This paper introduces a novel architecture of neural networks based on a homogeneous transformation matrix for model-based reinforcement learning in robotics. It shows how the homogeneous ...