The continuous enhancement of platforms and payloads have enabled imaging satellites to obtain greater societal benefits, while to bring challenges to imaging satellite task planning: refinement of comprehensive control, normalization of quick response, and complication of constraints. It is precisely because of the aforementioned changes and requirements, the contradiction between algorithm versatility and efficiency, between solution efficiency and accuracy are becoming increasingly acute. In order to alleviate these two pairs of contradictions, this book conducts research on imaging satellite task planning technology integrating with operations research and reinforcement learning. Preliminary research on the design of imaging satellite task planning system, bi-level optimization model, and learning-based combinatorial optimization algorithms are conducted. The effectiveness of the proposed method is verified in real-world task planning scenarios of "SuperView-1" constellation. In other combinatorial optimization problems with complex constraints, the methodology proposed in this book has enormous advantages and potential. We aspire to stimulate the interest of readers in researching related scientific issues through this book.