English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Reinforcement learning (RL) is a promising, upcoming topic in automaticcontrol applications. Where classical control approaches require a priorisystem knowledge, data-driven control approaches like RL allow a model-freecontroller design procedure, rendering them emergent techniques for systemswith changing plant structures and varying parameters. While it was alreadyshown in various applications that the transient control behavior for complexsystems can be sufficiently handled by RL, the challenge of non-vanishingsteady-state control errors remains, which arises from the usage of controlpolicy approximations and finite training times. To overcome this issue, anintegral action state augmentation (IASA) for actor-critic-based RL controllersis introduced that mimics an integrating feedback, which is inspired by thedelta-input formulation within model predictive control. This augmentation doesnot require any expert knowledge, leaving the approach model free. As a result,the RL controller learns how to suppress steady-state control deviations muchmore effectively. Two exemplary applications from the domain of electricalenergy engineering validate the benefit of the developed method both forreference tracking and disturbance rejection. In comparison to a standard deepdeterministic policy gradient (DDPG) setup, the suggested IASA extension allowsto reduce the steady-state error by up to 52 $\%$ within the consideredvalidation scenarios.

Steady-State Error Compensation in Reference Tracking and Disturbance Rejection Problems for Reinforcement Learning-Based Control