This thesis presents an architecture of a distributed problem solving system(DPS), focusing on the control and communication aspects. Practically the DPS is implemented using three node-computers, which have the computation and communication capability on physically-separated environment. These node boards use a CPU MC68020 and a token bus controller MC68824. These node boards communicate with each other using modem. In this thesis, the DPS system decomposes and solves the given problem, and synthesizes the results. This system has learning capability by writing acquired knowledge into the meta-knowledge in the process of solving the given problem, and focus-control-mechanism, which restricts the size of the explored solution space. The system is applied to logic circuit design problem and the performance is measured. By the learning capability and focus-control-mechanism it is shown that the number of the generated tasks and communication messages among the nodes are considerably reduced."