Retinal vascular puncture surgery using micro-robots faces challenges,master-slave tracking accuracy is insufficient,puncture force is difficult to control. Secondary puncture risks are caused by sudden force drops. An adaptive position-force hybrid control method was proposed. A non-singular fast terminal sliding mode controller was designed. High-precision trajectory tracking was achieved during the vascular approach phase. An adaptive impedance control strategy was introduced. A radial basis function neural network was utilized. Uncertainties from soft tissue deformation were estimated and compensated online. Precise puncture force control was enabled. A smooth switching factor was incorporated. The control mode was adjusted dynamically. A puncture stop strategy was combined to prevent secondary puncture risks. Experimental validation was conducted on simulated eyeball models and ex vivo porcine eyes. A positional error of only 2.3 μm is achieved by the robotic system. The preset curve is tracked consistently by the puncture force. The system is halted immediately upon vessel rupture. The risk of tissue damage is significantly reduced. The challenges of force-position switching are resolved effectively,a reliable technical solution is provided for retinal vascular injection surgery.