The tail‑sitter vertical takeoff / landing (VTOL) unmanned aerial vehicle (UAV) exhibits poor stability and limited compatibility with traditional landing gears. To address the aforementioned issues, a novel landing gear incorporating free-tail technology is proposed. The landing gear adopts a tandem multi-stage structure, which can ensure the length of the tail force arm in cruise condition while lowering the fuselage altitude. Furthermore, the dynamic landing process is regulated through the employment of virtual centroid force distribution techniques, streamlining the control process and facilitating seamless trajectory optimization during mode transition. Based on the single-point dataset of the cat center point, a neural network is used to train the landing gear control, which makes the landing gear adaptive takeoff and landing speed and accuracy effectively improved. Subsequently, multi-objective optimization and similarity conversion are executed in conjunction with parameter requirements of different modes of the UAV, effectively enhancing landing adaptability and stability of the tail-sitter VTOL UAV.