This invention provides a directed self-organizing network adjacent node discovery method based on probability optimum And the method is used for solving the technical problem of low efficiency in finding the adjacent nodes in the existing technology. The operating steps are that: Setting the beam selection probability vector of the nodes; Modifying the beam selection probability value on each beam directions; Modifying the times of executing the adjacent nodes selecting in all directions of the beam and the time gap length ratio between messages sending and receiving; Choosing the beam direction of executing the adjacent nodes selecting; Choosing the adjacent nodes selecting mode; Updating the adjacent nodes table and the number of found adjacent nodes according to the received messages. The method guides the next cycle of the adjacent nodes discovery according to the adjacent nodes number found in every cycle in all directions of the beam. The efficiency of the adjacent nodes discovery is improved according to optimizing the beam selection probability of executing the adjacent nodes discovery in all directions of the beam, and modifying the execution times in all directions of the beam and the time gap length ratio between messages sending and receiving.