Bimanual robotic manipulation still suffers from precise coordination control due to its strong coupling, high nonlinearity, and great uncertainty. To achieve precise and high-success-rate bimanual robotics control, we propose a Simplified Hierarchical Deep Imitation Learning method named SHDIL. This method consists of a high-level planning layer and a motion control layer and divides multimodal motion models into several primitive models. By considering the computational constraints when dealing with large data volumes and massive training, SHDIL design simplified and shallow network models while ensuring the model’s performance. To validate the proposed method’s effectiveness, we conduct the extensive simulation experiments based on Pybullet with two different manipulation tasks. The results show that our method outperforms the basic models, and reduces the parameter size by 32% compared to the state-of-the-art while slightly sacrificing the success rate.