The problem to select the best next view of the object at each modelling step is posed as an optimization problem where the robot entire posture needs to be defined at the same time as the robot cameras’ position and orientation. To achieve this, a differentiable formula has been designed to express the amount of unknown data visible from a specific viewpoint given only knowledge acquired in previous steps. A specific stability constraint is also designed to allow the robot to reach a pose where its feet can be moved away from their initial position.