Aiming at the adaptation problem of PI (Physical Internet) containers to goods, considering the standardization, modularization and extensibility of PI containers as well as the constraints of the volume, direction and fully support of the cargo packing, the modular reorganization of PI containers was carried out to build a combined PI container adapted to the cargo for the goal of maximizing containers space utilization. By combining the characteristics of PI container packing combination to adapt packing problem, a combination adaptive packing algorithm was designed, which including three steps: cargo classification, cargo packing and box combination. In the steps of cargo packing, the order and position of cargo packing were solved by using the improved meme algorithm introduced multi-population mutation strategy to improve the early search quality of the algorithm, and path reconnection technology and disturbance operation were introduced to prevent the algorithm from falling into a local optimum. It is solved in two experimental scenarios: more batch and less goods and less batch and more goods, and the effectiveness of the improved meme algorithm was verified by algorithm comparison.