On link estimation in dense RPL deployments

The Internet of Things vision foresees billions of devices to connect the physical world to the digital world. Sensing applications such as structural health monitoring, surveillance or smart buildings employ multi-hop wireless networks with high density to attain sufficient area coverage. Such applications need networking stacks and routing protocols that can scale with network size and density while remaining energy-efficient and lightweight. To this end, the IETF RoLL working group has designed the IPv6 Routing Protocol for Low-Power and Lossy Networks (RPL). This paper discusses the problems of link quality estimation and neighbor management policies when it comes to handling high densities. We implement and evaluate different neighbor management policies and link probing techniques in Contiki's RPL implementation. We report on our experience with a 100-node testbed with average 40-degree density. We show the sensitivity of high density routing with respect to cache sizes and routing metric initialization. Finally, we devise guidelines for design and implementation of density-scalable routing protocols.