Symbiotic Sensing: Exploring and Exploiting Cooperative Sensing in Heterogeneous Sensor Networks

During the last several years we have witnessed the emergence of smartphone-based sensing applications that include activity recognition, urban sensing, social sensing, and health monitoring. In fact, most smartphones have various sensors, wireless communication interfaces, a large memory capacity, powerful processors and operating systems. These features of smartphones have intrigued researchers to develop sensing systems in which smartphones support or even replace the traditional devices in Wireless Sensor Networks (WSNs}. However, smartphones are not deliberately designed as dedicated sensing devices. Using smartphones as sensing devices opens new sensing possibilities, but also comes with new challenges, which are introduced below. When numerous applications are running on a smartphone simultaneously, they are likely to conflict when acquiring resources such as sensors, memories, battery, and bandwidth. For example, sensing applications might fail when simultaneously using sensors on Android smartphones since the sensors like microphones are exclusive, and cannot be accessed by multiple applications at the same time. Moreover, continuously sampling data would result in the batteries of smartphones being depleted quickly. In addition, smartphones are non-deterministic platforms by design, which usually add considerable uncertainties to their sensory measurements. For example, acoustic data measured by the microphones of most Android smartphones are subject to considerable delay and clock synchronization errors of up to hundreds of miliseconds. Such tolerances lead to inaccurate estimates of distances between sound sources and the microphones, which are measured based on time of arrival. Furthermore, the dynamic mobility of smartphones carried by the users introduces daunting challenges to information retrieval. These challenges include the dramatic change of the sensing contexts, the sensing locations, and the background noises. For example, audio data of a person laughing in a pub is different from that recorded on a street. These unexpected noises significantly influence the development of human-centric sensing systems. Short-range radios such as WiFi and Bluetooth offer direct communication of sending data among phones by a method such as the store-carry-forward paradigm. However, this approach is challenging because of the dynamic mobility of smartphones carried by users, of which movement patterns are hard to be predicted. To this end, we designed a symbiotic-like architecture of smartphone sensing systems using a cooperative and distributed approach through three main stacks: data sampling, data processing and data dissemination. The architecture imitates the symbiosis in nature to pave the way for multiple sensing applications to run seamlessly on a smartphone. We explore and exploit opportunities given by smartphones to overcome the above challenges. We address techniques to support building and using smartphone-based sensing systems in the context of Heterogeneous Sensor Networks (HSNs). We consider the increasing number of smartphones, the growing number of applications, the diversity of sensing capabilities of smartphones, and the dynamic mobility of the smartphones' users. The results of simulations and experiments of our proposed techniques are consistent with the theoretical analysis. For further research, we will elaborate and integrate these algorithms into a complete sensing system for smartphone-based sensing applications in smart cities.