Section Editors' Introduction: WHAT IS NEXT FOR MOBILE SENSING?

Abstract

8 e now live in a world where forms of mobile sensing – for example, the monitoring of physical activity, sleep habits or commute patterns using a mixture of inertial, location and audio sensors – that until recently were firmly in the domain of research, are now rapidly becoming commonplace application features. They are well known to consumers, present in entry-level watches and phones, and are even available (in some cases) as simple API calls embedded in major mobile OSs. Given such advances in current practice, it is time for us to start thinking of what will be the next generation of key questions that will drive mobile sensing research moving forward. In this issue, we highlight three award-winning papers from ACM UbiComp 2015 – one that received a best paper award and two of which received honorable mentions at the conference. Each of these papers consider emerging topics within mobile sensing and, in this way, contribute to the search for what comes next. We begin with how additional sensory functions can be supported within already tight energy budgets by transparently integrating the use of sensor co-processors. Our next two papers study novel ways in which mobile sensor data may be used, and seek to go beyond increasingly familiar sensing tasks (such as simply counting the number of steps by a user) by automatically monitoring highly complex long-term behavioral outcomes (such as academic success) or quantifying how skilfully an activity (e.g., cooking, painting) is performed, for example. As the application horizons of mobile sensing continue to expand, this evolution will only increase the pressure that exists on the familiar resource bottleneck of mobile battery reserves. In " MobileHub: No Programmer Effort for Power Efficiency with Sensor Hub " – one of the winners of Best Paper at UbiComp 2015 – researchers from the University of Washington, Stony Brook University and Intel Research allow applications to benefit from emerging co-processor–based hardware support for low-power sensing. The key innovation is that through a combination of dynamic taint tracking and machine learning, this is achieved transparently to the developer, and requires no modification to existing application code. Although hardware support for sensing promises large leaps in sensor application energy efficiency, conventional approaches to using this hardware require the rewriting of existing sensing algorithms and developers to change the way they interact with sensors within their programs. In " SmartGPA: How Smartphones can Assess …