A GENERAL FRAMEWORK FOR MINIMIZING ENERGY CONSUMPTION OF SERIES ELASTIC ACTUATORS WITH REGENERATION

Area of Science:

• Robotics
• Control Systems
• Optimization Theory

Background:

• Series elastic actuators (SEAs) are crucial for human-robot interaction due to their inherent compliance.
• SEAs offer reduced energy consumption compared to rigid actuators, but performance hinges on elastic element design.
• Current design methods (natural dynamics, parameterized optimization) have limitations regarding constraints and trajectory flexibility.

Purpose of the Study:

• To address limitations in SEA elastic element design.
• To propose a non-parametric convex optimization approach for globally optimal SEA elastic element design.
• To ensure actuator constraints are respected during the optimization process.

Main Methods:

• Formulation of SEA elastic element design as a non-parametric convex optimization problem.
• Mathematical proof of convexity for periodic reference trajectories with energy-regenerative SEAs.
• Application and analysis of the method for human ankle motion and a nonlinear mass-spring system.

Main Results:

• The proposed method yields a globally optimal conservative elastic element.
• The optimization successfully respects actuator constraints.
• Demonstrated energy reduction for human walking and nonlinear spring motion tasks.

Conclusions:

• The non-parametric convex optimization approach overcomes limitations of existing SEA design methods.
• This method provides a robust way to design energy-efficient SEAs that adhere to operational constraints.
• The findings are applicable to various dynamic tasks, including human motion emulation.