We consider the problem of multiterminal source-channel communication where a number of distributed and possibly correlated sources are transmitted through an orthogonal multiple access channel to a common destination. We provide a characterization of the optimal tradeoff between the transmission cost Γ and the distortion vector D as measured against individual sources. Our approach consists of two steps: 1) a multiple-letter characterization of the rate-distortion region of the multiterminal source coding and 2) a source-channel separation theorem ensuring that all achievable pairs of (Γ, D) can be obtained by combining the rate-distortion region and the orthogonal multiple access channel capacity region. As a corollary, we determine the optimal power and distortion tradeoff in a quadratic Gaussian sensor network under orthogonal multiple access, and show that separate source and channel coding strictly outperforms the uncoded (amplify-forward) transmission, and is in fact optimal in this case. This result is in sharp contrast to the case of nonorthogonal multiple access for which separate source and channel coding is not only suboptimal but also strictly inferior to uncoded transmission.
- Gaussian sensor networks
- Multiterminal source coding
- Rate-distortion region
- Source-channel separation theorem
- Uncoded transmission