Reinforcement learning based engineering governance policy simulation optimization system

CN122113674AInactive Publication Date: 2026-05-29CHANGSHA UNIVERSITY OF SCIENCE AND TECHNOLOGY

Patent Information

Authority / Receiving Office: CN · China
Patent Type: Applications(China)
Current Assignee / Owner: CHANGSHA UNIVERSITY OF SCIENCE AND TECHNOLOGY
Filing Date: 2026-04-24
Publication Date: 2026-05-29
Estimated Expiration: Not applicable · inactive patent

AI Technical Summary

Technical Problem

In the governance scenario of multi-asset infrastructure clusters, existing technologies struggle to improve the trainability, executability, and rolling operation stability of governance solutions under conditions of multi-source heterogeneous data, complex rule constraints, and manual approval participation. This results in generated solutions being difficult to implement in management processes, leading to repeated adjustments to project schedules, increased resource consumption, and greater service risks.

Method used

By using a reinforcement learning-based engineering governance strategy simulation optimization system, feasible domain masks, evidence obligation vectors, and service bottom line templates are generated. Combined with a high-fidelity digital twin evaluator and approval memory, candidate action packages are classified for release. The digital twin evaluator and release classification thresholds are calibrated based on the approval results and execution receipts, and the master strategy model is updated.

Benefits of technology

It enables the organization of information on bridges, roads, tunnels, pumping stations, and pipelines under the same object semantics and time, reduces object misalignment and rule disconnection, improves the executability and rolling operation stability of governance solutions, reduces the frequency of unexecutable suggestions, and enhances the stability and efficiency of governance suggestions.

✦ Generated by Eureka AI based on patent content.

Smart Images

Figure CN122113674A_ABST

Patent Text Reader

Abstract

The application discloses an engineering governance strategy simulation optimization system based on reinforcement learning, relates to the technical field of engineering governance, and comprises the following contents: first, multi-source state data is collected and a unified state graph is constructed, governance rules are compiled into an executable constraint network, a feasible region mask, an evidence obligation vector and a service bottom line template are generated; second, a governance action package containing a target asset set, an intervention type, an execution time window, a resource package, a license package and a guarantee package is generated, and a set of actionable action packages is screened through a low-fidelity digital twin evaluator; then, the candidate action package is released and classified in combination with a high-fidelity digital twin evaluation result and an approval memory; finally, the digital twin evaluator and the release classification threshold are calibrated according to the approval result and an execution receipt, and a main strategy model is updated. The application can unify the expression of multi-source objects and governance rules, reduce unexecutable suggestions, and improve the stability of governance suggestions entering the approval and iteration.

Need to check novelty before this filing date? Find Prior Art