Abstract : To exploit the assets of robotics systems, missions are usually expressed through simultaneous goals to reach (objectives) while satisfying imperative conditions (constraints). These operational inputs being usually considered as unpredictable, reactive control laws algorithms are used. However, they almost always focus on the resolution of the control problem, assumed to be feasible; the simple case of a joint submitted to both deceleration and position limits shows that constraints incompatibilities frequently occur. To overcome such issues, a methodology is proposed to analyze and establish safety at the control level and case studies involving common robotics constraints are exposed. They lead to two main results: 1/ to remain compatible with joint accelerations limits, the intuitive expression of joint position limits is modified; 2/ operational acceleration being dependent of the robot configuration, their compatibility with obstacle avoidance cannot be proved, so dedicated alternative safe behaviors are proposed. These results are illustrated though experiments on a 6-DOFs manipulator. The control problem being adequately formulated, its resolution algorithm has various specifications among which safety, optimality, efficiency, etc. Safety being a prerequisite, two compromises between those specifications are proposed. First, the Constraints Compliant Control law (CCC) based on a passive avoidance principle shows performing and robust features. Then, the use of a virtual displaced configuration in the control problem resolution enables to obtain a compromise between efficiency and optimality through the first safe single iteration resolution method.