Concise and flexible programming of wireless sensor networks

As computers get smaller, their use is becoming more widespread and ubiquitous. Soon computers will be all around us, interacting with us in our daily lives as networks of, or sensing their environment and communicating its findings as wireless sensor networks. These wireless sensor network devices, or sensor nodes as we call them, communicate with each other using low data rate digital radios. The devices are battery-operated and need to be small and cheap to be economically viable. Naturally, research in this field focuses on making efficient use of the scarce resources available, such as computation time and communication. The limited memory complicates implementation of many of the features common on `larger' computers. Writing applications on severely limited sensor network nodes is challenging for a number of reasons. First, finding proper methods and abstractions with which to program wireless sensor networks effectively is an ongoing process. Small size of sensor network programs is crucial, to minimize development time, duration of transport and memory use. Powerful and effective abstractions and tools may reduce the size of programs and extend their capabilities. Second, the challenge is to program devices with their desired behavior. For some applications only once, or infrequently, in other cases regular reprogramming is required. The low bandwidth communication and little memory make the transport of programs and their execution a non-trivial task. Furthermore, for wirelessly accessible and (re)programmable devices protection from loading malicious applications is also highly desirable. This dissertation motivates and describes the design and implementation of the SensorScheme platform. SensorScheme is a programming language and interpreter designed to addresses the above challenges.