Packet Scheduling for Real-Time Surveillance in Multihop Wireless Sensor Networks With Lossy Channels

We consider the problem of real-time surveillance where remote sensors transmit data periodically to a control center through multihop transmissions. All data need to be delivered within a common deadline, which is the time that the control center makes control decisions. We propose a model that jointly considers the end-to-end delay constraints and delivery ratio requirements of flows, the need for multihop transmissions, and the unreliable nature of wireless transmissions. We develop a framework for designing feasibility-optimal policies. We then demonstrate the utility of this framework by considering two types of systems: one where sensors can transmit and receive packets simultaneously, possibly on different channels, and the other where sensors cannot. For the first type of systems, we propose an online distributed scheduling policy and prove that the policy is feasibility optimal. We also provide a heuristic for the second type of systems. We show that this heuristic is still feasibility optimal for some topologies. Simulation results show that both policies outperform state-of-the-art policies by large margins.