Control of an actuator on the basis of reinforcement learning

By employing reinforcement learning with discrete control parameters and output signal discretization, the challenges of parameterizing control circles for actuators are addressed, resulting in improved training success and actuator control accuracy.

WO2025067857A1 Publication Date: 2025-04-03ZF FRIEDRICHSHAFEN AG
6 Cites 0 Cited by

Patent Information

Application Number
PCT/EP2024/075148
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2023-09-29
Filing Date
2024-09-10
Publication Date
2025-04-03

AI Technical Summary

Technical Problem

The manual and time-consuming process of parameterizing control circles for actuators, especially electrohydraulic actuators, is challenging due to non-linearities and component influences, leading to uncertainty in assessing the influence of control parameter changes.

Method used

A procedure for controlling actuators using reinforcement learning, where a software agent selects discrete control parameters to maximize a reward, and discretization of continuous output signals to improve distinguishability and learning accuracy.

Benefits of technology

This approach enables reliable assessment of control parameter changes, improves the success of training, and enhances the control of actuators by clearly distinguishing the influence of individual parameters, thus optimizing the reward and actuator control.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention relates to a method for controlling an actuator on the basis of control parameters generated using reinforcement learning and on the basis of a corresponding device and to a corresponding training method. The method for controlling an actuator on the basis of control parameters generated using reinforcement learning has the steps of determining (S1) a regulation parameter set with a plurality of discrete regulation parameters for the respective actuator, receiving (S2) a control parameter generated by a software agent, mapping (S3) the control parameter to one of the regulation parameters of the regulation parameter set, and controlling (S4) the actuator on the basis of the regulation parameter. The invention additionally relates to a method for an improved training of a software agent for controlling an actuator and to a device for carrying out the method.
Need to check novelty before this filing date? Find Prior Art