Generating an efficient sensor network program by partial deduction

Partial deduction is an optimisation technique developed by the logic programming community. We propose the use of Partial deduction in the domain of wireless sensor network programming where programs are written for small computational platforms and energy is typically scarce. We show how, together with a declarative programming language which has been shown to be suitable for several demanding sensor network applications, it can address key issues such as rewriting a query using views and reducing redundancy of rewritings as long as some computation and abstraction can be performed at compile-time, which obviously leads to the improvement of energy efficiency at run-time. We argue that energy efficiency can be achieved with: (1) minimised sensor network programming workload by forcing the folding of goals into the view partially; (2) reduced redundant computation with fewer computation steps at network nodes by forcing the unfolding of simple goals; (3) reduced inter-node message transmission by more specific addressing of messages to nodes; and (4) reduced memory requirements by specialising network-wide programs to smaller programs for specific nodes. A partial deduction system is developed and an extended example is provided to demonstrate the potential performance improvement of the technique.