With the primary goal to give the research field of humanoid robotics a new impulse, the three technical universities in the Netherlands (Delft, Eindhoven and Twente) together with Philips have developed the humanoid robot TUlip. This robot is intended as key experimental platform for research on walking, dynamical analysis, control and artificial intelligence of humanoid robots. Using TUlip as a test bed, the work presented in this thesis focuses on the modeling and identification of humanoid robots, as well as on analysis of stability of bipedal walking. For modeling, different methods are critically reviewed: Newton-Euler, Lagrange-Euler and TMT. By making use of Denavit Hartenberg convention for modeling robot kinematics, we contribute an automatic algorithm for derivation of Lagrange-Euler equations of motions for a general humanoid robot. This algorithm is instantiated on TUlip. Furthermore, conditions for ground contact and impact expressions are derived, which are implemented in a numerical simulation together with the equations of motion. For numerical integration, the event detection and time-stepping methods are considered. Arguments in favor of the event detection method are provided and this method is used in the implementation of the model of TUlip. The model parameters of the robot TUlip are identified in dynamic experiments. To facilitate the dynamic identification a regressor form is derived using an automated algorithm from the equations of motion together with the actuator dynamics, friction and dynamics of the drive train. Persistently exciting trajectories are optimized using this regressor and these trajectories are used in experiments on the robot TUlip. These experiments provide estimates of the model parameters. The estimated parameters are validated in experiments. Finally, the most commonly used stability criteria have been critically argued. Among several candidates, two the most conventional ones, namely the zero moment point and Poincaré map have been applied for design and stability analysis of walking gaits for a model of a planar humanoid robot. The results of this analysis show that these methods are neither sufficient nor necessary to guarantee stability of bipedal walking. Consequently, further developments are needed in the research on formal design and stability analysis of walking gaits.