Convex Optimization-based Motion Planner for Linear Variable Stiffness Actuators

This paper is focused on control of variable stiffness actuators (VSA). One of the new approaches in this area is based on a linear VSA model, whose main features are the directly controlled linear stiffness dynamics and bilinear output torque model. The paper tackles the problem of motion planning for linear VSA. While the problem appears to be non-convex, the study demonstrates that it is possible to introduce a new parametrization of the model that produces a convex optimization formulation. It is proposed to limit the power input allowed for the controller, making sure that it does not exceed the physical capabilities of the motors implementing the trajectories. The paper shows the behavior of the planer when it reaches the limits of the allowed power input. The paper demonstrates the use of the proposed planner in planning trajectories for a lower-limb exoskeleton performing sit-to-stand transfer.