This project was conducted in the the frame of the course MICRO-507 Legged Robots at EPFL. The aim is to design, implement and analyse the walking control of a simple biped. The considered biped is 2-dimensional and made out of three links (two legs and a torso), each of them having its own punctual mass (see Figure 1.1). The biped, its kinematics, its dynamics and its control are modeled, simulated and animated in a Matlab environment. In this project, three different controllers are implemented: two classical control methods (PD-controller and Virtual Force Control) and one learning method (Deep Reinforcement Learning). In a first time, this report explains how the kinematics and dynamics of the biped are designed and how the controller are implemented and tuned so that the biped can walk. Then, in a second time, the three controller’s performance are analysed regarding different criteria (speed, robustness against perturbations, cost of transport, etc.). Eventually a synthetic comparison of the three methods is made.