Abstract. The Internet of Things plays a central role in the foreseen shift of the Internet to the Future Internet, as it incarnates the drastic expansion of the Internet network with non-classical ICT devices. It will further be a major source of evolution of usage, due to the penetration in the user's life. As such, we envision that the Internet of Things will cooperate with the Internet of Services to provide users with services that are aware of their surrounding environment. The supporting service-oriented middleware shall then abstract the functionalities of Things as services as well as provide the needed interoperability and flexibility, through a loose coupling of components and composition of services. Still, core functionalities of the middleware, namely service discovery and composition, need to be revisited to meet the challenges posed by the Internet of Things. Challenges in particular relate to the ultra large scale, heterogeneity and dynamics of the Internet of Things that are far beyond the ones of today's Internet of Services. In addition, new challenges also arise, pertaining to the physical-world aspect that is central to the IoT. In this paper, we survey the major challenges posed to service-oriented middleware towards sustaining a service-based Internet of Things, together with related state of the art. We then concentrate on the specific solutions that we are investigating within the INRIA ARLES project team as part of the CHOReOS European project, discussing new approaches to overcome the challenges particular to the Internet of Things.

Challenges the Internet of Things (IoT) is facing are directly inherited from today's Internet. However, they are amplified by the anticipated large scale deployments of devices and services, information flow and involved users in the IoT. Challenges are many and we focus on addressing those related to scalability, heterogeneity of IoT components, and the highly dynamic and unknown nature of the network topology. In this paper, we give an overview of a service-oriented middleware solution that addresses those challenges using semantic technologies to provide interoperability and flexibility. We especially focus on modeling a set of ontologies that describe devices and their functionalities and thoroughly model the domain of physics. The physics domain is indeed at the core of the IoT, as it allows the approximation and estimation of functionalities usually provided by things. Those functionalities will be deployed as services on appropriate devices through our middleware.

No abstract

This article presents two lifetime models that describe two of the most common modes of operation of sensor nodes today, trigger-driven and duty-cycle driven. The models use a set of hardware parameters such as power consumption per task, state transition overheads, and communication cost to compute a node's average lifetime for a given event arrival rate. Through comparison of the two models and a case study from a real camera sensor node design we show how the models can be applied to drive architectural decisions, compute energy budgets and duty-cycles, and to preform side-by-side comparison of different platforms. Based on our models we present a MAT-LAB Wireless Sensor Node Platform Lifetime Prediction and Simulation Package (MATSNL). This demonstrates the use of the models using sample applications drawn from existing sensor node measurements.

We present a method to identify and localize people by leveraging existing CCTV camera infrastructure along with inertial sensors (accelerometer and magnetometer) within each person's mobile phones. Since a person's motion path, as observed by the camera, must match the local motion measurements from their phone, we are able to uniquely identify people with the phones' IDs by detecting the statistical dependence between the phone and camera measurements. For this, we express the problem as consisting of a twomeasurement HMM for each person, with one camera measurement and one phone measurement. Then we use a maximum a posteriori formulation to find the most likely ID assignments. Through sensor fusion, our method largely bypasses the motion correspondence problem from computer vision and is able to track people across large spatial or temporal gaps in sensing. We evaluate the system through simulations and experiments in a real camera network testbed.