Author:

Abstract:

Mobile crowd sensing applications offer an innovative approach to realise people-centric and environmental-centric sensing applications in scenarios such as smart buildings, traffic and road condition monitoring, or e-healthcare. These applications involve large numbers of participants using heterogeneous mobile devices. These devices differ in terms of hardware (e.g. on-board sensors available), software (e.g. operating system), and external connectivity (e.g. WiFi, 3G, and online social networks). Participants may serve as opportunistic sensors and flexible actuators, while also accessing local sensing services. Nonetheless, contemporary mobile crowd sensing platforms provide poor consideration of user dynamism, which includes: mobility across networks, mobility across devices, and context-awareness. In addition, when focusing on the devices, all deployment and development efforts are closely related to the platforms supported. However, a cross-platform approach featuring flexibility (modularity, reusability and reconfigurability) is not still provided by state of the art platforms. In addition, application success relies on the number of participants, so an efficient user recruitment mechanism is also demanded. Due to their wide availability, online social networks realise a platform and device independent communication channel with participants (70% of participants replied in the first four hours since receiving a query) meanwhile increasing device reachability (by roughly 67% according to one of the performed studies). Furthermore, online social networks also provide an efficient recruitment medium due to their large user base. The contributions of this thesis are: (a) an abstraction of the user as a first-class software component (User Component), (b) a consistent communication and addressing abstraction that maps to multiple communication channels (User Binding) and (c) a modular, reconfigurable and secure runtime for web browsers with zero installed software base (ExNihilo). The User Component allows developers to reconfigure and monitor the user using regular software tools, and hence, provides a mechanism to deal with mobility between contexts. The User Binding supports selective communication with the user or with a device through multiple networks, including online social networks, and hence dealing with mobility between networks and devices. These contributions are validated through two prototypes which implement real-world applications. In the context of the SmartOffice use-case, the prototype @migo relies on the User Component and the User Binding contributions, and increases participation by 8% and availability by 24% without critically impacting battery lifetime. In the context of the SmartTeaching use-case, the prototype ExNihilo involved up to 47 regular off-the-shelf devices and improves distribution, heterogeneity, flexibility and security support within mobile crowd sensing applications. Based on these results, we believe that a middleware incorporating the User Component, the User Binding and ExNihilo enables worldwide, massive and fast release of mobile crowd sensing applications by reducing development, deployment and management efforts. In addition, the middleware provides seamless communication and coordination between humans and smart devices to support innovative applications. Finally, exploring additional mechanisms to extend middleware's applicability (e.g. quality of service aware applications) and functionality (e.g. flexible authorisation schemas) in such a dynamic environment is encouraged.