This paper addresses a data-driven locomotive strategy of an under-water vehicle-manipulator system (UVMS). Two modular designed biomimetic underwater propellers are deployed on the UVMS. Each propeller has a ribbon fin that consists of twelve rays linked by a thin latex membrane. Each ray is individually actuated. By distributing all rays in phases of undulating patterns, different form of waves can be generated from the propeller. A measurement platform is constructed to obtain the horizontal and vertical force generated by the waves of the undulating fin. The current in main circuit of the propeller is also measured. The measured data are calculated into five output results. Three types of waves are performed on the platform, including sinusoidal waves, fusiform sinusoidal waves and inward counter-propagating sinusoidal waves. Output results are varied by parameters of undulating patterns. By diversifying parameters of undulating patterns, multiple waves are carried out and their output results are obtained. With these data, a locomotive strategy for the UVMS is established. The strategy is to select the output results of the two propellers by optimization methods to meet the requirements of the UVMS on standards of driving force, stability and power consumption. With this strategy, two biomimetic propellers can perform suitable undulating patterns when the UVMS is required to implement certain locomotions.