The invention discloses a motion optimization based self-instruction generating method for imaging satellite, a method of converting multiple meta-tasks into on-satellite manipulation-level instructions. According to the method, firstly, based on the motion information equipped to the meta-tasks and the timing sequence relationships among the meta-tasks and through the use of the greedy algorithm, the same type of motions of opening and closing the machine between two adjacent meta-tasks are maximized. Therefore, the execution complexity of software is reduced; the motion sequence of the satellites is effectively planned. And according to the motion sequence and the working state of the current satellite's equipment (for instance, sequestration uses backup to save, digital transmission servo uses the original, and etc.), the corresponding manipulation-level instructions are specified so as to effectively make the satellites more operational, achieve better efficiency and provide technical support to enable the single orbit of agile satellites to complete multiple tasks. In addition, according to the method, abstract definitions are made between the meta-tasks and their corresponding motion information as well as between the timing sequence relationships of the meta-tasks, which allows a user to define themselves. Therefore, the method is very flexible and universal and can be applied to various remote sensing satellites.