The classical approach to decrease the energy consumption of high-speed robots is by lowering the moving elements mass in order to have a lightweight structure. Even if this allows reducing the energy consumed, the lightweight architecture affects the robot stiffness, worsening the accuracy of the mechanism. Recently, variable stiffness actuators (VSAs) have been used in order to reduce the energy consumption of high-speed pick-and-place robots. The idea is to smartly tune online the stiffness of VSA springs so that the robot is put in near a resonance mode, thus considerably decreasing the energy consumption during fast pseudo-periodic pick-and-place motions. However, the serial configuration of springs and motors in the VSA leads to uncontrolled robot deflections at high-speeds and, thus, to a poor positioning accuracy of its end-effector. In order to avoid these drawbacks and to increase the energy efficiency while ensuring the accuracy, this paper proposes the use of parallel arrangement of variable stiffness springs (VSS) and motors, combined with an energy-based optimal trajectory planner. The VSS are used as energy storage for carrying out the reduction of the energy consumption and their parallel configuration with the motors ensure the load balancing at high-speed without losing the accuracy of the robot. Simulations of the suggested approach on a five-bar mechanism are performed and show the increase on energy efficiency. 1 INTRODUCTION It is well-known that in industrial applications, such as high-speed pick-and-place operations, parallel robots are widely used [1, 2]. Repeatability and accuracy are typically the most important criteria to measure their performance. Nevertheless, the design trends to operate at high speeds are shifting to the design of robots with lightweight architectures [3] in order to decrease the energy consumed by the motors, and measure as well the robot performance based on its energy efficiency [4]. For slow motions, gravity-balancing techniques [5-8] have been proposed in order to compensate the input efforts required to move the links of a pick-and-place robot, and thus to avoid consuming energy. Even if these methods have shown their effectiveness at slow speeds, it is not the case for high-speed operations in which the inertial effects are preponderant. A first solution introduced the series elastic actuators (SEAs) [9] to cope with the energy storage issues. The SEAs are compliant actua-tors composed by a motor which is linked to a spring in series that serves as energy storage, and whose stiffness is set by the spring constant. SEAs were first used to absorb contact shocks and to reduce the peak forces due to the impacts in bipedal walking robots [10]. The limitation of the SEAs is that the stiffness is fixed and cannot be altered during motion, thus limiting the level of compliance to adapt for different tasks. Therefore, a recent second solution proposed the use of variable stiffness actuators (VSAs) [11-13] to handle with energy storage issues. VSAs con