Due to the continuous development in hardware-, radio-, and sensor technologies, and
the efforts of standardization organizations, the Internet of Things is not just a
vision anymore, but it slowly becomes a part of our everyday life. The number of deployed
sensors and actuators in our environment is increasing day-by-day transforming the
physical world into an intelligent environment enabling context-aware services. To
fully support this transformation we need to adapt the basic principles of communication.
We do not want to know the IP addresses of individual sensor for example, we would
rather like to query them based on their context. Also, we are often interested in
the information itself, no matter which device provides it. In this paper we extend
our formerly proposed addressing scheme for RPL networks (CAEsAR) to make it even
more efficient. CAEsARv2 uses RPL trees and aggregates context information in Bloom-filters
(BF) or bit vectors along the tree. With this addressing scheme the RPL protocol itself
is enhanced to support context-based multicast, service-discovery and datacentric
communication. Compared to our original proposal, in CAEsARv2 we get shorter update
messages, as a result of assigning distinct data structures (Bloom filters or bit
vectors) to each of the context parameters. We also show that by storing IP addresses
also in Bloom filters, similarly to other context parameters, routing entries become
shorter and evenly distributed among the nodes. Through simulations we demonstrate
that the efficiency of Bloom-filter and bit vector aggregation in CAEsARv2 is not
affected significantly by the radio ranges of the nodes in the network. Finally, through
experimental results we show that, in case of correlation between geographical proximity
and measured values, CAEsARv2 can adapt more efficiently to context changes than the
centralized publish/subscribe messaging systems.
