System Energy Efficiency Lab
Home People Research Publications Sponsors Contacts
   
Current research
Mobile air quality sensing
Completed projects
Energy Efficient Design
Reconfigurable Computing
People
Resources
Publications
Sponsors
 

Energy Efficient Design of Heterogeneous Wireless Healthcare Sensing Systems


Wireless Healthcare System Architecture

Overview

We conducted clinical trials in collaboration with research group of UCSD school of Medicine, Center for wireless and population health systems. One of the key challenges observed during the clinical trials was extending the battery lifetime of the system while keeping the mobile devices smaller and lighter for user convenience. A longer battery life of the system is required for real time behavior monitoring of the subjects. In such systems, dynamically selecting which processing, sensing, and communication tasks should run at what point in time and on which device is difficult, but critically important if a good tradeoff among battery lifetime, latency, and high fidelity of sensed and processed data are to be achieved. Since task binding optimization has been shown to be NP complete, we use an optimal and static ILP formulation of this problem as a baseline comparison to the dynamic and heuristic algorithm we developed. Our heuristic algorithm adapts to changing conditions in the environment and user's needs by selecting and binding the tasks to the appropriate nodes in the system at run time with very low overhead.


Acti-Heart Sensor

Motivation and Goal:

We have a three tiered wireless healthcare system containing heterogeneous nodes with limited battery capacities and dynamically changing system characteristics. The assignment of tasks to different parts of the system affects the overall lifetime of the system. Our focus is to analyze the overall energy and performance tradeoffs and design policies capable of adapting at run-time to changing conditions in a wireless medium, as well as to the changing energy availability at mobile nodes. The goal of the policies is to maximize the battery lifetime while delivering the needed performance with very low overhead.

Our Solution:

Energy efficiency is an important requirement for wireless networks consisting of energy constrained sensor nodes and data aggregators that are battery operated. Furthermore, in a wireless health monitoring system, the sensor devices and data aggregator should be small in size and relatively light so that daily life and patient mobility are not constrained. We are planning to use the following approaches to achieve low power consumption at the system level while delivering the needed performance:


Real time Energy Expenditure Data
ref. Kevin Patrick, UCSD School of Medicine

Effective task binding: We decide whether to perform processing on the sensor nodes or to transmit the raw sampled data to the local aggregator. Similar decisions need to be made for the local aggregator and the back-end interactions. We compare the energy cost of processing the measured data before transmission on sensor device/local aggregator with the cost of transmitting the raw samples and appropriately partitioning the processing task in order to achieve lower power consumption. Thus, we decide when, where, and how much computation and communication should occur at what energy cost.

A context aware and cooperative duty cycling algorithm: Context can be used to choose a different duty cycle (sample more/less frequently) for a set of sensor devices. For example, in the health monitoring application, PAEE (physical activity energy expenditure), sensors can sample more frequently when the user is actively running and sample less frequently when the user is idle (sleeping). Similarly, the decision of when to send data to the backend server or the local aggregator can be context aware as well. For example, we can send data immediately in case of an emergency and we can send it in the future if the situation is not critical. In the same way, we can also think about doing cooperative duty cycling. Multiple sensor sampling intervals as well as the amount of sampled data can change depending on the energy availability and the required accuracy. We plan to integrate both context aware and cooperative duty cycling with the overall energy management strategy in order to achieve the best results.