Approximate Semantics for Wirelessly Networked Applications

Approximation saves energy in computation and storage by allowing data to diverge from precise representations. In contrast, networks guarantee that transmissions are received precisely by enforcing integrity checks at multiple layers of the protocol stack; these checks result in onerous retransmissions and increased network contention especially in wireless networks. When energy is limited, such as on embedded systems, conservative network semantics can needlessly consume an application’s energy budget. Why should approximate data be transmitted precisely?