ABSTRACT
CONTENTS 13.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 378 13.2 Background and Related Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379
13.2.1 Mobile Crowdsourcing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379 13.2.2 Mobile P2P Search . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 381 13.2.3 Multi-Objective Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 382
13.3 System Model and Problem Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383 13.3.1 System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 384 13.3.2 Optimization Problem Formulation . . . . . . . . . . . . . . . . . . . . . . . . 385
13.4 The SmartOpt Framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 386 13.4.1 The Optimizer Module . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 386
13.4.3 The P2P Search Module . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 391
13.5 The Smartphone Prototype System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 391 13.5.1 The SmartLab Programming Cloud . . . . . . . . . . . . . . . . . . . . . . . . 393
13.6 SmartP2P Prototype Evaluation on SmartLab . . . . . . . . . . . . . . . . . . . . . . . 394 13.6.1 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 394 13.6.2 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 397
13.7 Conclusions and Potential Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 400 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 401
13.1 Introduction The advent of social networks and the widespread deployment of smartphone devices have brought a revolution in location-aware social-oriented applications and services on mobile phones. A smartphone social network is a structure made up of individuals carrying smartphones, which are used for sharing and collaboration [1] (i.e., content, interests, comments and places.) For example, Google Latitude, Foursquare, and Facebook Places enable users to check-in to favorite places, provide their location history, as well as numerous other functions. Smartphones can also unfold the full potential of crowdsourcing, allowing users to transparently contribute to complex and novel problem solving. A crowd of smartphone users that is constantly moving and sensing providing large amounts of opportunistic/participatory data [4, 9, 3] can offer optimality to location-aware search and similarity services [5]. There is already a proliferation of innovative applications founded on opportunistic/participatory crowdsourcing that span from assigning tasks to mobile nodes in a given region to providing information about their vicinity using their sensing capabilities (e.g., noise-maps [38]) to estimating road traffic delay [41] using WiFi beams collected by smartphones rather than invoking expensive GPS acquisition and road conditions (e.g., PotHole [17].)
Currently, the bulk of social networking services, designed for smartphone communities, rely on centralized or cloud-like architectures. In particular, in order to enable content sharing and community search over crowdsourced data, the smartphone clients upload their captured objects (e.g., images uploaded to Twitter, video traces uploaded to YouTube, etc.) to a central entity that subsequently takes care of the content organization and dissemination tasks. Although certain types of objects, such as text-based micro-blogs, will behave reasonably well under this model, significant challenges arise for captured multimedia and sensor data (e.g., data captured by the camera, microphone, WiFi RSS readings, etc.) We claim that the centralization of these object types will be severely hampered in the future due to several constraints including (i) data-disclosure: continuously disclosing user-captured objects to a central entity might compromise user privacy in very serious ways and (ii) energy consumption: smartphones have asymmetric communication mediums with a slow up-link, thus continuously transferring massive amounts of data to a query
battery faster, increase query response time, and quickly degrade the network health.
