This paper describes a kind of parallel genetic algorithm that is based on the idea of multi-agent cooperation.The algorithm consists of several computing units

in each of which a simple genetic algorithm is maintained

thus each computing unit can be regarded as an independent autonomous agent.The algorithm allocates computing resources to each unit according to the resource-allocating vector and carries through exchange of individuals between units according to the individual-migrating matrix.From the viewpoint of multi-agent system

the allocation of computing resource represents the coordination between agents

while the migration of individuals represents the collaboration between them.The algorithm can be implemented easily in a serial computer and it has the quasi-parallel feature in this case.The analyses to such a quasi-parallel genetic algorithm in two dimension show that since the cooperation between computing agents is taken into account in the algorithm

the computing resources can be utilized in a more effective way and thus better performances are presented when the algorithm deals with different kinds of optimizing problems.