Practical Approach to Characterize Realistic Motor Dynamics for Robotic Simulation Independent of the Use Case

Abstract

Incorporating realistic actuator dynamics in robotic simulations is key to a successful simulation-to-reality transfer. But real actuation chains are often complex and impossible to model with analytical methods alone. Although it is feasible to reverse-engineer the actuator dynamics from hardware measurements, this requires the completed robotic system to be already available. To enable the inclusion of realistic actuator dynamics in robot models also during the design phase or for initial controller tuning, this work presents an alternative hands-on approach for actuator characterization. Based on actuator measurements taken independently of the overall system integration, a model expression for the actuator is derived. This can be added to the simulation of any robotic system. To showcase this concept, we present the workflow for a robotic leg with a Series Elastic Actuation chain. We create a simulation of the leg incorporating the derived actuator model and show its validity through comparison with analogous hardware. The observed motor and link dynamics of both cases show close correspondence without increasing the needed computation times with respect to a simulation without actuation. Thus, the proposed method offers a promising approach to include realistic actuator dynamics during the design and development process of robotic applications.