Efficiently Running Continuous Monitoring Applications on Mobile Devices using Sensor Hubs

Smartphone applications that continuously monitor user context are becoming popular for applications ranging from health-care to lifestyle monitoring to participatory sensing. Unfortunately, these emerging apps make poor use of the already scarce energy resources because they continually wake the CPU and other mobile device components for periodic sensing, computation, and communication. We report measurements of three representative Android smartphone apps that show 77% of the battery is wasted by constantly waking up the device components. Moreover, we find that existing power optimization techniques provide only modest benefits for this new class of apps. Using a trace-driven study of the three apps, we estimate that reducing 3G overhead, offloading to the cloud, offloading 3G traffic to WiFi, and using sensor duty cycling, can reduce only 5-10% of the energy costs. For greater energy savings, we make the case for platform support in the form a sensor hub. Sensor hubs are dedicated subsystems that interface with the sensors and radios; they use a micro-controller to support low cost sensing and computation. Our trace-driven analysis shows that sensor hubs can reduce the power consumption of continuous monitoring apps by 61%.