CN-122015840-A - BIM-based construction site intelligent navigation and task allocation method

Abstract

The invention provides a construction site intelligent navigation and task allocation method based on BIM, and belongs to the technical field of building informatization and robot automatic construction. The method comprises the steps of extracting semantic information such as wall materials, construction period, room information and the like from a BIM model, detecting construction key nodes such as doors and windows, inflection points, wall positions and the like from images by combining a neural network model, extracting feature vectors and fusing the feature vectors and the semantic information. And then, constructing a hierarchical topological map based on materials, rooms and construction periods through multi-dimensional hierarchical and dynamic weight distribution, and realizing the clustering of construction nodes and the selection of central nodes. The topological map combines the meta controller and the sub controller, so that efficient path planning and obstacle avoidance control are realized, and the construction efficiency and safety are improved. The invention obviously optimizes the construction sequence, reduces the navigation cost and energy consumption, and is effectively applicable to complex construction environments.

Inventors

• LI LIQIANG
• JIANG ZHIJUN
• LIU JIN
• ZHANG XIAOGUO
• YANG YUAN
• LI ZIQIANG
• WANG YIHAO
• WANG ZHENGDONG
• ZHAO XIAOBING
• SHEN YUE
• MENG XIANGYIN
• ZU YU
• CHENG LE

Assignees

• 中交建筑集团有限公司
• 中交建筑集团第一工程有限公司
• 东南大学

Dates

Publication Date

20260512

Application Date

20251231

Claims (6)

1. 1. The construction site intelligent navigation and task allocation method based on BIM is characterized by comprising the following specific steps: Processing a BIM model to obtain room semantic information and construction key node information, combining the nodes with the semantic information, and extracting indoor building information; The method comprises the steps of deriving IFC information from BIM, using an IFC library to read IFC files, loading model data, searching related entities through scripts to extract attributes, extracting information including wall materials, construction period and room information semantic information, converting the BIM model into images, converting the models into house integral and room images through an automatic script and a 3D rendering engine, setting camera view angles and rendering and storing the images in batches, detecting corresponding key nodes of door windows, room inflection points and wall positions from the images through a ResNet-50 and CBAM attention mechanism, extracting feature vectors, emphasizing important parts of input features through a weight matrix through the attention mechanism, and finally combining the identified node position information with semantic information; Step two, constructing a layered topological map by node multidimensional layering, clustering combination and dynamic weight distribution and combining construction information; In the process of constructing the hierarchical topological map, node sets are extracted from the BIM model, construction of walls of different materials is considered to be performed respectively, construction of nodes in the same room is more convenient, construction period urgency principles are utilized to arrange construction sequences, then grouping nodes are aggregated into dense clusters through a clustering algorithm, central nodes of each group of nodes are calculated, then the central nodes are fused into a main node of the topological map by utilizing dynamic weight information, the advantages of construction in terms of materials, rooms and construction period are ensured, through verifying reachable relations, other nodes are used as secondary nodes to be connected with the main node, meanwhile connection penetrating through obstacles is eliminated, and finally, the hierarchical topological map is constructed accurately; Step three, a kinematic model of the robot is established, a hierarchical controller is designed, a primary node and a secondary node are selected in a hierarchical topological map by a meta controller, a sub controller performs an obstacle avoidance function to ensure that collision does not reach a sub target, and a navigation target is completed; When a kinematic model of the robot is established, the meta controller preferentially selects construction nodes close to each other through regional compactness analysis, adopts a neighborhood node priority strategy to select secondary nodes close to the main nodes and having urgent construction periods, follows the unified principle of construction procedures, preferentially processes nodes with consistent materials and task types, finally evaluates and selects the shortest path, and the sub controllers are responsible for avoiding barrier functions to achieve collision-free arrival of selected target nodes, and generate collision-free tracks according to the current state and the selected nodes, sense barriers in the environment in real time, and control the movement of the robot according to the kinematic modeling of the robot so as to safely navigate to the target nodes.
2. 2. The intelligent navigation and task allocation method based on BIM according to claim 1, wherein the specific process of the step one is as follows: given an input image Extracting feature map by ResNet-50 The method is characterized by comprising the following steps: ; Wherein the method comprises the steps of Representation ResNet-50 convolutional neural network for image feature extraction, comprising 50 layers, using residual blocks to solve the gradient vanishing problem for deep networks, one residual block comprising three convolutional layers, assuming input as Output of residual block The method comprises the following steps: ; Wherein the method comprises the steps of Is a residual function, and has the specific structure as follows: ; Here, the The weights of the three convolutional layers respectively, Is the corresponding bias term; feature enhancement is carried out by using CBAM to obtain an enhanced feature map : ; Wherein the method comprises the steps of Representative attention module comprising two modules, a channel attention module and a spatial attention module, the channel attention module The method comprises the following steps: ; Wherein the method comprises the steps of Is a Sigmoid function of the code, Is a multi-layer sensing machine, which comprises a main body, And Global average pooling and global maximum pooling, respectively; the spatial attention module F s calculates: ; Wherein, the Is a convolution operation and is performed by, Is a splicing operation, and the method comprises the steps of, And Respectively carrying out average pooling and maximum pooling; The final weighted feature map F' is: ; Wherein the method comprises the steps of Representing element-by-element multiplication; Identifying the target to determine its two-dimensional position coordinates in the image based on the obtained feature vector of the target, assuming that the detected node is Its position coordinates are Fusing location information with semantic information to create a comprehensive node description expressed as, for each node Comprehensive description thereof Represented as a tuple containing the location coordinates and associated semantic attributes 1; ; Wherein, the Is a node The coordinates of the position in the image, Is a collection containing semantic information; ; Combining the comprehensive descriptions of all nodes to form a complete model description : ; Wherein the method comprises the steps of The total number of detected nodes is made.
3. 3. The intelligent navigation and task allocation method based on BIM according to claim 1, wherein the second step is as follows: Firstly, dividing all nodes into a plurality of layers according to different attributes of the nodes, wherein the attributes are wall materials, rooms and construction period urgency of the nodes, each subset comprises nodes with the same attributes, and the nodes are arranged For a set of all the nodes, Represented as a tuple of one of the elements, To have the same semantics Is then: ; Wherein, the Is a room type, wall material, and construction period attribute; node subset at each semantic attribute In the method, a clustering algorithm is applied to divide the nodes into a plurality of clusters, and For grouping Is used for the number of clusters of (a), Is the first The clusters are as follows: ; Wherein, the Is a node An index of the assigned cluster; For each cluster Calculating the center node of the center The central node is the average value of the spatial position attribute values of all nodes in the cluster and represents the characteristics of the cluster, and is provided with Is the first Center node of each cluster, then: ; Wherein, the Representing clusters The number of nodes in the (c) tree, Representing nodes Coordinates of (c); Calculating a master node of a topological map according to the central nodes of each layer of clusters, wherein each layer of central nodes is required to be allocated with a dynamic weight, and the weight is set based on a plurality of factors including construction period urgency, room type and wall materials As a total number of weight factors, Is the first The weight vector of the individual weight factors, As a central node In the first place Score on individual weight factors, then center node Is the integrated weight of (2) The method comprises the following steps: ; Using optimization method to calculate master nodes of topological map according to hierarchical center nodes, aiming at calculating position capable of minimizing weighted distance between nodes in whole topological map, setting And Respectively being central nodes And Is used for the weight of the (c), As a central node And The distance between them, the optimization problem is expressed as: ; verifying reachability between the master nodes is accomplished by a path search algorithm in graph theory to ensure that all master nodes are interconnected, then connecting the master nodes and the secondary nodes together to form a topology map, A topological map is represented and is displayed, Representing a set of master nodes, Representing the edge set, then: ; finally, a hierarchical topology is constructed that takes into account the various attributes, ensuring that tasks face the constraints of materials, rooms and construction periods.
4. 4. The intelligent navigation and task allocation method based on BIM according to claim 3, wherein in the second step, in the process of constructing the hierarchical topological map, node sets are extracted from the BIM model, and the construction sequence is arranged by considering that walls with different materials are required to be constructed respectively, the construction of nodes in the same room is more convenient, and the construction period urgency principle is as follows: the method comprises the steps of carrying out most layering on nodes, dividing the nodes into spatial layering, grouping the nodes according to room information so that the nodes in the same room can be efficiently constructed, carrying out material layering, dividing the nodes into a plurality of groups according to different wall materials, ensuring consistency of processing modes and used materials in the construction process, carrying out construction period layering, sorting and grouping the nodes according to construction period urgency, and ensuring reasonable distribution of construction progress.
5. 5. The intelligent navigation and task allocation method based on BIM according to claim 3, wherein the three-way controller selects the primary node and the secondary node in the hierarchical topological map, and the specific process is as follows: firstly, analyzing the compactness of a region, preferentially selecting nodes which are concentrated in construction tasks and have close space distances, and defining a compactness function To evaluate the compactness of nodes in a region, and set the main nodes in the region as a set Node And The distance between them is , Is a node Is a construction task amount. Then: ; Wherein, the Smaller, representing nodes The more concentrated the construction task, the closer the space distance, the more compact the node distribution; Then, the meta-controller is based on Value of (2) to master node Is selected from the group consisting of: ; Wherein the method comprises the steps of Refers to one node in the set N; Then adopting a neighborhood node priority strategy to preferentially select secondary nodes close to the main node and having urgent construction period so as to reduce navigation cost, and setting Is a node Is used for solving the problem of the urgency of the construction period of the (a), For the distance between the primary node and the secondary node, the priority of the secondary node The method comprises the following steps: ; the meta-controller is based on the priority of each secondary node Sorting and selecting secondary node set with larger priority : ; According to the unified principle of construction procedures, nodes with identical materials and consistent task types are preferentially processed so as to reduce the switching of robot working modes and define the task types And wall material Representing nodes respectively Construction information of (a) and construction of process similarity functions between nodes : ; Wherein, the In order to indicate the function, The larger the material and task type of the representing node are unified with the main node, the secondary node selection is adjusted according to the result Wherein Representing nodes The process similarity of (2) is: ; finally, evaluating the total running path, evaluating the running paths from the current node to the main node and the secondary node, selecting the shortest path to reduce the energy consumption, and calculating the total running path : ; Wherein the method comprises the steps of Optimizing secondary node combinations for distances between two nodes based on results : ; Finally, the meta-controller selects the optimal master node And a secondary node set 。
6. 6. The intelligent navigation and task allocation method based on BIM according to claim 3, wherein the step three subcontroller performs the obstacle avoidance function to ensure collision-free arrival of the subcontact, and the specific process is as follows: the sub-controller is responsible for generating a collision-free track and controlling the robot to safely reach the sub-target First, sub-targets selected according to meta-controllers Position coordinates of (a) Converting it into a target position in a local obstacle avoidance coordinate system The specific conversion formula is as follows: ; Wherein, the The current orientation angle of the robot; Next, based on the target position in the local obstacle avoidance coordinate system Calculating the target direction to which the robot should be directed : ; Target direction The direction in which the robot should move under the condition of no obstacle; In the obstacle avoidance process, the robot needs to dynamically adjust the angular speed according to the distribution of the obstacles Assuming that the robot is driven by differential speed, its linear velocity Constant, angular velocity The calculation process of (2) is as follows: Selecting the current direction angle of the robot Taking a look-ahead point under the local obstacle avoidance coordinate system, wherein the look-ahead point is 1 unit distance away from the origin, and obtaining: ; Wherein R is the radius of an arc formed by the current coordinate and the forward looking point, namely the turning radius of the differential driving robot, so that: ; according to turning radius of differential driving robot Sum linear velocity Relation of (2) Calculating angular velocity : ; Wherein, the The target angle to which the robot should turn; In order to avoid overlarge angular speed fluctuation of the robot in the obstacle avoidance process, a first-order IIR filter is adopted for diagonal speed And performing smoothing treatment, wherein a difference equation of the filter is as follows: ; Wherein, the As the angular velocity at the previous moment in time, A new angular velocity calculated for the current moment, In order to obtain the angular velocity after the filtering, Is a weight constant.

Description

BIM-based construction site intelligent navigation and task allocation method Technical Field The invention relates to the technical field of building informatization and robot automatic construction, in particular to a construction site intelligent navigation and task allocation method based on BIM. Background Along with the informatization and digital development of the building industry, the Building Information Model (BIM) technology is widely applied in China. BIM technology provides abundant information support for links such as design, construction, operation and the like by constructing a three-dimensional building model. However, how to efficiently utilize information in the BIM model in the construction process to realize efficient planning of the indoor building robot facing the indoor complex construction problem is still a problem to be solved. At present, the navigation and task allocation of the construction site are not scientific in construction path planning, the path planning of the construction site is often based on experience, and the selection of an optimal path is difficult to realize, so that the energy consumption is high and the efficiency is low in the construction process. The task scheduling is disordered, the task scheduling on the construction site is lack of optimization, and the problems of unreasonable task execution sequence, influence on the construction progress and the like are easily caused. In order to solve the above problems, researchers have proposed various solutions, such as a construction resource optimization allocation method based on a genetic algorithm, a construction path planning method based on an ant colony algorithm, and the like. However, these methods still have the limitations of high complexity of the model, slow calculation speed, too much dependence on empirical parameters and poor versatility in practical application. Compared with the technology of Chinese patent CN120147573A 'a three-dimensional map key feature extraction and topology map construction method' of a transformer substation construction site: The data source is different from the semantic information processing mode. The application (BIM-based construction site intelligent navigation and task allocation method) extracts information from a Building Information Model (BIM), and the BIM is a structured data source containing rich semantics (such as wall materials, construction period and room functions). Chinese patent CN120147573a is based on three-dimensional point cloud and image data collected on site, these are raw, unstructured data, and one of the key of the methods is to extract key features and semantic information from these raw data using Mamba algorithm. Briefly, the present application utilizes existing design semantic information, while chinese patent CN120147573a is directed to inferring semantics from post-collected geometric data. The core technology is essentially different from the algorithm. The present application uses convolutional neural network (ResNet-50 and CBAM) based methods to identify building structure nodes such as doors, windows, walls, etc. from BIM converted images. Chinese patent CN120147573a uses Mamba algorithm (a state space model) to process point cloud data to extract key information of power equipment, power line, etc. The two are completely different in the deep learning model for feature extraction. The idea of map construction is different from that of emphasis. The application constructs a layered topological map serving construction flow optimization, which performs layering, clustering and dynamic weighting on nodes according to a plurality of dimensions such as construction period, materials, space and the like, and aims to optimize construction sequence and robot task allocation. The key point of the construction of the topological map in the China patent CN120147573A is to truly and reliably reflect the space geometric relationship of the field environment, and the steps of connection relationship reliability evaluation and map optimization adjustment are designed to ensure the accuracy of the map. The final application targets are different. The application aims to provide a complete intelligent navigation and task distribution system, a constructed map is an intermediate tool for realizing efficient path planning and task scheduling of a robot, and a meta controller and a sub controller are designed to realize the closed loop system. The core output of chinese patent CN120147573a is the "topological map construction method" itself, which provides more reliable model support for subsequent applications such as path planning, target positioning, etc., but its implementation mainly focuses on the map construction stage. Disclosure of Invention In order to solve the technical problems, the invention provides a BIM-based construction site intelligent navigation and task allocation method, which is used for effectively extracting the sem