CN-121981508-A - City update intelligent building platform and intelligent scheduling method

CN121981508ACN 121981508 ACN121981508 ACN 121981508ACN-121981508-A

Abstract

The invention discloses an intelligent building platform for urban updating and an intelligent scheduling method, and relates to the technical field of urban updating, wherein a system comprises a data fusion and building unit, an intelligent capsule bin cluster, a self-adaptive scheduling unit and an augmented reality man-machine cooperation unit. The method and the system can improve the instantaneity and the executable performance of the scheduling decision, and can finely control the construction process, thereby improving the construction efficiency of urban updating construction.

Inventors

LI YONGWEN
LI YONGMING
ZHU XIAONING
ZHANG BAOMIN
JIN LIXIU
SUN YUE
ZHANG MING
Xiao Quanqin

Assignees

彩旺控股集团有限公司

Dates

Publication Date: 20260505
Application Date: 20260409

Claims (10)

1. The city updating intelligent building platform is characterized by comprising the following modules: The data fusion and construction unit is used for acquiring multi-source data of the urban updating area and constructing a four-level construction scene digital twin model comprising a geometric layer, a physical layer, a semantic layer and a dynamic layer, wherein the dynamic layer maps intelligent capsule bin position data, personnel position data, environment monitoring data and dynamic obstacle data in real time; the intelligent capsule bin cluster comprises a plurality of intelligent capsule bins with heterogeneous functions, wherein each intelligent capsule bin is integrated with an environment sensing and communication module and a pluggable task module; the environment sensing and communication module is used for sensing the surrounding environment in real time, constructing a local occupation grid map and transmitting data; The pluggable task module is interchangeably arranged on the intelligent capsule bin through a standardized interface and is used for executing specified construction procedures; The self-adaptive scheduling unit is used for generating scheduling instructions based on deep reinforcement learning and comprises a state space construction module and a decision scheduling module; The state space construction module is used for extracting system state characteristics at the current moment from the digital twin model to form a high-dimensional state vector; and the decision scheduling module is deployed with a pre-trained deep reinforcement learning model and is used for outputting a cooperative scheduling strategy according to the high-dimensional state vector, generating a scheduling instruction and transmitting the scheduling instruction to a corresponding intelligent capsule bin.
2. The city update intelligent building platform of claim 1, wherein in the data fusion and construction unit, the semantic layer is configured to store semantic attribute information in a city update region, the semantic attribute information including a region function attribute, a time constraint attribute, and a security attribute; The self-adaptive scheduling unit further comprises a conflict prediction and potential field construction module, wherein the conflict prediction and potential field construction module is used for constructing a space-time combined potential field for describing the risk distribution of the operation space based on semantic attribute information and intelligent capsule bin position data, personnel position data, environment monitoring data and dynamic obstacle data mapped in real time in a dynamic layer, and taking the gradient or potential field value of the space-time combined potential field as the input characteristic of a high-dimensional state vector, and the deep reinforcement learning model is trained to learn and output a cooperative scheduling strategy under the constraint of the potential field.
3. The urban updating intelligent construction platform according to claim 2, wherein in the conflict prediction and potential field construction module, the space-time combined potential field comprises a static potential field and a dynamic potential field, the static potential field is constructed according to regional functional attributes in a semantic layer, the potential field strength of the static potential field has direction dependence, the dynamic potential field is constructed and updated in real time according to intelligent capsule bin position data and is used for representing mutual avoidance requirements among intelligent capsule bins, and the rejection strength function of the dynamic potential field is as follows: , wherein, For Euclidean distance between the current intelligent capsule bin and the obstacle intelligent capsule bin, q represents pose vector of the current intelligent capsule bin in the configuration space, comprises position coordinate and orientation angle information, The radius is affected for a preset repulsive force, For the relative speed between the current smart capsule bin and the obstacle smart capsule bin, For a preset maximum relative speed, η is a first gain factor and α is a second gain factor.
4. The urban updating intelligent construction platform according to claim 3, wherein in the data fusion and construction unit, the semantic attribute information further comprises marked anisotropic constraint areas, and the anisotropic constraint areas comprise personnel passing directions, cultural relic protection area directional forbidden and narrow channel passing directions; in the conflict prediction and potential field construction module, a dynamic potential field is constructed based on an anisotropic constraint area marked by a semantic layer so as to reflect anisotropic constraint in an urban updating scene, the rejection strength of the dynamic potential field changes along with the included angle of the intelligent capsule bin relative to the constraint direction, and the rejection strength function of the dynamic potential field is further updated as follows: wherein θ is the movement direction angle of the current intelligent capsule bin relative to the obstacle intelligent capsule bin, Beta is the third gain factor, which is the principal direction of constraint marked by the semantic layer.
5. The urban updating intelligent construction platform according to claim 1, wherein in the decision scheduling module, a multi-agent depth deterministic strategy gradient algorithm is adopted, each intelligent capsule bin is regarded as an agent, a distributed cooperative strategy is learned, and the training process of the depth reinforcement learning model comprises the steps of constructing a high-fidelity simulation environment based on a digital twin model, and designing a reward function R= X progress rewards + X-efficient rewarding- Xconflict penalty- Energy consumption penalty- X disturbing people punishment, wherein, As a first weight coefficient, a first set of weights, As a result of the second weight coefficient, As a result of the third weight coefficient, For the fourth weight coefficient, the first weight coefficient, Performing off-line training in a simulation environment for a fifth weight coefficient, and performing on-line fine adjustment on model parameters based on real feedback data; The self-adaptive dispatching unit further comprises a feedback learning module, wherein the feedback learning module is used for collecting an actual execution result of the intelligent capsule bin after executing the dispatching instruction to form a feedback sample, and the feedback sample is used for carrying out incremental training on the deep reinforcement learning model regularly to realize self-adaptive optimization of a dispatching strategy.
6. The urban updating intelligent construction platform according to claim 1, further comprising an augmented reality man-machine cooperative unit, wherein the augmented reality man-machine cooperative unit is used for receiving a scheduling instruction generated by the decision scheduling module and converting the scheduling instruction into augmented reality guiding information to be pushed to a field personnel terminal, and the augmented reality guiding information comprises a personnel positioning and tracking module, a dynamic security fence generating module and an instruction generating and pushing module; the personnel positioning and tracking module is used for tracking the position of the personnel on site in real time and updating the position to the dynamic layer of the digital twin model; The dynamic safety fence generation module is used for generating a dynamic safety fence in real time according to the motion trail of the intelligent capsule bin and triggering grading early warning when personnel enter; The instruction generation and pushing module is used for automatically generating and pushing the augmented reality guide information comprising the operation point position marks, the operation step guide and the safety prompt.
7. The urban updating intelligent construction platform according to claim 1, wherein in the data fusion and construction unit, a geometric layer is used for storing three-dimensional space coordinates, dimensions and topological relations, and a physical layer is used for storing material properties, structural loads and equipment performance parameters; The pluggable task module is selected from one or more of a detection module, a spraying module, a reinforcement module and a carrying module, wherein the standardized interface comprises a standardized mechanical interface and an electrical interface, and the scheduling instruction comprises one or more of a task allocation instruction, a path adjustment instruction, an avoidance coordination instruction and a cooperative instruction.
8. Urban updating intelligent scheduling method, applying the urban updating intelligent construction platform according to any one of claims 1-7, characterized by comprising the following steps: The step of constructing a digital twin model, which is to acquire multi-source data of an urban updating area and construct a four-level construction scene digital twin model comprising a geometric layer, a physical layer, a semantic layer and a dynamic layer, wherein the dynamic layer maps the position of an intelligent capsule cabin, the position of personnel, environmental monitoring data and dynamic barrier data in real time; a step of updating the dynamic layer in real time, which is to sense the surrounding environment of the intelligent capsule bin in real time, construct a local occupation grid map, transmit data and update the dynamic layer in real time; extracting state characteristics, namely extracting system state characteristics from a digital twin model at the current moment to form a high-dimensional state vector; a step of generating a scheduling instruction, which is to input the high-dimensional state vector into a pre-trained deep reinforcement learning model, output a cooperative scheduling strategy, generate the scheduling instruction and issue the scheduling instruction to a corresponding intelligent capsule bin; and executing the scheduling instruction, namely executing the designated construction procedure by the intelligent capsule bin according to the scheduling instruction through the installed standardized interface.
9. The method for intelligently scheduling city update according to claim 8, wherein in the step of constructing the digital twin model, the semantic layer is used for storing semantic attribute information in a city update area, the semantic attribute information comprises an area function attribute, a time constraint attribute, a security attribute and a marked anisotropic constraint area, and the anisotropic constraint area comprises a personnel passing direction, a cultural relic protection area directivity forbidden and a narrow channel passing direction; A step of constructing a space-time combined potential field is further arranged between the step of extracting the state characteristics and the step of generating the scheduling instruction, wherein the space-time combined potential field describing the risk distribution of the operation space is constructed based on semantic attribute information stored by a semantic layer and data mapped in real time by a dynamic layer, and the gradient or potential field value of the space-time combined potential field is fused into the high-dimensional state vector; The space-time combined potential field comprises a static potential field and a dynamic potential field, the static potential field is constructed according to regional functional attributes in a semantic layer, the potential field strength of the static potential field has direction dependence, the dynamic potential field is constructed and updated in real time according to intelligent capsule bin position data, and a direction modulation factor is introduced based on an anisotropic constraint region and used for representing the mutual avoidance requirement between intelligent capsule bins, the rejection strength of the dynamic potential field changes along with the included angle of the intelligent capsule bins relative to the constraint direction, and the rejection strength function of the dynamic potential field is as follows: , wherein, For Euclidean distance between the current intelligent capsule bin and the obstacle intelligent capsule bin, q represents pose vector of the current intelligent capsule bin in the configuration space, comprises position coordinate and orientation angle information, The radius is affected for a preset repulsive force, For the relative speed between the current smart capsule bin and the obstacle smart capsule bin, For the preset maximum relative speed, eta is a first gain coefficient, alpha is a second gain coefficient, theta is the movement direction angle of the current intelligent capsule bin relative to the obstacle intelligent capsule bin, Beta is the third gain factor, which is the principal direction of constraint marked by the semantic layer.
10. The urban updating intelligent scheduling method according to claim 8, further comprising the step of augmented reality man-machine cooperation, wherein the step of receiving the generated scheduling instruction, converting the scheduling instruction into augmented reality guiding information and pushing the augmented reality guiding information to a field personnel terminal, wherein the augmented reality guiding information comprises operation point position marks, operation step guiding and safety prompts, simultaneously tracking the position of the field personnel in real time and updating the position of the field personnel to a dynamic layer of a digital twin model, generating a dynamic safety fence in real time according to the motion trail of the intelligent capsule bin, and triggering grading early warning when personnel enter.

Description

City update intelligent building platform and intelligent scheduling method Technical Field The invention relates to the technical field of city updating, in particular to an intelligent building platform and an intelligent scheduling method for city updating. Background With the urban process of China entering the stock upgrading stage, urban updating becomes the main battlefield of urban construction. The city updating is necessary and planned reconstruction activities for areas which are not suitable for modern city social life in a city built-up area, and comprises the projects of old district reconstruction, historical block repair, existing building function improvement, infrastructure updating and the like, and relates to various engineering types such as pipe network reconstruction, facade repair, elevator installation, road repair and the like. Compared with a newly built project, the urban updating project is usually positioned in a densely populated and space-limited built area, and has the characteristics of complex construction environment, narrow site, multiple peripheral interference factors, sensitivity to the influence of resident life and the like. The project often needs to be constructed in a normal living environment of residents, and higher requirements are put on construction organization, construction period control and environmental protection. How to efficiently organize multi-professional construction in a limited space, furthest reduce interference to residents and ensure construction safety and quality becomes a key technical problem to be solved in the field of urban updating. In early practice of urban updating projects, the method mainly depends on a traditional manual construction mode, and adopts a staged and serial operation mode in terms of construction organization, namely, the construction is carried out by entering the steps in sequence. For example, in the engineering of outer wall reconstruction, a scaffold is firstly erected, then the foundation treatment and spraying operation are manually carried out, finally the scaffold is dismantled, and the subsequent pipeline reconstruction team can enter. In the aspect of field management, mainly rely on manual inspection and paper record, the security officer needs to examine the construction standard execution condition at the construction site itinerary, and the quality officer needs to carry out manual inspection to key processes, lacks effective cooperation mechanism between construction machinery and the operating personnel. In order to solve the above problems, some intelligent construction management techniques have appeared in recent years. The construction process digital management platform is used for acquiring construction progress images and environment data through deployment of site cameras and sensors, and is combined with a building information model to carry out visual display, a manager can remotely check site conditions through the platform, track construction progress, and part of the platform also has simple data statistics and report generation functions. The technology realizes the digital acquisition and visual presentation of construction information, but still takes the manual decision as the main part, and the platform only provides information support. Another type of technology is the single machine intellectualization of automated construction equipment, for example, an intelligent spraying robot can autonomously complete wall surface spraying operation and has a simple obstacle avoidance function, an intelligent detection robot can perform structural scanning along a preset path, and an intelligent feeding car can convey materials according to site requirements or preset requirements. The equipment realizes automatic operation on specific procedures, improves the efficiency of single procedures, but lacks the capability of unified scheduling and collaborative operation among the equipment. In spite of the progress of the prior art, the prior dispatching system mainly has the defects that the prior dispatching system mostly adopts centralized static planning or simple time window distribution, cannot respond to dynamic changes of a construction site in real time, particularly has extremely limited avoidance space among devices under the scenes of old community narrow roadways and the like, is difficult to adjust a dispatching strategy in time when sudden conditions such as device faults, personnel walking, material temporary stacking and the like occur, causes device waiting, path conflict and even operation interruption, adopts a uniform static constraint model, the progress, quality and safety data collected by the construction site cannot be fed back to a dispatching decision link in real time, and has a large number of constraint conditions of dynamic changes in a city updating scene, such as the situations that residents need to pause material transportation in a peak period, the peripher