Graph Codes for Distributed Instant Message Collection in an Arbitrary Noisy Broadcast Network

摘要: We consider the problem of minimizing number broadcasts for collecting all sensor measurements at a sink node in noisy broadcast network. Focusing first on arbitrary network topologies, we provide: 1) fundamental limits required data gathering and 2) general in-network computing strategy to achieve an upper bound within factor $\log N$ limits, where $N$ is agents Next, focusing two example networks, namely, geometric networks random Erdos-Renyi provide improved schemes that are optimal they attain i.e., lower bounds tight scaling sense. Our main techniques three distributed encoding techniques, called graph codes, which designed, respectively, above-mentioned scenarios. work, thus, extends unifies previous works such as those Gallager Karamchandani function computation special while bringing novel e.g., from error-control coding circuits, both bounds.