Natural Gradient Routing: Sink convergence using data as guide

During the last decade, WSN routing algorithms diversity has evolved and clustered onto several categories. Although most of them still rely on the use of individual node addresses as an input for the routing algorithm, there is a renewed interest in the use of addressless routing, specially from the proposed IETF (RFC6550) standard of the RPL algorithm, which defines an incremental rank value on each node to establish the route towards the Sink. In this paper, we propose a new routing algorithm called Natural Gradient Routing (NGR), which uses sensed values in a gradient field as a rank, and a single forwarding rule to approach the Sink. With extensive simulation work we compare our proposal against two reference routing algorithms: Flooding, which represents the absence of predefined routes and AODV, a deterministic reactive shortest-path approach. We show that first, our proposal has a similar consumption, delay and overhead performance than AODV with minimal complexity, and second, the NGR configuration period is reduced to zero, thus decreasing the baseline consumption until the first packet is sent.