Robust Energy-Efficient Design of Series Elastic Actuators
Bolívar Nieto, Edgar Alberto
MetadataShow full item record
In the same manner that muscles benefit from the elasticity of tendons to be more energy efficient and powerful, electric motors can benefit from elastic elements to improve the efficiency and reliability of robots. For more than twenty years, roboticists have purposefully used springs to connect the motor and the moving joint aiming for efficient and reliable motion. However, there is no clear understanding of how to select the torque-elongation profile of the spring to minimize the motor energy consumption while satisfying motor speed-torque limitations and maximum elongation of the spring. Existing methods cannot guarantee a torque-elongation profile that is better than any other when analyzing arbitrary periodic motion of the joint, and cannot guarantee that the speed-torque of the motor and elongation of the spring are within safe limits when the parameters of the motor, tolerance in manufacturing of the spring, and definition of this periodic motion are uncertain. This dissertation addresses this issue by formulating the design of the torque-elongation spring profile as a robust convex optimization program. The resulting spring profile is guaranteed to be the best among all increasingly monotonic profiles, linear or nonlinear, and satisfies motor and spring constraints as long as the uncertainty stays within the uncertainty sets. As a case study, this dissertation applies the proposed formulation to the design of a series elastic actuator of a powered prosthetic ankle.