A multi-robot cooperative measurement method and system based on crowd wisdom emergence

Abstract

The application discloses a kind of multi-robot cooperative measurement method and system based on crowd intelligence emergence, to solve the problems of single point failure, poor environmental adaptability and insufficient scalability in existing centralized, pre-planned multi-robot cooperative measurement method.The application builds a shared digital environment maintained by all robots through decentralized cooperation, introduces a decentralized task decision mechanism based on digital pheromone, and regards each measurement robot as an autonomous agent, independently makes measurement target decisions by sensing the digital pheromone concentration in the shared digital environment determined by measurement data quality dynamically.The application enables the robot cluster of multi-robot cooperative measurement to spontaneously emerge globally optimal cooperative measurement behavior without a central control node, significantly improving the robustness, adaptability and scalability of the measurement system, and achieving efficient and reliable three-dimensional measurement of large components.

Description

Technical Field

[0001] This invention discloses a multi-robot collaborative measurement method and system based on swarm intelligence emergence, belonging to the field of robot intelligent control technology. Background Technology

[0002] In the high-end manufacturing of large components, utilizing multiple mobile robots to perform collaborative 3D measurement of these components has become a key technological approach for achieving automated and high-precision quality control. Currently, the dominant technological paradigm in the field of multi-robot collaborative measurement technology is based on a centralized, top-down command and control architecture.

[0003] In this architecture, a powerful central server acts as the "brain" of the system, responsible for global planning, task allocation, and data fusion. For example, Chinese patent application CN120599564A discloses a centralized multi-robot collaborative SLAM method based on an FPGA platform. This method inputs color images into the FPGA platform of individual robots, converts them into grayscale images, performs image pyramid scaling, corner detection, and feature description, and transmits the data to the individual robot's processing system. The processing system generates observation information, which in turn generates image frames. Keyframes are generated and sent to the central server. The central server performs loop closure detection on the keyframes, merges the local maps established by different individual robots to obtain a global map, and optimizes the global map; finally, it constructs a 3D dense point cloud map. The workflow typically involves pre-dividing static regions based on the initial model of the components, assigning fixed scanning tasks and paths to each robot, and then performing unified offline data processing by the server after all robots have completed data acquisition. However, this traditional architecture, which heavily relies on a central node, suffers from inherent defects that become a fundamental bottleneck restricting the improvement of system performance and reliability when applied to complex and ever-changing real-world measurement scenarios. This architecture not only has the risk of a single point of failure due to its reliance on a central server, but its static, pre-planned operating mode also makes it inflexible and unable to cope with dynamic uncertainties in the field.

[0004] Based on the aforementioned architecture that relies on a central server to control all robots, the theory of swarm intelligence emergence, derived from the study of collective biological behavior in nature, provides a novel paradigm for robot swarm collaboration. The core idea of ​​swarm intelligence emergence is that a system composed of multiple autonomous intelligent agents, without central coordination, can spontaneously and from the bottom up exhibit complex, intelligent global behaviors that transcend individual capabilities through local interactions and environmental perception. However, unlike tasks such as target search and environmental exploration, industrial measurement requires deterministic, quantifiable, and comprehensive high-precision data acquisition. Directly applying swarm intelligence emergence theory to high-precision industrial measurement still faces a significant technological gap. Therefore, designing a novel multi-robot collaborative measurement method that fully utilizes the adaptive and robust advantages of swarm intelligence emergence while ensuring that the final measurement results meet industrial-grade accuracy and completeness requirements has become a pressing technical challenge in this field. Summary of the Invention

[0005] The technical problem solved by this invention is to provide a multi-robot collaborative measurement method and system based on swarm intelligence emergence, which addresses the problems of single point of failure, poor environmental adaptability and insufficient scalability in existing centralized and pre-planned multi-robot collaborative measurement methods.

[0006] This invention is achieved using the following technical solution:

[0007] This invention first discloses a multi-robot collaborative measurement method based on swarm intelligence emergence, comprising the following steps:

[0008] S1. All measuring robots in the multi-robot system jointly construct and maintain a shared digital environment containing the geometric information layer of the measured component in real time by running the C-SLAM algorithm. In the shared digital environment, attracting pheromones are seeded on the geometric information layer based on the standard three-dimensional model of the measured component to construct a digital pheromone layer.

[0009] S2. Each measuring robot independently senses the distribution state of the attracting pheromones in the digital pheromone layer, and obtains its own local environmental information and motion state. It uses a probabilistic decision model to select its own target measurement area from the geometric information layer of the component being measured, and combines the states of neighboring measuring robots to plan a travel path within its own target measurement area using cluster rules.

[0010] S3. The measuring robots perform measuring actions according to the planned travel path within their respective target measuring areas, collect new measuring data of the measured components, integrate the new measuring data into the geometric information layer of the measured components in the shared digital environment, and dynamically update the digital pheromone layer.

[0011] S4. Continue executing steps S2 and S3. During the process, the global concentration of attracting pheromones in the shared digital environment is detected in real time. When the global concentration of attracting pheromones is lower than the preset termination threshold, the measurement process is terminated, and the measurement three-dimensional model of the measured component is output from the geometric information layer of the final fused new measurement data in the shared digital environment.

[0012] In a multi-robot collaborative measurement method based on swarm intelligence emergence of the present invention, the digital pheromone layer is a spatially aligned dynamic data field established on the geometric information layer of the component under test. Each geometric coordinate of the geometric information layer is stored as an initial numerical data as a digital pheromone representing the measurement uncertainty of the corresponding coordinate area. The digital pheromone includes an attraction pheromone for attracting the robot to the measurement uncertainty area and a repulsion pheromone for driving the robot away from the completed measurement area.

[0013] In a multi-robot collaborative measurement method based on swarm intelligence emergence of the present invention, further, in step S1, the concentration of the attracting pheromone is positively correlated with the measurement uncertainty of the corresponding region.

[0014] In a multi-robot collaborative measurement method based on swarm intelligence emergence of the present invention, step S2 further includes the following sub-steps:

[0015] S21. The measuring robot independently queries the shared digital environment to obtain the pheromone distribution state required for its decision-making. The pheromone distribution is then used to calculate the probability of the measuring robot going to each potential measuring area through a probabilistic decision model.

[0016] ,

[0017] in, robot exist Choose to go to the potential measurement area at any time The probability, For potential measurement areas exist The concentration of attraction pheromones at any given moment For robot k to reach the potential measurement area The heuristic information is the reciprocal of the distance from the robot to the potential measurement area. For pheromone weighting factors, As a heuristic information weighting factor, Given the set of all available potential measurement regions for robot k, the measurement robot selects the target measurement region from the set of potential measurement regions based on the probability of each potential measurement region.

[0018] S22. Obtain the motion state of the measurement robot in the global coordinate system of the shared digital environment, and calculate the velocity vector of the measurement robot at the next moment through cluster rules to plan the travel path. The velocity vector of the travel path... From cohesion vector Separation vector and alignment vector Sure:

[0019] ,

[0020] in, To measure robot k in The final velocity vector of the motion state at any given moment. It is the cohesive vector, and its direction points to the target measurement area selected in step S21. This is the separation vector, and its direction is opposite to that of the neighboring measuring robot. The alignment vector is the average motion direction of the neighboring measurement robots. , , These are the weight coefficients for their respective vectors;

[0021] S23. Using the velocity vector obtained in step S22 as the kinematic constraint of the measurement robot, and combining the local environmental information of the measurement robot, plan the travel path of the measurement robot in the target measurement area.

[0022] In a multi-robot collaborative measurement method based on swarm intelligence emergence of the present invention, further, in step S21, the set of all currently selectable potential measurement regions for robot k includes the following two subsets:

[0023] A local potential measurement area set, wherein the measurement robot queries the shared digital environment for all potential measurement areas near its location with an attractive pheromone concentration greater than zero;

[0024] A global priority measurement area set is defined, in which the measurement robot queries the top N potential measurement areas with the highest pheromone concentration from the global scope of the shared digital environment, where N is a preset value.

[0025] In a multi-robot collaborative measurement method based on swarm intelligence emergence of the present invention, further, in step S22, the cohesion vector It can be expressed by the following formula:

[0026] ,

[0027] in, The center point of the target measurement area, To measure the precise pose of robot k in the global coordinate system at time t;

[0028] The separation vector It is expressed by the following formula:

[0029] ,

[0030] in, This represents the precise pose of the neighboring measurement robot j in the global coordinate system at time t. Let k be the set of neighboring measuring robots;

[0031] The alignment vector It can be expressed by the following formula:

[0032] ,

[0033] in, To measure the number of neighboring measurement robots of robot k, Let t represent the motion state of the neighboring measurement robot j in the global coordinate system at time t.

[0034] In a multi-robot collaborative measurement method based on swarm intelligence emergence of the present invention, further, in step S3, the measurement robot moves along a planned path to the selected target measurement area, adjusts its scanner posture to scan the target measurement area, collects new measurement data of the component under test, and fuses the new measurement data into the geometric information layer of the shared digital environment through pose transformation of the C-SLAM framework. Simultaneously, the quality of the new measurement data collected within the target measurement area is evaluated, and the digital pheromone layer is dynamically updated based on the quality evaluation results.

[0035] When the quality assessment result of the new measurement data of the target measurement area by the measurement robot is not up to standard, the concentration of attracting pheromones in that area is increased.

[0036] When the measurement robot's assessment of the quality of new measurement data for the target measurement area meets the standards, the attractive pheromone for that area is updated to a repulsive pheromone.

[0037] In a multi-robot collaborative measurement method based on swarm intelligence emergence according to the present invention, further, in step S3, the quality of new measurement data in the target measurement area is evaluated by the point cloud density of the measurement data:

[0038] If the point cloud density of the measurement data in the target measurement area is less than the target density threshold, the quality assessment result will not meet the standard.

[0039] When the point cloud density of the measurement data in the target measurement area reaches the target density threshold, the quality assessment result meets the standard.

[0040] In a multi-robot collaborative measurement method based on swarm intelligence emergence of the present invention, further, in step S4, steps S2 and S3 are continuously repeated in all measurement robot clusters of the multi-robot system to monitor in real time the sum of the attracted pheromone concentrations within the selectable potential measurement areas of all measurement robots in the shared digital environment, and compare it with a preset termination threshold. The measurement task ends when the following termination conditions are met:

[0041] ,

[0042] in, For potential measurement areas exist The concentration of attracting pheromones at any given time, and allvoxels representing the number of potential measurement areas for all measuring robots. It is a preset global termination threshold. After the measurement task ends, it automatically terminates the measurement activities of all measurement robots and exports the final geometric information layer model of the measured component and the measurement data from the shared digital environment as the final measurement 3D model output of the measured component.

[0043] The present invention also discloses a multi-robot collaborative measurement system based on swarm intelligence emergence, comprising a multi-robot system composed of two or more detection robots, wherein the multi-robot system measures the component under test using the multi-robot collaborative measurement method based on swarm intelligence emergence described above.

[0044] The present invention, by adopting the above technical solution, has the following beneficial effects:

[0045] (1) This invention eliminates the risk of single point of failure through a decentralized architecture, abolishes the central control node, and decentralizes intelligence to each individual measurement robot, avoiding robot failures caused by failures of the central control terminal. Furthermore, it achieves automatic task takeover after a single robot fails through a pheromone mechanism. When a single measurement robot fails, the attracting pheromone corresponding to its unfinished measurement task continues to exist in the shared digital environment, automatically attracting other healthy measurement robots to take over, thus realizing the fault tolerance capability of a single robot and possessing strong robustness and fault tolerance.

[0046] (2) This invention utilizes a dynamic digital pheromone mechanism, enabling a robot swarm to adaptively focus measurement resources in real time on complex or incomplete areas with high attractant pheromone concentrations. The attractant pheromone in this invention is stored as numerical data greater than 0 on each grid cell of the geometric information layer of the measured component in the shared digital environment. Initially, the attractant pheromone concentration is the same in all incomplete areas. Then, the measuring robot measures a certain area. If the measurement data of the area is found to be substandard after quality assessment, the attractant pheromone concentration of the area will be increased. The higher the pheromone concentration, the more complex the area, attracting more measuring robots to measure the area. Each robot, as an autonomous intelligent agent, independently makes measurement target decisions by sensing the attractant pheromone concentration and repulsive pheromone in the shared environment, which are dynamically determined by the quality of the measurement data, without human intervention.

[0047] (3) The present invention uses the local quantity region set and the global priority measurement region set to make probabilistic decisions on the target measurement region of the measurement robot, so as to realize the full coverage of the surface region of the measured component by the measurement robot and effectively improve the integrity and consistency of the surface detection of the measured component.

[0048] (4) The present invention adopts an indirect communication method based on a shared environment. All measurement robots jointly construct and maintain a shared digital environment in real time, which includes the geometric information layer and digital pheromone layer of the measured component. The robot cluster does not need a central control node. It can spontaneously generate globally optimal collaborative measurement behavior by interacting with the environment by following unified local rules. This reduces the communication bandwidth pressure from the measurement robot to the central processor and back to the measurement robot, eliminates the central bottleneck, and has excellent ability to scale to a large-scale cluster.

[0049] (5) This invention achieves real-time fusion of measurement data of the measured component through the framework of decentralized collaborative real-time localization and mapping (C-SLAM), which fundamentally avoids the problem of offline data stitching and ensures the global consistency and high accuracy of the final three-dimensional model.

[0050] In summary, this invention constructs a shared digital environment that all measurement robots can access and modify, and introduces an indirect communication and incentive mechanism using digital pheromones. This allows efficient and optimal global measurement strategies for the measured components to emerge spontaneously from the local interactions of multiple measurement robot groups. It features decentralization and self-organization, significantly improving the robustness, adaptability, and scalability of the measurement system, and achieving efficient and reliable 3D measurement of large components.

[0051] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. Attached Figure Description

[0052] Figure 1 This is a flowchart illustrating a multi-robot collaborative measurement method based on swarm intelligence emergence according to the present invention. Detailed Implementation

[0053] Example

[0054] See Figure 1 The figure shows a flowchart of a multi-robot collaborative measurement method based on swarm intelligence emergence according to the present invention, which specifically includes the following steps:

[0055] S1. All measurement robots in the multi-robot system jointly construct and maintain a shared digital environment containing the geometric information layer of the measured component in real time by running the C-SLAM algorithm. In the shared digital environment, attracting pheromones are seeded on the geometric information layer based on the standard three-dimensional model of the measured component to construct a digital pheromone layer.

[0056] The specific process is as follows:

[0057] S11. The measuring robots jointly construct a 3D point cloud or mesh model of the measured object as a geometric information layer of the shared digital environment through a decentralized collaborative real-time localization and mapping algorithm, namely the C-SLAM algorithm. All measuring robots execute the C-SLAM algorithm throughout the entire measurement process to maintain the geometric information layer of the shared digital environment. The basic concepts and implementation methods of the C-SLAM algorithm are well-known in the field of multi-robot technology, and will not be elaborated here.

[0058] S12. A spatially aligned dynamic data field is established on the geometric information layer as the digital pheromone layer for storing and updating digital pheromones. The digital pheromone layer is distributed with attracting pheromones to attract the robot to the measurement uncertainty area and repulsive pheromones to drive the robot away from the measurement completed area.

[0059] The geometric information layer is a three-dimensional grid map, with each grid cell having a unique coordinate. The digital pheromone layer establishes a spatially aligned dynamic data field on the geometric information layer of the component under test, which can be implemented using data structures such as hash tables. Both layers use the same set of spatial coordinates and indices for alignment. This data structure maps each grid cell in the geometric information layer to a stored pheromone value. Inputting a grid position allows the corresponding pheromone magnitude to be retrieved through the mapping. Each geometric coordinate in the geometric information layer is associated with an initial numerical value, which serves as a digital pheromone representing the measurement uncertainty of the corresponding coordinate region. This digital pheromone includes an attracting pheromone to attract the robot to the measurement uncertainty area and a repulsive pheromone to drive the robot away from the completed measurement area. In this invention, both the attracting and repulsive pheromones are represented by specific numerical values; a positive value and a larger value indicate a higher concentration of attracting pheromones, while a value of 0 indicates a repulsive pheromone. In the initially constructed digital information layer, the concentration of attracting pheromones is positively correlated with the measurement uncertainty of the corresponding region of the component under test; that is, regions with higher uncertainty receive higher concentrations of attracting pheromones.

[0060] The geometric information layer obtained by C-SLAM contains a dynamic 3D map of the component under test, including the initial outline of the component under test and information about the surrounding environment. In order to accurately obtain the unmeasured area of ​​the component under test, a digital pheromone layer is constructed by seeding the geometric information layer with attracting pheromones based on the standard 3D model of the component under test. The standard 3D model is the CAD model of the component under test.

[0061] The geometric information layer of the shared digital environment stores information using 3D point clouds, representing the geometric information of objects in the environment in the form of point clouds. Digital pheromones are one of the attributes of points in the 3D point cloud, and the pheromone layer is an abstraction of the digital information of all point clouds. The shared digital environment uses distributed storage; each measurement robot stores a copy of the shared digital environment. When one robot updates its shared digital environment information, all measurement robots synchronously update their stored shared digital environment information through network broadcasting of the multi-robot system.

[0062] S2. Each measuring robot independently senses the distribution state of the attracting pheromones in the digital pheromone layer, and obtains its own local environmental information and motion state. It uses a probabilistic decision model to select its own target measurement area from the geometric information layer of the component being measured, and combines the states of neighboring measuring robots to plan a travel path within its own target measurement area using cluster rules.

[0063] Step S2 specifically includes the following sub-steps:

[0064] S21. The measuring robot independently queries the shared digital environment to obtain the pheromone distribution state required for its decision-making. It calculates the probability of the measuring robot proceeding to each potential measuring area using a probabilistic decision model based on the pheromone distribution. The measuring robot then selects a target measuring area from the potential measuring area set based on the probability of each potential measuring area. The potential measuring area set includes two subsets: a local potential measuring area set and a global priority measuring area set. The local potential measuring area set consists of all areas near the measuring robot's location with a pheromone concentration greater than zero, as queried from the shared digital environment. The global priority measuring area set consists of the top N potential measuring areas with the highest pheromone concentration, queried from the global scope of the shared digital environment, where N is a preset value. The local potential measuring area set ensures high efficiency of the measuring robot in the local measurement process, while the global priority measuring area set guarantees the completeness of the global measurement. The final measurement will only terminate when the measurement results of all areas on the surface of the component being measured meet the standards. If any area fails to meet the standards, the total pheromone concentration will be greater than 0, and the robot will continue to measure areas with pheromone concentrations greater than 0 until all areas have been measured.

[0065] S22. Obtain the motion state of the measurement robot in the shared digital environment's global coordinate system. The motion state includes the motion parameters of each joint of the robot at the previous moment and the position and orientation of the measurement robot in its own three-dimensional space. Calculate the velocity vector of the measurement robot at the next moment using cluster rules to plan the travel path. The global coordinate system is a unified reference coordinate system of the shared digital environment jointly maintained by all measurement robots in step S1. The origin of the coordinate system can be selected from some salient features in the environment that are not easily changed, or the pose of the first robot when it starts C-SLAM can be used as the origin.

[0066] S23. Using the velocity vector obtained in step S22 as the kinematic constraint of the measurement robot, and combining it with the local environmental information of the measurement robot, the robot's path in the target measurement area is planned. Local environmental information refers to the environmental features perceived in real time by the measurement robot's sensors, such as other nearby robots or temporary obstacles not modeled in the global map. The robot uses these environmental features for specific motion planning and obstacle avoidance. The smooth path of the measurement robot is calculated by the motion planning module. The velocity vector at the next moment serves as the kinematic constraint for the robot's desired velocity and direction. This instruction is input into the motion planning module, which combines the robot's own kinematic constraints, the surrounding environment, and known algorithms such as A* or RRT* within the module to calculate a collision-free and smooth path. The specific robot path planning algorithm is a known prior art and will not be elaborated upon in this embodiment.

[0067] S3. The measuring robot performs measurement actions according to the planned travel path within its respective target measurement area, collects new measurement data of the component under test, and integrates the new measurement data into the geometric information layer of the shared digital environment. Then, based on the quality assessment results of the new measurement data of the component under test, the digital pheromone layer is dynamically updated, including increasing the attractant pheromone concentration in the area where the assessment is substandard, decreasing the attractant pheromone concentration in the area where the assessment is satisfactory, or disseminating repulsive pheromones.

[0068] The measurement robot moves along a planned path to the selected target measurement area. It adjusts its scanner posture to scan the component under test within the target area, collecting new measurement data. This new data is then fused into the geometric information layer of the shared digital environment through pose transformation within the C-SLAM framework. The digital pheromone layer is dynamically updated based on the quality assessment of the new measurement data. If the measurement robot's assessment of the measurement data quality for the target area is substandard, the concentration of attracting pheromones in that area is increased; conversely, if the assessment is satisfactory, the concentration of attracting pheromones in that area is decreased while the concentration of repulsive pheromones is increased.

[0069] Specifically, the quality of new measurement data of the measured component can be evaluated by measuring the point cloud density of the measurement data in the target measurement area. When the point cloud density of the measurement data in the target measurement area is less than the target density threshold, the measurement data quality evaluation result is unsatisfactory. When the point cloud density of the measurement data in the target measurement area reaches the target density threshold, the measurement data quality evaluation result is satisfactory.

[0070] This invention evaluates data quality by calculating the point cloud density in new measurement data of the target measurement area. A higher point cloud density indicates higher measurement data quality, meaning the point cloud data collected in that area is sufficient to represent the real model and no further robot measurement is needed. In this case, the attracting pheromone in that area is updated to a repulsive pheromone to prevent the robot from measuring again. If the quality is insufficient, it means that the area still cannot meet the requirements after measurement. Therefore, the attracting pheromone concentration in that area is increased to attract the current robot or surrounding robots to supplement the measurement.

[0071] S4. Continue executing steps S2 and S3. During the process, the global concentration of attracting pheromones in the shared digital environment is detected in real time. When the global concentration of attracting pheromones is lower than the preset termination threshold, the measurement process is terminated, and the measurement three-dimensional model of the measured component is output from the geometric information layer of the final fused new measurement data in the shared digital environment.

[0072] Steps S2 and S3 are continuously repeated across all measurement robot clusters in the multi-robot system. The sum of the attracted pheromone concentrations within all potential detection areas of the measurement robots in the shared digital environment is monitored in real time and compared with a preset termination threshold. The measurement task ends when the following termination conditions are met:

[0073] ;

[0074] in, Potential detection area exist The concentration of attracting pheromones at any given time, and allvoxels represents the number of potential detection areas for all measuring robots. It is a preset global termination threshold, which is a parameter that is dynamically calculated or pre-set according to the preset measurement accuracy requirements or total coverage requirements. After the measurement task ends, it automatically terminates the measurement activities of all measurement robots and exports the final 3D model and measurement data from the shared digital environment as the final measurement output of the measured component.

[0075] The geometric information layer within the shared digital environment is continuously maintained and updated by all measurement robots. Initially, the measurement data in the geometric information layer is relatively coarse. As the measurement robots conduct collaborative measurements, new measurement data is incorporated, and the data in the geometric information layer is gradually improved. The accuracy of the measurement data of the measured component gradually increases. Once the measurement is completed, the model data of the measured component in the geometric information layer can meet the requirements. The model data is then separated and processed through noise reduction and other methods to obtain the final measurement output 3D model of the measured component.

[0076] To enable those skilled in the art to better understand the technical solution of this invention, the following detailed description of the working process of a specific multi-robot collaborative measurement system further clarifies and completes the technical solution of this invention. Obviously, the embodiments described below are merely some embodiments of this invention, and not all embodiments.

[0077] To perform comprehensive, high-precision 3D digital measurement of a large component, a multi-robot system consisting of 10 identical mobile measurement robots is deployed. Each mobile measurement robot in this robot cluster is equipped with a high-precision 3D laser scanner and has the ability to communicate wirelessly via a mobile ad hoc network (MANET). The present invention implements a multi-robot collaborative measurement method based on swarm intelligence emergence according to the following specific steps.

[0078] S1. At the start of the measurement task, 10 measurement robots jointly execute the decentralized collaborative instant localization and mapping algorithm (C-SLAM) to jointly build and maintain a shared digital environment that includes the geometric information layer and digital pheromone layer of the measured component in real time.

[0079] In sub-step S11, specifically, all measuring robots perform distributed backend optimization by exchanging keyframe and pose map information, thereby fusing their respective local scan data and constructing a high-precision 3D grid point cloud map with unified global coordinates, serving as the initial geometric information layer of the shared digital environment. In this embodiment, the size of the grid cell (Voxel) of this geometric information layer is set to 5cm x 5cm x 5cm.

[0080] In sub-step S11, a spatially strictly aligned digital pheromone layer is established above the geometric information layer. This digital pheromone layer can be implemented as an efficient octree, mapping the three-dimensional index of each grid cell in the three-dimensional grid point cloud map of the geometric information layer to a numerical structure that stores the concentrations of attracting and repelling pheromones.

[0081] Based on the standard CAD 3D model of the component under test, an initial seeding of attracting pheromones is performed in the digital pheromone layer of the shared digital environment. This process involves dense point sampling on the surface of the standard CAD model of the component under test, and assigning the initial attracting pheromone concentration of the corresponding grid cell in the geometric information layer to the location of each sampling point. This allows for the pre-establishment of a strong signal representing "incomplete measurement" on the surface of the component under test across the entire geometric information layer.

[0082] S2. At any point in the measurement process, each measurement robot in the multi-robot system independently and asynchronously executes a cycle of perceiving the environment and its state. Taking one of the measurement robots, k, as an example, at the beginning of its decision cycle, it queries the shared digital environment to obtain the distribution of digital pheromones within its perception range, mainly focusing on the concentration of attracting pheromones, and obtains its precise pose in the global coordinate system of the 3D grid point cloud map. and motion state Then, based on the comprehensive information obtained above, the measurement robot k makes autonomous decisions and plans its path to the target measurement area.

[0083] In sub-step S21, firstly, a probabilistic decision model is used to select the next target measurement area for the measurement robot k. The parameters of the probabilistic decision model are set as follows: pheromone weight factor. Heuristic information weighting factor The probability that the robot will choose to go to the target measurement area. Determined by the following formula:

[0084] ;

[0085] in, For potential measurement areas exist The concentration of attraction pheromones at any given moment To measure robot k to the potential measurement area The heuristic information is typically the reciprocal of the distance from the measuring robot to the potential measurement area. For pheromone weighting factors, As a heuristic information weighting factor, Let k be the set of all available potential measurement regions for robot k. The measurement robot selects the target measurement region from the set of potential measurement regions based on the probability of each potential measurement region.

[0086] Specifically, probability This represents the probability that the potential measurement area will be selected by the measurement robot. The measurement robot randomly selects the target measurement area from the set of potential measurement areas using randomly generated numbers. A potential measurement area with a higher probability corresponds to a larger range of random numbers, thus increasing the likelihood that the robot will select it as the target measurement area. Assuming there are three potential measurement areas, their probabilities of selection are calculated to be 0.7, 0.2, and 0.1, respectively. The measurement robot randomly generates an integer from 0 to 9. If the number is 0 to 6, potential measurement area 1 is selected; if the number is 7 or 8, potential measurement area 2 is selected; and if the number is 9, potential measurement area 3 is selected. Clearly, potential measurement area 1 is more likely to be selected, while potential measurement areas 2 and 3 also have a probability of being selected. There are many other ways to implement area selection, which are well-known technologies and will not be elaborated upon in this embodiment.

[0087] Potential measurement area set High-attractant-concentration potential measurement areas within the global scope of the shared digital environment are also included in the selectable potential measurement areas. The probability of a measurement robot selecting a potential measurement area is also related to the distance between the measurement robot and the potential measurement area. For example, if a potential measurement area has a high attractant concentration but is very far from the robot, then the probability of that potential measurement area being selected by the measurement robot is very small. If multiple measurement robots select the same potential measurement area, then the allocation is based on the motion cost of the measurement robots moving to that potential measurement area, with priority given to measurement robots with lower motion costs. The movement distance from the robot to the target area can be used as the motion cost; a shorter movement distance naturally results in a lower motion cost.

[0088] The potential measurement region set comprises two subsets: a local potential measurement region set and a global priority measurement region set. The local potential measurement region set consists of all regions near the measurement robot's location with a pheromone concentration greater than zero, queried from the shared digital environment. The global priority measurement region set consists of the top N potential measurement regions with the highest pheromone concentrations, queried from the global scope of the shared digital environment, where N is a preset value. The local potential measurement region set ensures high efficiency of the measurement robot in the local measurement process, while the global priority measurement region set guarantees the completeness of the global measurement. The final measurement will only terminate when the measurement results of all regions on the surface of the component being measured meet the standards. If any region fails to meet the standards, the total pheromone concentration will be greater than 0, and the robot will continue to measure regions with pheromone concentrations greater than 0 until all regions have been measured.

[0089] In sub-step S22, the measurement robot uses swarm rules to calculate its velocity vector for the next moment to plan its travel path, where the velocity vector... From cohesion vector Separation vector and alignment vector Determined by weighted average:

[0090] .

[0091] in, To measure robot k in The final velocity vector of the motion state at any given moment. This is the cohesive vector, and its direction points towards the target measurement area selected in step S31. This is a separation vector, its direction is opposite to that of neighboring measurement robots, used to avoid other measurement robots to prevent physical collisions and sensor interference. The alignment vector is the average motion direction of the neighboring measurement robots, used to make all the detection robots in a multi-robot system form a cooperative array with the same motion direction. , , These are the weight coefficients for each vector, used to dynamically balance the robot's behavioral preferences. The weight ratios are adjusted based on engineering experience and on-site debugging. In this embodiment, the weight coefficients for each vector are set as follows: cohesive vector weights... Separate vector weights Alignment vector weights .

[0092] Specifically, cohesion vector The direction points towards the center point of the target measurement area. The final velocity vector is primarily driven by the following equation:

[0093] ;

[0094] in, To measure the center point of the target measurement area for robot k, To measure the precise pose of robot k in the global coordinate system at time t.

[0095] Separation vector To avoid collisions during mobile measurement, it calculates the distance between the robot and all neighboring measurement robots. The repulsive force between them is expressed by the following formula:

[0096] .

[0097] in, This represents the precise pose of the neighboring measurement robot j in the global coordinate system at time t. Let k be the set of neighboring measuring robots.

[0098] Alignment vector This is used to form a cooperative array with neighboring measurement robots, achieved by calculating the average motion vector of the neighboring measurement robots, as expressed by the following formula:

[0099] .

[0100] in, To measure the number of neighboring measurement robots of robot k, Let t represent the motion state of the neighboring measurement robot j in the global coordinate system at time t.

[0101] S23. Using the velocity vector obtained in step S22 as the kinematic constraint of the measurement robot, and combining it with the local environmental information of the measurement robot, the RRT* algorithm is used to calculate a collision-free and smooth trajectory path of the measurement robot in the target measurement area.

[0102] S3, Measurement Robot The underlying motion controller is based on the calculated velocity vector Drive measurement robot The robot moves along the planned path towards the target measurement area. Upon arrival, the robot adjusts its scanner posture to the optimal scanning angle and triggers the scanner to perform a high-precision 3D scan of the measured component within the target measurement area, acquiring a frame of point cloud data. After the measurement operation is completed, the robot immediately fuses the newly acquired point cloud data into a globally unified high-precision 3D raster point cloud map in the shared digital environment through pose transformation within the C-SLAM framework. Based on the quality assessment results of the new data, the digital pheromone layer is dynamically updated.

[0103] Specifically, the measurement robot k measures the enhanced region after the new measurement data has been integrated. Perform a rapid data quality assessment and calculate the average point cloud density of all raster cells within the region. .

[0104] Based on the point cloud density assessment of the new measurement data, the measurement robot will dynamically update the digital pheromone concentration corresponding to the target measurement area. The pheromone concentration is superimposed with a dynamically updated increment of the digital pheromone. This increment The calculation method is as follows: a piecewise function.

[0105] ;

[0106] in, It is the target density threshold of the data point cloud. It is a positive enhancement coefficient. It is a negative penalty coefficient. If the point cloud density of the new measurement data does not meet the standard, a positive enhancement is given to the attractive pheromone of the region, attracting the measurement robot to perform more measurements in the region; if it has met the standard, a large negative penalty is given, updating the attractive pheromone of the region to 0, turning it into a repulsive pheromone, and repelling the measurement robot from measuring the region again.

[0107] S4. The entire cluster of measurement robots in the multi-robot system continuously and asynchronously repeats the perception, decision-making, action, and update cycle described in S2 to S3. Simultaneously, the total pheromone concentration attracted by all potential measurement areas within the shared digital environment is monitored in real time. The task terminates when the following termination condition is met:

[0108] ;

[0109] in, For potential measurement areas exist The concentration of attracting pheromones at any given time, and allvoxels representing the number of potential measurement areas for all measuring robots. It is a preset global termination threshold, which is a parameter that is dynamically calculated or preset according to the preset measurement accuracy requirements or total coverage requirements. After the measurement task ends, it automatically terminates the measurement activities of all measurement robots and exports the final 3D model and measurement data from the shared digital environment. After denoising and meshing, it is used as the final measurement output of the measured component.

[0110] The above embodiments describe the basic principles and main features of the present invention, as well as its advantages. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the specific working principles of the present invention. Various changes and modifications can be made to the present invention without departing from its spirit and scope. All such changes and modifications fall within the scope of the present invention as claimed, which is defined by the appended claims and their equivalents.

Claims

1. A method for swarm intelligence emergent multi-robot cooperative measurement, characterized in that, Comprising the following steps: S1, all measurement robots in the multi-robot system jointly and in real time construct and maintain a shared digital environment containing a geometric information layer of the measured member by running a C-SLAM algorithm, in which a digital pheromone layer is constructed based on a standard three-dimensional model of the measured member; S2, each measurement robot independently perceives the distribution state of the attractive pheromone in the digital pheromone layer, obtains its local environment information and motion state, uses a probabilistic decision model to select a target measurement area from the geometric information layer of the measured member, and uses a swarm rule to plan a travel path in the target measurement area in combination with the state of adjacent measurement robots; S3, the measurement robot executes a measurement action in the target measurement area according to the planned travel path, moves along the planned travel path to the selected target measurement area, adjusts the attitude of the scanner to scan the target measurement area, collects new measurement data of the measured member, and fuses the new measurement data into the geometric information layer of the shared digital environment through pose transformation of the C-SLAM framework, while performing quality assessment on the new measurement data of the target measurement area through metrology data point cloud density: When the measurement data point cloud density of the target measurement area is less than the target density threshold, the quality assessment result is not up to standard; When the measurement data point cloud density of the target measurement area reaches the target density threshold, the quality assessment result is up to standard; According to the quality assessment result, dynamically update the digital pheromone layer: When the quality assessment result of the new measurement data of the target measurement area by the measurement robot is not up to standard, increase the concentration of attractive pheromone in this area; When the quality assessment result of the new measurement data of the target measurement area by the measurement robot is up to standard, update the attractive pheromone in this area to repulsive pheromone; S4, continuously execute steps S2 and S3, and in the process, real-time detect the global concentration of attractive pheromone in the shared digital environment, and when the global concentration of attractive pheromone is lower than a preset termination threshold, terminate the measurement process, and output the measured three-dimensional model of the measured member in the final fused geometric information layer in the shared digital environment.

2. The multi-robot collaborative measurement method based on swarm intelligence emergence according to claim 1, characterized in that: The digital pheromone layer is a dynamic data field established on the geometric information layer of the measured member, in which each geometric coordinate of the geometric information layer is stored with an initial numerical data as a digital pheromone representing the measurement uncertainty of the corresponding coordinate area, and the digital pheromone includes attractive pheromone for attracting robots to measure uncertain areas and repulsive pheromone for driving robots away from completed measurement areas.

3. The multi-robot collaborative measurement method based on swarm intelligence emergence according to claim 2, characterized in that: In step S1, the concentration of the attractive pheromone is positively correlated with the measurement uncertainty of the corresponding area.

4. The crowd-sourced multi-robot cooperative measurement method according to claim 1, characterized in that: Step S2 includes the following sub-steps: S21, the measurement robot independently queries the shared digital environment to obtain the attractive pheromone distribution state required for its decision, and calculates the probability of the measurement robot going to each potential measurement area through a probability decision model: ， in, Measuring robot k in Choose to go to the potential measurement area at any time The probability, For potential measurement areas exist The concentration of attraction pheromones at any given moment To measure robot k to the potential measurement area The heuristic information is the reciprocal of the distance from the robot to the potential measurement area. For pheromone weighting factors, As a heuristic information weighting factor, Given the set of all available potential measurement regions for the measurement robot k, the measurement robot selects the target measurement region from the set of potential measurement regions based on the probability of each potential measurement region. S22, obtain the motion state of the measurement robot in the global coordinate system of the shared digital environment, calculate the velocity vector of the measurement robot at the next moment through the cluster rule to plan the travel path, the velocity vector of the travel path is determined by a cohesion vector , a separation vector and an alignment vector . ， wherein, is a final velocity vector of the measurement robot k at the time instant, is a final velocity vector of the measurement robot k at the time instant, is a cohesion vector whose direction points to the target measurement region selected by step S21, is a separation vector whose direction is opposite to the neighboring measurement robot, is an alignment vector whose direction is the average movement direction of the neighboring measurement robot, , , are weight coefficients of the respective vectors; S23, taking the speed vector obtained in step S22 as the kinematic constraint of the measurement robot, and combining the local environment information of the measurement robot to plan the travel path of the measurement robot in the target measurement area.

5. The multi-robot collaborative measurement method based on swarm intelligence emergence according to claim 4, characterized in that: In step S21, the current set of all optional potential measurement areas of the measurement robot includes the following two subsets: The local potential measurement area set, the measurement robot queries all potential measurement areas with an attractive pheromone concentration greater than zero near its location from the shared digital environment; The global priority measurement area set, the measurement robot queries the top N potential measurement areas with attractive pheromone concentration from the global scope of the shared digital environment, N being a preset value.

6. The multi-robot collaborative measurement method based on swarm intelligence emergence according to claim 4, characterized in that: In the step S22, the cohesion vector is represented by the following equation: ， wherein, is a center point of the target measurement area, is the accurate pose of the measurement robot k in the global coordinate system at time t; the separation vector is represented by the following formula: ， wherein, is the accurate pose of the neighboring measuring robot j in the global coordinate system at time t, is the set of neighboring measuring robots of measuring robot k; The alignment vector is represented by the following formula: ， wherein, is the number of neighboring measuring robots for measuring robot k, is the motion state of neighboring measuring robot j in the global coordinate system at time t.

7. The multi-robot collaborative measurement method based on swarm intelligence emergence according to claim 1, characterized in that: In step S4, the set of all measurement robots of the multi-robot system continuously repeats steps S2 and S3, real-time monitors the sum of the attractive pheromone concentration in all optional potential measurement areas of the measurement robots in the shared digital environment, and compares it with the preset termination threshold, when the following termination conditions are met, the measurement task is ended: ， wherein, is the potential measurement area at the pheromone concentration at the moment, allvoxels is the number of potential measurement areas of all measurement robots, is a preset global termination threshold, after the measurement task is completed, the measurement activities of all measurement robots are automatically terminated, and the final fused measured component geometric information layer model and measurement data are exported from the shared digital environment as the final measurement three-dimensional model of the measured component. 8.A multi-robot cooperative measurement system based on crowd-sourced emergence, characterized in that: A multi-robot system composed of two or more detection robots, the multi-robot system measures the measured component through the multi-robot collaborative measurement method based on swarm intelligence emergence according to any one of claims 1-7.

Images