Patent · US Active

Randomized reinforcement learning for control of complex systems

US11164077B2 · kind B2 · utility

1Cited by

1References

14Claims

0Family size

Assignee

Siemens Aktiengesellschaft · DE

Inventors

Siegmund Düll · Hersbruck, DE
Kai Heesche · Munich, DE
Raymond S. Nordlund · Orlando, US
Steffen Udluft · Birkenfeld, DE
Marc Christian Weber · Munich, DE

Key dates

Filing date	Nov 2, 2017
Grant date	Nov 2, 2021
Priority date	—
Expiry date	Sep 3, 2040

Classification

Technology area (CPC G)Physics
CPC primaryG06N20/00
WIPO fieldComputer technology
WIPO sectorElectrical engineering

Abstract

A method of controlling a complex system and a gas turbine being controlled by the method are provided. The method comprises providing training data, which training data represents at least a portion of a state space of the system; setting a generic control objective for the system and a corresponding set point; and exploring the state space, using Reinforcement Learning, for a control policy for the system which maximizes an expected total reward. The expected total reward depends on a randomized deviation of the generic control objective from the corresponding set point.

Source: USPTO / EPO open patent data. Objective bibliographic and citation counts.