Reliable end-to-end data transmission in wireless sensor networks

Wireless Sensor Networks (WSNs) are used in a variety of applications today and will be used in even more applications in the future. However, WSNs tend to be less reliable than the wired networks we commonly work with mainly due to cheap hardware, low-power radios (with limited range) and limited energy supplies. An increasing number of applications especially in the context of industrial scenarios require a highly reliable network. In this thesis, the mismatch between available network technology and application demand is investigated and solution approaches are given. To cover the wide spectrum of possible applications as good as possible, two seemingly opposed types of networks are examined; both share the foundation of WSN. So-called real-time wireless sensor networks guarantee certain reliability and timeliness of transported messages and are well suited for industrial applications such as plant monitoring and closed loop control. This thesis outlines the design of a real-time networking architecture relying on WSN hardware and TDMA medium access control with fixed slot assignments. Furthermore, mechanisms to programmatically create said TDMA schedules are presented, wherein the schedules can cope with a predefined level of unreliability for the underlying wireless links. Also, mechanisms to adapt the transmission power are discussed to reduce the inter-network interference of neighbouring networks. Based on the strong foundation of the EU FP7 project GINSENG, experimental results from a testbed in a live industrial facility in Portugal in addition to more results from office environments and laboratory experiments are presented. Insight into the design of said industrial testbed including experiences when working in such an environment are outlined as well. The results show that programmatically determined TDMA schedules significantly improve reliability while reducing interference considerably. The results also show that adapting the transmission power to changes of the wireless link further reduces interference. On the contrary, delay-tolerant wireless sensor networks potentially exhibit a high delay and applications have to be able to handle those. Reliability in said networks is ensured, if network capacity allows. For this purpose, DTN technology known from PC-grade implementations is adapted to WSN networks. In this thesis, the feasibility of said approach is demonstrated and implementations for WSN nodes as well as appropriate adapters for the PC are shown to enable seamless integration between delay-tolerant wireless sensor networks and common delay-tolerant networks. On this basis, the necessary low-overhead communication paradigms are presented and evaluated. The performance of a delay-tolerant wireless sensor network as well as its network capacity is shown using an exemplary use case. The cumulative dissertation at hand shows that reliability can be achieved in real-time and in delaytolerant wireless sensor networks given a number of reasonable assumptions.