Looking back at Weiser's 1991 vision of ubiquitous computing, many of his predictions have been surpassed, often by several orders of magnitude: information appliances sport ever more powerful processors, store in excess of 64 GB of data onto solid state chips, and have constant high-bandwidth access to the network. Yet one prediction that has not been realized, as anybody who uses mobile computers today can attest, is the ability to use information appliances for several days before recharging them. While battery life is the cornerstone (without sufficient battery life the burden of maintaining multiple mobile devices outweighs the advantages), three emerging technologies may significantly alter the energy footprint of information appliances: bi-stable displays which only consume energy when they are refreshed; Magnetic RAM (MRAM) which combines the speed of SRAM, the density of DRAM, and the non-volatility of FLASH memory; and a new generation of powerful embedded processors that support aggressive power saving strategies, and which offer a preview of the potential of energy efficient asynchronous processors. Combined, these technologies will make it possible to create systems where power consumption is near zero while quiescent, a significant departure from the behavior of current devices. The PI's goal in this project is to investigate how the various interaction techniques we know of today might benefit a low energy architecture enabled by the aforementioned emerging technologies. The team has extensive experience with hardware design, hardware simulation, and empirical evaluation. Project outcomes will contribute to a better understanding of the parameters influencing the design of very low power interfaces, and will include: an openly available hardware test bed for evaluating interaction technique energy signatures; the first systematic evaluation of the energy footprint of command selection and navigation techniques to complement the extensive performance data already gathered; an evaluation of the potential of sensor-assisted techniques to reduce the energy consumption of information appliances; and the first evaluations of the potential of asynchronous design to enable very low power information appliances. Evaluations will be performed in the lab and through longitudinal deployments, considering a variety of tasks, to further increase the external validity of the results.
Broader Impacts: Project outcomes will establish the empirical foundations for very low energy interface designs that will help researchers and designers better understand the energy implications of various interaction techniques. They will offer researchers and practitioners the tools and toolkits they need to quickly implement and evaluate the overall energy footprint of a design, and thus will significantly lower the barrier to entry into this research area. Furthermore, they will support new curricula that focus on energy consumption. Given that information appliance use is accelerating, and since the distinction between information appliances and personal computers is blurring, this work will have a great impact on reducing the overall energy consumed for our everyday information needs.
The power efficiency of everyday information appliances such as general-purpose slates or specialized ebook readers has made striking progress in recent years. Most of this progress can be traced back to improvements in the underlying hardware such as advanced integrated power management systems inside microprocessors, the use of solid state disks, and - in some cases - the use of bi-stable displays. Bi-stable displays only consume power when refreshing pixels and are used extensively in black and white ebook-readers. Redesigning the interface can further reduce power consumption, but often comes at the cost of a less attractive interface. In this grant, we explored ways of reducing the power consumption of information appliances using a new generation of colored and interactive bi-stable displays requiring little or no change to the appearance of the interface. Key to our approach is the observation that many common reading tasks, such as turning a page or annotating it, can be performed using a slow, yet highly efficient micro-controller without having the main power-intensive processors involved (akin to playing music in the background). We demonstrated the potential of this approach by implementing it on an off-the-shelf smartphone processor, the TI OMAP 4460, featuring two high- performance ARM Cortex-A9 cores and two low power ARM Cortex-M3 cores. We found that our approach offers a similar end-user experience as traditional implementations, but significantly improves battery life for tasks such as inking (×1.7 in battery life), typing on a virtual keyboard (×2.3), and page flipping (×2.1). When combining this approach with the Space Filling Thumbnails technique (a technique in which navigation is achieved by picking the target page on an overview of all the pages in the document) we can improve battery life by a factor of up to 9.4 compared to the status-quo. Importantly, it is possible to take advantage of this approach without having to change existing applications, although – relative to creating dedicated interfaces optimized for our approach - the resulting energy savings are up to 20% lower. This approach could not only significantly improve the battery life of many of the devices we are using every day; if widely deployed, it could also considerably limit the carbon footprint of information technology with minimal impact on user experience.
