Building robot multi-machine cooperative construction planning method and system

Abstract

The invention discloses a building robot multi-machine cooperative construction planning method and system, and belongs to the technical field of building construction automation, and the method comprises the steps: collecting three-dimensional space data of a construction site, combining with a timestamp, constructing a four-dimensional space-time semantic map, recognizing a process stage, dynamically allocating tasks, generating an initial path through an improved A * algorithm, and constructing a three-dimensional space-time semantic map. And the path is adjusted in real time by adopting a dynamic window method. According to the three-dimensional collaborative modeling design method of the building robot, the problem of dynamic environment adaptability is solved through combination of four-dimensional map construction, Nash equilibrium allocation and an improved A * algorithm.

Description

Technical Field

[0001] The present invention belongs to the technical field of construction automation, and particularly relates to a multi-robot collaborative construction planning method and system for construction robots. Background Art

[0002] Existing construction robots mostly adopt an independent operation mode and lack a dynamic collaboration mechanism. For example, Patent CN116822826A proposes a multi-robot collaborative path planning method, but does not consider the impact of dynamic environments (such as temporary obstacles and construction progress changes) on the path during the construction process, resulting in rigid task allocation and frequent manual adjustments required during actual construction. In addition, Patent CN119203335A plans a static path through a BIM model, but does not involve real-time interaction and task reallocation of robots, making it difficult to meet the requirements of cross-construction in multiple process stages.

[0003] Therefore, we propose a multi-robot collaborative construction planning method and system for construction robots to solve the above problems. Summary of the Invention

[0004] The purpose of the present invention is to solve the problems of rigid task allocation and difficulty in adapting to cross-construction of multiple processes in the prior art, and to propose a multi-robot collaborative construction planning method and system for construction robots.

[0005] To achieve the above purpose, the present invention adopts the following technical solutions:

[0006] A multi-robot collaborative construction planning method for construction robots, comprising:

[0007] S1: Collect three-dimensional spatial data of the construction site through lidar and drones, combine the construction progress timestamp, construct a four-dimensional spatio-temporal semantic map, and use a graph neural network to perform cross-domain association on multi-source data in the three-dimensional spatial data;

[0008] S2: According to the construction process logic, construct a process dependency graph G=(V, E), where the vertex V represents the process stage, the edge E represents the process constraint, and generate a feasible construction sequence through topological sorting; establish a task allocation model:

[0009]

[0010] where, U task is the task utility function, which is positively correlated with the construction urgency; U robot is the robot utility function, which is inversely correlated with the power and load; α, β are weight coefficients, which are dynamically adjusted by the gradient descent method; T j is the attribute of the defined jth task, including the workload required for the task, the construction urgency, and the process constraint; S iis the state of the i-th robot, including the remaining battery power, three-dimensional spatial coordinates, and load factor;

[0011] S3: Generate an initial path by improving the A* algorithm, and introduce a risk distance factor into the cost function; at the same time, use the dynamic window method to adjust the path in real time. When there is a path conflict among multiple robots, start distributed voting. Each robot broadcasts its own path priority based on the task urgency, and the robot with a lower priority re-plans the path to avoid the conflict area.

[0012] Preferably, a graph neural network is used to perform cross-domain association on multi-source data in three-dimensional space data, specifically including: mapping the lidar data point cloud into graph nodes, and the node attributes include coordinates (x, y, z) and obstacle type labels;

[0013] Extract the construction progress features from the UAV images through a convolutional neural network and map them as graph edge weights to represent the construction urgency of the area.

[0014] Preferably, in step S1, receive the incremental data of the UAV and lidar at each preset time interval, update the map through the iterative closest point algorithm, label the timeliness of the temporary obstacles, and trigger an alarm if they are not removed after the timeout.

[0015] Preferably, in step S2, define the state of the i-th robot as:

[0016] S i =(E i , P i , L i )

[0017] where E i is the remaining battery power, P i =(x i , y i , z i ) is the three-dimensional spatial coordinates of the robot, and L i is the load factor of the robot = current load / rated load;

[0018] Define the attributes of the j-th task as:

[0019] T j =(D j , γ j , τ k )

[0020] where D j is the workload required for the task, in man-hours; γ j is the construction urgency, ranging from [0, 1], and τ j is the process stage constraint.

[0021] Preferably, in step S2, the probability distribution of robot i selecting task j is defined as p ij , and each robot calculates the expected utility according to the current strategy and adjusts the strategy p ij to maximize its own utility, and repeats the iteration until the strategy change amount Δp ij < ∈, where ∈ is a preset convergence threshold.

[0022] Preferably, in step S2, the weight coefficients α and β are adjusted according to the construction stage. When in the emergency construction period, the weight of α is increased to give priority to completing tasks; when in the normal construction period, the weight of β is increased to extend the battery life of the robot.

[0023] Preferably, when it is detected that the construction progress deviation exceeds the threshold or the robot has an abnormality, task reallocation is triggered; the construction urgency of the lagging area is calculated, the task priority is updated, and the nearest robot is reallocated to perform the emergency task.

[0024] Preferably, in step S3, the initial path is generated by the following method:

[0025] f(n) = g(n) + h(n) + λ·d risk (n)

[0026] where g(n) is the actual cost from the starting point to the current node; h(n) is the heuristic estimated cost from the current node to the end point; d risk (n) is the reciprocal of the Euclidean distance between node n and the nearest obstacle; λ is a dynamic weight, which is adjusted according to the construction urgency, and the obstacle avoidance priority is reduced in case of emergency.

[0027] A multi-robot collaborative construction planning system for construction robots, including

[0028] The data acquisition layer, which collects environmental data in real time through lidar, drones, BIM models, and millimeter-wave radars, is used to fuse BIM models, lidar point clouds, drone aerial photos, and millimeter-wave radar data to construct a four-dimensional spatio-temporal semantic map, and dynamically annotate the positions of obstacles, construction progress tags, and the real-time trajectories of robots;

[0029] The decision-making layer, including a dynamic task allocation module, a process stage identification unit, and a resource scheduling engine; is used to link with the four-dimensional spatio-temporal semantic map, and realize task allocation through the Nash equilibrium game model to ensure the load balance of each robot and the seamless connection of the process stage

[0030] The path planning layer is used to interact with the task allocation module in real time. When construction delays are detected, path replanning is triggered and broadcast to relevant robots;

[0031] The execution layer includes a robot body and a quick-release actuator.

[0032] In summary, the technical effects and advantages of the present invention are as follows: The three-dimensional collaborative modeling design method and system of the construction robot improve the collaborative efficiency of the robots through dynamic task allocation combined with the Nash equilibrium model, shorten the response time to sudden obstacles, reduce the path conflict rate, and improve the obstacle avoidance success rate in complex scenarios by real-time fusing the aerial drone data with the four-dimensional semantic map. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a schematic flowchart in the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0034] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments.

[0035] As shown in the figure, a multi-robot collaborative construction planning method and system for a construction robot includes S1: collecting three-dimensional spatial data of a construction site through a lidar and an aerial drone, constructing a four-dimensional spatio-temporal semantic map in combination with a construction progress timestamp, and performing cross-domain association on multi-source data in the three-dimensional spatial data by using a graph neural network;

[0036] S2: constructing a process dependency graph G=(V, E) according to the construction process logic, where the vertex V represents a process stage, and the edge E represents a process constraint, and generating a feasible construction sequence through topological sorting; establishing a task allocation model:

[0037]

[0038] where, U task is a task utility function, which is positively correlated with the construction urgency; U robot is a robot utility function, which is inversely correlated with the battery power and load; α and β are weight coefficients, which are dynamically adjusted by the gradient descent method; T j is the attribute of the defined j-th task, including the workload required for the task, the construction urgency, and the process constraint; S i is the state of the i-th robot, including the remaining battery power, the three-dimensional spatial coordinates, and the load rate;

[0039] S3: generating an initial path through an improved A* algorithm, introducing a risk distance factor into the cost function; at the same time, using the dynamic window method to adjust the path in real time. When there is a path conflict among multiple robots, start distributed voting, and each robot broadcasts its own path priority based on the task urgency. The robot with a lower priority re-plans the path to avoid the conflict area.

[0040] Embodiment 1

[0041] The scenario of this embodiment is the coordinated construction of wall plastering and tile paving. An RGB-D camera and a lidar are used to construct the point cloud of the construction scene, which is registered with the BIM model through the Iterative Closest Point (ICP) algorithm, and the error is controlled within ±2 cm. The process dependency relationship in the BIM model is analyzed to generate a process stage diagram (such as "structural layer → plastering layer → tile layer") to avoid process conflicts. After registration with the BIM model, the plastering area (high priority) and the tile area (medium priority) are marked.

[0042] Three robots (with battery levels of 80%, 50%, and 75%) are assigned tasks through the Nash equilibrium model, and the load balancing error is <5%.

[0043] The global path cost of robot R1, f(n) = 15.2 + 0.3d risk , bypassing the temporary storage area; there is a path conflict between R2 and R3 at the corner. Through the voting mechanism, it is determined that R2 (task priority 0.9) has the right of way first, and R3 re-plans its path.

[0044] The technical solutions in the embodiments of the present application at least have the following technical effects or advantages: Through dynamic task allocation combined with the Nash equilibrium model, the collaborative efficiency of the robots is improved. By real-time fusing the aerial photography data of the drone through the four-dimensional semantic map, the response time to sudden obstacles is shortened, the path conflict rate is reduced, and the obstacle avoidance success rate in complex scenarios is improved.

[0045] Embodiment 2

[0046] By integrating a millimeter-wave radar and an infrared sensor, high-precision positioning (error < ±3 cm) in rainy and foggy weather is achieved, making up for the perception limitations of the lidar in harsh environments. When the millimeter-wave radar detects rain interference, the positioning error compensation model is activated:

[0047] ΔP = k·ρ rain ·v robot ·Δt

[0048] where k is the compensation coefficient, with the unit of s 2 / mm, calibrated through regression analysis, with the default value of k = 0.02, Δt is the control period, which is set to 0.1 second in this embodiment, and ρ rain is the rainfall density, with the unit of mm / h; v robot is the speed of the robot.

[0049] The corrected position of the robot: p i ’ = p i + Δp, ensuring that the positioning error is < ±3 cm.

[0050] After deployment, k is optimized online through the Kalman filter to adapt to the actual environmental differences.

[0051] Effect verification data:

[0052]

[0053]

[0054] The technical solutions in the embodiments of the present application at least have the following technical effects or advantages: By establishing a physical model of rainfall density - robot speed - positioning error, combining real-time sensor data with a dynamic compensation algorithm, the problem of the decline in the positioning accuracy of construction robots in rainy environments is effectively solved. Compared with traditional methods, the systematic error is greatly reduced, and the universality of the solution is ensured through parameter calibration and an adaptive mechanism, meeting the reliability requirements of complex construction scenarios.

[0055] The embodiments of the present application also provide a method and system for multi-robot collaborative construction planning of construction robots.

[0056] Step 1: System initialization and environment modeling

[0057] 1.1 Data acquisition and fusion

[0058] The lidar scans the construction site to generate a 3D point cloud (resolution 2 cm), and registers it with the BIM model through the ICP algorithm, with the error controlled within ±2 cm.

[0059] The drone takes an aerial photo every 30 seconds, extracts construction progress features (such as the wall completion degree) through a convolutional neural network (CNN), and annotates them to a four-dimensional spatio-temporal semantic map.

[0060] The millimeter-wave radar detects rain and fog weather parameters (such as rainfall density σ rain ), and inputs them into the dynamic compensation model.

[0061] 1.2 Four-dimensional semantic map construction

[0062] Input multi-source data into a graph neural network (GNN). The node attributes include coordinates, obstacle types, and construction urgency; the edge weights represent process dependencies. Provide static process constraints through the BIM model, update the progress labels dynamically by the drone, and correct the geometric accuracy of the map by the lidar.

[0063] Step 2: Dynamic task allocation and process stage identification

[0064] 2.1 Process dependency analysis

[0065] Extract the process logic (such as "structural layer → plaster layer → tile layer") from the BIM model to generate a topological sorting sequence, and prohibit the allocation of tasks that violate the order.

[0066] 2.2 Game theory task allocation

[0067] The decision-making layer calculates the optimal allocation plan according to the robot state S i =(E i , P i , L i ) and the task attribute T j =(D j , γ j , τ j ), and calculates the optimal allocation plan through the Nash equilibrium model. The four-dimensional map provides real-time γ j and P j , corresponding to the construction urgency and construction coordinates respectively.

[0068] Step 3: Adaptive path planning and dynamic obstacle avoidance

[0069] 3.1 Global path generation

[0070] The path planning layer adopts an improved A* algorithm, and the cost function introduces a risk distance factor:

[0071] f(n)=g(n)+h(n)+λ·d risk (n)

[0072] Dynamic weight adjustment: If the construction urgency γ j >0.8, reduce λ and allow approaching obstacles to shorten the path.

[0073] 3.2 Local obstacle avoidance and conflict resolution

[0074] During the movement of the robot, the speed and angular velocity are adjusted in real time through the dynamic window method (DWA):

[0075] Q(v, ω)=0.7·v+0.3·min(d ods )

[0076] When a path conflict is detected, the communication layer starts a distributed voting mechanism, and the robot with a lower priority replans the path.

[0077] Step 4: Communication synchronization and fault tolerance processing

[0078] Each robot broadcasts a status data packet every 0.1 second. The cloud decision center integrates the data and updates the global task queue; if a certain robot goes offline (no response for more than 3 seconds), the adjacent robot takes over its task.

[0079] The technical solutions in the embodiments of the present application at least have the following technical effects or advantages: By combining the four-dimensional semantic map and the distributed game model, the problem of dynamic environment adaptability is solved.

[0080] The working principle is as follows: Through the distance attenuation term, tasks with shorter distances are automatically prioritized to reduce the time spent on movement; the urgency weight ensures that critical tasks are executed first to avoid progress bottlenecks; the robot utility term enables high-load and low-battery robots to automatically select lightweight tasks; dynamically adjusting β can prevent low-battery robots from shutting down prematurely.

[0081] The above are only the preferred specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution of the present invention and its inventive concept, equivalent substitutions or changes should be covered within the protection scope of the present invention.

Claims

1. A multi-robot collaborative construction planning method for construction robots, characterized in that including S1: Collect three-dimensional spatial data of the construction site through lidar and drones, combine the construction progress timestamp, construct a four-dimensional spatio-temporal semantic map, and use graph neural networks to perform cross-domain association on multi-source data in the three-dimensional spatial data; S2: According to the construction process logic, construct a process dependency graph G = (V, E), where the vertex V represents the process stage, and the edge E represents the process constraint. Generate a feasible construction sequence through topological sorting; establish a task allocation model: Among them, U task is the task utility function, which is positively correlated with the construction urgency; U robot is the robot utility function, which is inversely correlated with the power and load; α and β are weight coefficients, which are dynamically adjusted by the gradient descent method; T j is the attribute of the j-th task defined, including the workload required for the task, the construction urgency, and the process constraint; S i is the state of the i-th robot, including the remaining power, the three-dimensional space coordinates, and the load rate; S3: Generate an initial path through an improved A* algorithm, introduce a risk distance factor into the cost function; at the same time, use the dynamic window method to adjust the path in real time. When there is a multi-robot path conflict, start distributed voting, and each robot broadcasts its own path priority based on the task urgency. The robot with a lower priority re-plans the path to avoid the conflict area.

2. The multi-robot collaborative construction planning method for construction robots according to claim 1, wherein, The cross-domain association of multi-source data in the three-dimensional spatial data by using graph neural networks specifically includes: mapping the lidar data point cloud into graph nodes, and the node attributes include coordinates (x, y, z) and obstacle type labels; Extract the construction progress features from the drone images through a convolutional neural network and map them to the graph edge weights to represent the construction urgency of the area.

3. The multi-robot collaborative construction planning method for construction robots according to claim 1, wherein In step S1, receive the incremental data of the drone and lidar once every preset time interval, update the map through the iterative closest point algorithm, label the timeliness label for the temporary obstacle, and trigger an alarm if it is not removed after the timeout.

4. The multi-robot collaborative construction planning method for construction robots according to claim 1, wherein, In step S2, define the state of the i-th robot as: S i = (E i , P i , L i ) Among them, E i is the remaining power, P i =(x i , y i , z i ) is the three-dimensional space coordinate of the robot, L i is the load rate of the robot = current load / rated load; Define the attributes of the j-th task as: T j = (D j , γ j , τ j ) Among them, D j is the workload required for the task, with the unit of man-hours; γ j is the construction urgency, ranging from [0, 1], and τ j is the process stage constraint.

5. The multi-robot collaborative construction planning method for construction robots according to claim 1, characterized in that, In step S2, the probability distribution of robot i selecting task j is defined as p ij , and each robot calculates the expected utility according to the current strategy and adjusts the strategy p ij to maximize its own utility, and repeats the iteration until the strategy change amount Δp ij < ∈, where ∈ is a preset convergence threshold.

6. The method for multi-robot collaborative construction planning of a construction robot according to claim 1, characterized in that In step S2, adjust the weight coefficients α and β according to the construction stage. When in the emergency construction period, increase the α weight to give priority to completing tasks; when in the normal construction period, increase the β weight to extend the robot's battery life.

7. The method for multi-robot collaborative construction planning of a construction robot according to claim 1, characterized in that When it is detected that the construction progress deviation exceeds the threshold or the robot appears abnormal, trigger task reallocation; calculate the construction urgency of the lagging area, update the task priority, and reallocate the nearest robot to perform the emergency task.

8. The method for multi-robot collaborative construction planning of a construction robot according to claim 1, wherein In step S3, generate the initial path through the following method: f(n) = g(n) + h(n) + λ·d risk (n) Among them, g(n) is the actual cost from the starting point to the current node; h(n) is the heuristic estimated cost from the current node to the end point; d risk (n) is the reciprocal of the Euclidean distance between node n and the nearest obstacle; λ is a dynamic weight, which is adjusted according to the construction urgency. When it is urgent, the obstacle avoidance priority is reduced.

9. A multi-robot collaborative construction planning system for construction robots, characterized in that The data acquisition layer collects environmental data in real time through lidar, drones, BIM models, and millimeter-wave radars, and is used to fuse BIM model, lidar point cloud, drone aerial photography, and millimeter-wave radar data to construct a four-dimensional spatio-temporal semantic map, and dynamically label the obstacle positions, construction progress labels, and real-time trajectories of robots; The decision-making layer includes a dynamic task allocation module, a process stage recognition unit, and a resource scheduling engine; It is used to link with the four-dimensional spatio-temporal semantic map, and realize task allocation through the Nash equilibrium game model to ensure the load balance of each robot and the seamless connection of the process stage The path planning layer is used to interact with the task allocation module in real time. When construction delay is detected, trigger path replanning and broadcast it to relevant robots; The execution layer includes a robot body and a quick-release actuator.

Images