Dependability-focused Evaluation of Sensor Networks – MURPHY

Wireless Sensor Networks (WSNs) hold great promises as an enabler of a wide spectrum of pervasive applications (e.g., urban monitoring, smart home or emergency scenario), and consequently generate huge opportunity for the National and European opportunities. With sensors, the key constraint remains resource saving. As the range of applications of wireless sensor networks extends in the fields of industrial and mission-critical, additional requisites relating to robustness and reliability appear. We are confronted to a wireless sensor network that is noteworthy for its gathering ability but that fails in ensuring its own reliance and its proper operation; such assertion is witnessed by the ever growing number of deployment campaigns that oscillate between failure and non-trivial deployments . With regards to the above requirements, our project aims at meeting the challenging task of easing the development of dependable and pervasive applications built on top of robust wireless sensor networks, thus providing a mean for detecting at an early stage possible failures and evaluating dependability measures. This endeavour is undertaken by providing:
- Enhanced fault detection capabilities grounded upon a monitoring system that provides advanced in-network event processing and correlation thus ensuring timely identification of any fault, misbehaviour or dysfunction,
- Fault injection which attempts to accelerate the occurrence of faults so as to easily judge the quality of the error handling and hence, more generally, facilitate the evaluation of dependability properties,
- Advanced code dissemination across sensor networks, which is intended to (i) enable the dynamic and distributed insertion of faults and (ii) hide to the end user the complexity related to this task.
The above components permit to detect faults, diagnose possible causes and select appropriate corrective actions, and therefore to consolidate the resiliency and dependability of sensor applications.

At the cross road of next generation of ambient intelligent wireless sensor networks, dependable systems and automatic networking, our project towards an early stage fault detection and diagnosis based on a in-network and distributed fault injection is particularly challenging mainly due to the gap between the high resilience requirements coming from pervasive applications and the fault prone nature of sensors. Thus, cross-layer, context & network-aware in-network fault detection, code distribution and abstractions will be devised, in order to ease the implementation of dependable application.
The above area of interest will be addressed by specification, prototype implementation serving as a proof-of-concept, as well as a use-cases-driven test and evaluation of the implemented technologies. Our approach is to use two representative WSN scenarios (an emergency monitoring and tracking application and a parking lot monitoring) to iteratively determine solutions for the key WSN issues, dependability, energy-efficiency, scalability and clearly understanding on real-world requirements.
For this purpose, our project is articulated around three main technical work packages including WP1: use case–driven requirements, architecture and abstraction model, WP2: fault injection-based dependability evaluation - engine and tool, WP3: use case-based testing, demonstration and validation. As a substantial advance in the state of the art, our project results constitute a necessary step to progress the field of wireless sensor network-based future and emerging technologies in France and Europe as well as an enabler for advancing the competitiveness of industry.