本文在未标定手眼关系及摄像机模型的情况下 ,建立了机器人立体视觉跟踪问题非线性视觉映射关系模型 ,并据此设计了基于人工神经网络的视觉跟踪控制器 .仿真结果表明该算法能完全消除稳态跟踪误差 ,具有很强的环境适应性和容错能力 ,算法简单 ,易于实时实现 . In this paper, we establish a nonlinear visual mapping model of robot stereo vision tracking problem without calibrating hand-eye relationship and camera model, and design a visual tracking controller based on artificial neural network. Simulation results show that the algorithm can completely eliminate Steady-state tracking error, with strong environmental adaptability and fault tolerance, the algorithm is simple and easy to implement in real time.