Abstract: The objective of this research is to implement an autonomous learning approach to robotic hand-eye coordination ability, so as to bring higher adaptive ability to robots in the practical environment. The approach is inspired by human infant's developmental procedure, a brain-like computational structure is constructed to simulate human brain cortices of controlling hand-eye coordination; and then, a behavioral pattern is adopted from infant development when forming hand-eye coordination. The combination of the computational structure and the behavioral pattern is applied to building a novel robotic hand-eye coordination learning algorithm. This work is supported by experimental evaluation, which shows that this approach is able to drive the robot to learn hand-eye coordination successfully; the robot also shows staged behavior change, which is similar to the features of human infant development; in addition, the robot exhibits fast learning speed.