Construction site resource optimization scheduling system based on deep learning

CN122434201APending Publication Date: 2026-07-21SOUTH VIETNAM CONSTR MANAGEMENT CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
SOUTH VIETNAM CONSTR MANAGEMENT CO LTD
Filing Date
2026-06-03
Publication Date
2026-07-21

AI Technical Summary

Technical Problem

Existing construction resource scheduling methods are difficult to effectively couple the spatiotemporal physical constraints and task logic dependencies in unstructured environments, and cannot balance computational real-time performance with the physical feasibility of scheduling schemes when dealing with large-scale dynamic scheduling problems.

Method used

A resource optimization and scheduling system for construction sites employing deep learning works collaboratively with a deep neural network and a mathematical programming solver to output resource scheduling results that satisfy preset physical constraints and task logic constraints. The system includes a data acquisition interface unit, a dynamic graph tensor construction unit, a dual-headed perception prediction unit, a hot-start solution unit, and a scheduling instruction execution unit. It utilizes a deep neural network to process heterogeneous graph tensors and achieves hot-start solution for resource scheduling through physical interference potential field characteristic calculation and Lagrange multiplier prediction.

Benefits of technology

It reduces the risk of spatial conflicts between mechanical equipment, shortens the solution time in dynamic scheduling scenarios, improves the adaptability and constraint satisfaction of the model in similar construction scheduling scenarios, and ensures the physical feasibility and real-time response capability of the scheduling scheme.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122434201A_ABST
    Figure CN122434201A_ABST
Patent Text Reader

Abstract

The application relates to the technical field of intelligent construction and engineering management, and discloses a building construction site resource optimization scheduling system based on deep learning, which comprises a data acquisition interface unit, a dynamic graph tensor construction unit, a double-head dual perception prediction unit, a hot start solving unit and a scheduling instruction execution unit. The system converts construction site data into a heterogeneous graph tensor containing physical interference potential field characteristics, uses a double-head deep network to predict a quadratic programming target parameter and a Lagrange multiplier vector of a KKT condition in parallel, and uses the multiplier vector as an initial dual variable of an original dual interior point method solver to perform hot start solving. By introducing physical potential field characteristic embedding and Lagrange multiplier hot start mechanism, the application can introduce spatial constraint information and task logic relationship of the construction site into the resource scheduling solving process together, generate a scheduling solution meeting preset constraint conditions under a constraint model, reduce the iteration number of the original dual interior point method solver, and improve the solving efficiency under a dynamic scheduling scenario.
Need to check novelty before this filing date? Find Prior Art