Robust algorithms for infrastructure establishment in wireless sensor networks

A wireless sensor network consists of many small, battery-powered computing devices (nodes) that can communicate with each other using radio links, and are equipped with sensors for measuring physical quantities in surrounding environment. Network infrastructure is the basic computational environment required for execution of application-level algorithms: network graph (pairs of nodes that can talk to each other), geographic locations of nodes, time synchronization etc. In many applications, network infrastructure cannot be configured prior to deployment (e.g., because of uncertainty in node placement), but has to be established autonomously by the network. We propose new algorithms for the following infrastructural tasks: (i) computing a communication graph in the presence of interference, (ii) inferring geometric and topological features of the deployment from network connectivity, and (iii) information brokerage among mobile data sources and sinks. The main goal is to relax the requirements on node capabilities, and increase robustness to variations in link quality and uncertainty in node placement. To this end, we organize computation around very simple, local and stateless elementary operations, and show improvement over existing approaches by a mix of theoretical and experimental results.