Protocol construction using genetic search techniques

Abstract

The construction of transport protocols which offer reliable communication is a complicated task. The communicating agent must quickly adapt to changes in the network in order to maintain optimal performance. This adaptive element is an extremely difficult component to construct as the highly dynamic environment in which the protocol operates is difficult to predict. Our work attempts to automate the design process by converting it from a design problem to one of optimisation, in which genetic algorithms are used to search the space of possible protocol designs in an attempt to find the optimal solution. We present results from experiments in which we have evolved alternating bit protocols and also windowed flow control protocols, which have high channel utilisation.