Collaborative opportunistic network coding for persistent data stream in disruptive sensor networks

In an energy-harvesting sensor network for perpetual lifetime, the operation of sensor nodes are synchronized with the energy fluctuations, causing the network connectivity to be disruptive and unstable. The unpredictable network disruptions and challenging communication environments make the traditional communication protocols inefficient and require a new paradigm-shift in design. In this thesis, we address several issues in collaborative data collection and storage in disruptive sensor networks. Our solutions are based on erasure codes and probabilistic network coding operations. The proposed set of algorithms improve data throughput and persistency because they are inherently amenable to probabilistic nature of transmission in wireless networks. Our contributions consist of five parts. First, we propose a collaborative data delivery protocol to exploit multiple energy-synchronized paths based on a new max-flow min-variance algorithm. In consort with this data delivery protocol, a localized TDMA MAC protocol is designed to synchronize nodes' duty-cycles and mitigate media access contentions. Second, we present Opportunistic Network Erasure Coding protocol, to collaboratively collect data in dynamic disruptive networks. ONEC derives the probability distribution of coding degree in each node and enable opportunistic in-network recoding, and guarantee the recovery of original sensor data can be achieved with high probability upon receiving any sufficient amount of encoded packets. Third, we present OnCode, an opportunistic in-network data coding and delivery protocol that provides good quality of services of data delivery under the constraints of energy synchronization. It is resilient to packet loss and network disruptions, and does not require any end-to-end feedback message. Fourth, we present a network Erasure Coding with randomized Power Control (ECPC) mechanism for data persistence in disruptive sensor networks, which only requires each node to perform a single broadcast at each of its several randomly selected power levels. Thus it incurs low communication overhead. Finally, we study an integrated algorithm and protocol middleware to preserve data persistency with heterogeneous disruption probabilities across the network.