Abstract: Evolutionary algorithms often suffer from premature convergence because of the loss of population diversity at the early stage. Coevolutionary algorithm is a hot research topic in computational intelligence, which aims at improving conventional evolutionary algorithms. Inspired by the principle of natural selection, coevolutionary algorithms are search methods in which processes of mutual adaptation occur amongst agents that interact strategically. The outcomes of interaction reveal a reward structure that guides evolution towards the discovery of increasingly adaptive behaviors. Much of the work on coevolutionary algorithms has focused on two kinds of interaction: competitive coevolutionary systems and cooperative coevolutionary systems. Competitive coevolutionary algorithms are natural models for evolving objects such as game playing programs for which it is difficult to write an external fitness function, but quite simple to define fitness in terms of competitive success against other programs in the evolving population. Cooperative coevolutionary algorithms are natural models for evolving complex objects by decomposing them into subassemblies that coevolve, and subassembly fitness is determined by how well it works with the other subassemblies in producing a complete object. The research state and advances in the coevolutionary algorithms are discussed and surveyed. The implementation techniques and main applications of the coevolutionary algorithms are outlined. Further research directions are indicated.