Abstract: The sea-air-space integrated network (SASIN) and collaborative data transmission mechanism were proposed. By complementing cruising nodes with access points, such as UAVs and unmanned boats, the buoy/submarine buoy data was partially collected, processed and forwarded to the data center, thus increasing the overall data transmission efficiency and reducing the transmission delay. This paper studied the joint scheduling control mechanism of virtual machine allocation, task scheduling and computation task offloading in SASIN. In this paper, a Q-learning based computational offloading method was proposed in dynamic network conditions, so as to deal with the multi-dimension resource scheduling of the SASIN.