Evolving Gene Regulatory Networks to Control Wireless Sensor Networks

Andrew Markham ( OUCL )

Low level programming of wireless sensor networks is a difficult problem, as
it is not clear how to relate high-level goals (e.g. to monitor temperature
over an area with a certain fidelity) to actions to be undertaken in a
distributed fashion (e.g. rules for specifying when to communicate to
neighbours).  This has limited the adoption of wireless sensor networks, as
high level users of sensor networks have to turn to embedded experts in
order to reprogram a network to accomplish a particular sensing task.  To
tackle this issue, we turn to biology for inspiration, and note that cells
all run the same "code", namely DNA, yet perform markedly different actions
depending on protein concentrations in their local neighbourhood.
Abstracting this, we examine how we can evolve code that can be run on nodes
in the network to accomplish a certain high level task, without resorting to
low level programming.  We present some preliminary results demonstrating
how code has been automatically generated that meets a defined sensing goal,
subject to constraints on energy and other resources.  Future research
directions will also be discussed.