We investigate data transmission schemes from sources to sinks through a multiple-relay network. We study the achievable rate of existing coding schemes (which we term omniscient schemes, for all nodes are considered in the coding design at each node). We find that, when maximising the achievable rate, not all nodes "cooperate" with all other nodes in terms of coding and decoding. This leads us to suggesting a constrained network, whereby each node only considers a few neighboring nodes during encoding and decoding. We term this myopic coding. We calculate the achievable rate region under myopic coding and we show by examples that this region is close to that of omniscient coding, when nodes transmit at low SNR. Myopic coding has practical advantage of being more robust to topology changes. It mitigates the high computational complexity and large buffer/memory requirements of existing omniscient coding schemes. On practical system where the nodes are constrained to communicating only with neighbouring nodes, the achievable rate of myopic coding gives a tighter bound than that of omniscient coding.