The thesis is divided into two main parts, kinematics analysis and walking pattern generation. Comparing to the traditional wheeled robot and quadruped robot, the humanoid robot has not only the humanoid shape, the feeling system and the control system, but also the intelligent behavior like human. In addition, it can have flexible walking ability to adapt complex terrain. For the sake of achieving smooth and steady movement in real environmental conditions, it is vital to analyze kinematics and dynamics of the robot. Part one: DARwIn-OP is built for the research in kinematics analysis, dynamics analysis, imaging recognition and so on. To begin with, for the sake of analyzing the forward kinematics and inverse kinematics of the humanoid robot, each joint which is driven by servo motor MX-28T should be considered. In addition, the analysis of the both legs (12-dof) is a very difficult and useful problem. As a result of that, this thesis is concentrated on 6-dof right leg for analyzing kinematics and gait planning. In this thesis, Denavit-Hartenberg(DH) method is applied to all joint angles in right leg. In addition, analysis of the right leg is similar to analysis of the left leg. Based on the joint angles constrain, the only correct one can be obtained in different solutions on the basis of mathematics model. Part two: Firstly, discuss the traditional method to do dynamics analysis, and then put forward 3D linear inverse pendulum model assumption to take the place of the conventional method. Secondly, COM trajectory and ZMP trajectory are generated based on footprints. Thirdly, generate the ankle trajectories and correct COM trajectory by fourth order spline interpolation method. In addition, apply an optimization method into the walking trajectory to let the robot walk more stable. Last but not least, the real MATLAB simulation and program have been generated successfully.