Disentangling wireless sensing from mesh networking

The resource demands of today's wireless mesh networking stacks hinder the progress of low-cost, low-power wireless sensor nodes. Optimizing wireless sensors means reducing costs, increasing lifetimes, and locating sensors close to the action. Adding mesh networking functions like IP routing and forwarding increases RAM and ROM requirements and demands substantial idle listening to forward others' traffic, all of which adds cost and increases power draw. We argue that an architectural separation between sensor and router, similar to what ZigBee and traditional IP networks advocate, would allow each node class to be better optimized to the task, matched to technology trends, and aligned with deployment patterns. Although trivial to implement on current platforms, for example by turning off router advertisements in an IPv6/6LoWPAN stack, reaping the full benefits of this approach requires evolving platform designs and revisiting the link and network layers of the stack. We examine the resulting implications on the system architecture.