Book Description

The popularity and reduced cost of low-power multi-sensor Internet of Things (IoT) devices provide many opportunities for human-centered ubiquitous computing. These personal and environmental embedded systems continuously collect the context information about the user activity or the environment, and provide digital assistance (e.g., turn up the A/C, unlock a door). While many existing always-connected IoT solutions rely on the success of cloud computing, there is a growing interest in discovering the benefits of utilizing local resources and on-device processing. To make the IoT applications work without global connectivity, we need to overcome the limitation of the IoT devices in terms of sensor equipment and energy consumption. On the other hand, the manners of device interactions have changed with advances in wireless communication. Today nearly all mobile devices are capable of short-range wireless communication (e.g., via Bluetooth, Zigbee, etc). This provides a great opportunity for allowing the co-located IoT devices to communicate and interact free of user intrusion. In this thesis, I advocate that we utilize the pervasive and opportunistic connections around us for facilitating collaboration in the internet of things. With the help from IoT proximity networks, the costly sensing tasks can be distributed to improve sustainability. Heterogeneous devices can leverage each other’s capability in the spontaneous context neighborhoods. This dissertation details the technical solutions to the following challenges: A key enabler of collaboration in IoT proximity networks is the ability to continuously identify nearby resources. We develop a low duty cycle protocol BLEnd to let the IoT hosts automatically discover neighboring devices in range of wireless communication. To effectively coordinate context sensing tasks among the IoT devices via device-to-device (D2D) communication, we propose a generic collaborative sensing framework SCENTS which enables application-transparent collaboration for context sharing. SCENTS incorporates an active request-response model to let devices in proximity sense context information as a fleet, while balancing sensing fulfillment and the fairness of energy consumption. In complement to the request-response approach of SCENTS we architect the PINCH framework, which proactively distributes the context information in local network based on various context demand models. To further optimize the sensing task assignment in an ad hoc grouping of devices, we develop the Stacon system. It entails a distributed algorithm that quickly adjusts the neighborhood’s sensing task assignments based on the heterogeneity and dynamics of resources in proximity. To exhibit the real world implications of the proposed solutions, we deploy a IoT sensor testbed and acquire a data collection from user-carried mobile devices. In the end, we evaluate the sensing collaboration with this rich-context, real-life dataset