Research‎ > ‎

Utility-Based Communication Optimization

Faculty: Greg Pottie
Postdocs: Katarzyna Wac
Students: Jay Chien


The FieldStream architecture includes interactions of multiple types, including autonomous operations within the Body Area Wireless Sensor Network (BAWSN) systems components worn by a subject, and interactions between study designers and subjects which affect system goals and resources.  The communications system consists of wireless links between BAWSN sensors and a cell phone, and a public infrastructure for wide-area connection of the cell phone (acting as a gateway) to the back-end server, making use of WiFi whenever available. Research issues emerge due to battery lifetime constraints on of all the devices, and because of the unreliable connection of sensors to the cell phone (e.g., cell phone is off or out of communication range, or a subject forgets to carry his/her cell phone). In the latter case sensors have to rely on own computing and storage resources.

Even when the cell phone is reachable from the sensors (enhancing the amount of processing and storage resources now available), connection to the back end server may still be absent (e.g., cell phone out of coverage area of the cellular towers, in airplane, etc.). These variations in the connection status and processing/storage resources available, in turn, require dynamic decision making (i.e., in-network processing) about the sampling, storage, processing, and transport policies, given the current connectivity status, resources available at the moment  (especially the devices’ battery lifetime), and quality of service requirements.

A central component of the research is thus the cross-layer optimization from BAWSN sensors sampling through to data transport and the application level adaptation techniques. The utility-based optimization research focuses on how the overall system should be modeled, and how the data processing techniques in FieldStream (model-driven sampling, compressive sensing and inference) should be optimized in conjunction with the transport and application layer adaptation techniques and that in order to ensure maximum system energy efficiency.


Data processing at the BAWSN and back-end nodes will reflect awareness of the available communications resources, in terms of their monetary cost, energy efficiency, bandwidth, and delay. An intrinsic characteristic of FieldStream is the variability of these resources, and consequently we must determine the network state and apply the appropriate data collection, processing, inference and communication strategy. Periods of BAWSN’s good   connectivity may be exploited for delay-tolerant bulk transfer of raw data, wherein raw data that is stored on local flash is transferred efficiently to the cell phone and from there to the back-end server on a wide-area network. In yet other circumstances, we may only have the communication resources to send high-priority summary information to enable real-time feedback from a back-end server, while waiting for better communication circumstances to send the raw data.

When the cell phone is not reachable from the sensors, a sensor should be able to autonomously process data and control its sampling without relying on the centralized control. Such autonomous control is also important to deal with cases where the end-to-end delay is significant due to network delay either in the BAWSN or in a wide-area network. Thus the nodes must be aware of the network state (requiring probe signals) and adjust their processing, storage, and communication strategies to best meet the quality of service requirements and especially the devices’ battery lifetime requirements.

FieldStream uses a variety of models at its core including models of communication connectivity for different classes of subjects, personalized models for model-driven sensing, psychological models for behavior events, and sub-Nyquist sampling techniques such as compressive sensing for physiological data capturing by a BAWSN. There are energy optimization issues (e.g. to meet battery lifetime requirements), and also interactions with the transport and application layers. Our work will also address how these model-driven techniques can be exploited to maximize the utility of real-time data transferred from BAWSN sensors to a central back-end server given varying network conditions. These utility functions will be constructed in collaboration with study organizers so as to design an overall optimization of feasible complexity that reflects the value put on data timeliness and resolution for recognition of different types of subject behavior. Real time data presents particular challenges, given the possibility of intermittent connectivity and a limited system battery lifetime.

Latest Work

We have formulated the overall optimization problem, and will be progressively refining models for greater realism, as data is gathered from the experimental program.  A report on progress will be posted at the end of the first quarter of 2010. Please return to this page for updates on our progress.

Related Work

[1] S. Boyd and L. Vandenberghe, Convex Optimization. Cambridge University Press: 2004.

[2] T. Hastie, R. Tibshirani, and J. Friedman, The Elements of Statistical learning. Springer: 2001.

[3] Ni, K. and Pottie, G., “Bayesian selection of non-faulty sensors,” IEEE ISIT, (2007).

[4] K. Ni, N. Ramanathan, M.N. Hajj Chehade, L. Balzano, S. Nair, S. Zahedi, G. Pottie, M. Hansen, and M. Srivastava, "Sensor Network Data Fault Types,"  ACM Transactions on Sensor Networks, 2009.

[5] K. Wac, M. Bargh, B.J. van Beijnum, R. Bults, P. Pawar, A. Peddemors,  “Power- and Delay-Awareness of Health Telemonitoring Services: the MobiHealth System Case Study”, IEEE JSAC, Special Issue on Wireless and Pervasive Communications in Healthcare, 27(4): 525-536, IEEE Press, 2009.

[6] Chalmers, D. and M. Sloman, "A survey of Quality of Service in mobile computing environments." IEEE Communications Surveys and Tutorials, vol. 2, no. 2, 1999.

[7] Istepanian, R.S.H.; Philip, N.Y.; Martini, M.G., "Medical QoS provision based on reinforcement learning in ultrasound streaming over 3.5G wireless systems," IEEE JSAC, vol. 27, no.4, pp. 566-574, May 2009

[8] Kakousis, K., Paspallis, N., and Papadopoulos, G. A. 2008. Optimizing the Utility Function-Based Self-adaptive Behavior of Context-Aware Systems Using User Feedback. OTM 2008. R. Meersman and Z. Tari, Eds. LNCS, vol. 5331. Springer-Verlag, Berlin, Heidelberg, 657-674, 2008.