Large model-based algorithm migration and automatic parameter tuning method and related equipment
By collaborating with a large model and a rule engine, an intermediate representation is constructed and simulation results-driven closed-loop optimization of parameter tuning is performed. This solves the problems of insufficient determinism and poor traceability in the algorithm migration and parameter tuning process in complex control systems, and realizes deterministic generation and parameter tuning of simulation platform components.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SUZHOU TONGYUAN SOFT CONTROL INFORMATION TECH CO LTD
- Filing Date
- 2026-06-12
- Publication Date
- 2026-07-17
AI Technical Summary
In the research and development of complex control systems, the algorithm migration and simulation parameter tuning processes suffer from insufficient determinism and poor traceability. Existing technologies make it difficult to directly integrate algorithms into the simulation environment and tune parameters, and the output uncertainty of large models cannot meet the deterministic and reliable requirements of industrial simulation scenarios.
By acquiring the business algorithm code, performing syntax parsing and business semantic annotation using a large model, constructing an intermediate representation, and combining it with the pre-defined platform development specifications, using a rule engine to perform deterministic transformation, generating simulation platform components that conform to the specifications, and executing closed-loop optimization driven by simulation results until the evaluation indicators meet the requirements.
It achieves high determinism and strong traceability in the algorithm migration and parameter tuning process, ensuring that the conversion results meet the interface dimension constraints and engineering specifications, forming a controllable closed loop, and solving the problem of insufficient determinism in the algorithm migration and parameter tuning process.
Smart Images
Figure CN122411901A_ABST