Section: Research Program

Large-scale mobile sensing and actuation

In the past decade, the increasingly low cost of MEMS (Micro-Electro-Mechanical Systems.) devices and low-power microprocessors has led to a significant amount of research into mobile sensing and actuation. The results of this are now reaching the general public, going beyond the largely static use of sensors in scenarios such as agriculture and waste-water management, into increasingly mobile systems. These include sensor-equipped smartphones and personal wearable devices focused on the idea of a “quantified self", gathering data about a user's daily habits in order to enable them to improve their well-being. However, in spite of significant advances, the key challenges of these systems arise from largely the same attributes as those of early envisioned mobile systems, introduced in  [76] and re-iterated in  [75]: relative resource-poverty in terms of computation and communication, variable and unreliable connectivity, and limitations imposed by a finite energy source. These remain true even though modern mobile devices are significantly more powerful compared to their ancestors; the work we expect them to do has increased, and the computation and storage abilities available through fixed infrastructure such as the cloud are larger by order of magnitudes than any single mobile device. The design of algorithms and protocols to efficiently coordinate the sensing, processing, and actuation capabilities of the large number of mobile devices in future systems is a core area of MiMove's research.

Precisely, the focus of MiMove's research interests lies mostly in the systems resulting from the increased popularity of sensor-equipped smart devices that are carried by people, which has led to the promising field of mobile phone sensing or mobile crowd-sensing  [58], [55]. The paradigm is powerful, as it allows overcoming the inherent limitation of traditional sensing techniques that require the deployment of dedicated fixed sensors (e.g., see work on noise mapping using the microphones in users' telephones  [70]). Specifically, we are interested in the challenges below, noting that initial work to address them already exists, including that by team members:

• How to efficiently manage the large scale that will come to the fore when millions, even billions of devices will need to be managed and queried simultaneously (e.g.,  [81], [45])?

• How to efficiently coordinate the available devices, including resource-poor mobile devices and the more-capable cloud infrastructure (e.g.,  [68], [36], [74], [64])?

• How to guarantee dependability in a mobile computing environment (e.g.,  [34], [80], [30])?

• How to ensure that the overhead of sensing does not lead to a degraded performance for the user (e.g.,  [56], [36])?