The invention belongs to the field of robots and discloses a robot control software module partitioning method. First, a system is partitioned according to minimum functions to obtain an element sample set of minimum function particles. Then, relevance analysis is conducted on sample set elements from the perspectives of functions and structures, and elements with certain relevance are merged and clustered through a fuzzy tree graph clustering method. A partitioning scheme of subtrees on different levels is obtained by selecting different thresholds. Finally, a module partitioning scheme is evaluated comprehensively according to an information entropy idea, and optimal partitioning results are selected. By combining fuzzy clustering and level analysis, relevance among system elements is quantized, effects of human factors in module partitioning are reduced, and difficulty of a modularity process of robot control software is reduced. A mathematical evaluation model is provided through the information entropy idea, a reasonable and effective solving scheme for a robot software system module partitioning particle problem is provided, a developing period of module partitioning is shortened, and system developing cost is reduced.