation of technologies to justify extensive manual optimization to address manufacturing variability. Therefore, in the last decade there have been a flurry of research interests, both in academia and industry, to address statistical timing analysis (STA) in the presence os manufacturing variability. More recently, the scope of research was enhanced to include synthesis and in particular optimization of IC in the presence of implementation variability. Although my research goal is essentially completely aligned with other researchers in terms of the problem formulation and emphasis on practicality there is a sharp difference. To the best of my knowledge, while all other approaches assume parametric statistical models, my objective is to address the analysis and synthesis of IC in the presence of MV using data driven non-parametric statistical methods. Non-parametrical statistical methods do no impose any assumptions on the properties or functional dependencies of the underlying statistical distributions that capture properties of MV. Therefore, while data driven non-parametric statistics-based approaches for MV is not amiable to theoretical and closed formula based analysis, it exactly addresses all the features and properties of IC subject to MV and exclusively relies on data actually collected from fabricated designs.
- Non-parametric Statistics-based Approach for Statistical Timing Analysis
- A Priori Wirelength Estimation
- Statistical Timing Analysis for Variable Voltage-based Power Minimization
- J. L. Wong, V. Khandelwal, A. Srivastava and M. Potkonjak , "Statistical Timing Analysis using Kernel Smoothing," UCLA Computer Science Technical Report #060003. [PDF]
- J. L. Wong, A. Davoodi, A. Srivastava and M. Potkonjak, "A Priori Wirelength Estimation: Statistical Models, Bounds and Applications," UCLA Computer Science Technical Report #060007. [PDF]
- J. L. Wong, A. Davoodi, V. Khandelwal, A. Srivastava and M. Potkonjak , "A statistical methodology for wire-length prediction," accepted for publication in IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 2006.
- J. L. Wong, A. Davoodi, V. Khandelwal, A. Srivastava, and M. Potkonjak , "Wire-length prediction using statistical techniques," in IEEE/ACM International Conference on Computer Aided Design. IEEE Press, 2004, pp. 702-705. [PDF]