Composite robot collaborative machining system

Abstract

This invention discloses a composite robot collaborative processing system, relating to the field of automated production technology. The composite robot collaborative processing system includes at least one processing unit for processing materials; each processing unit is equipped with a loading / unloading execution unit for loading / unloading the materials within the processing unit; a transport unit for transporting the materials between the processing units; and a control and management unit electrically connected to the processing unit, the loading / unloading execution unit, and the transport unit. The technical solution provided by this invention enables the transport and movement of materials between processing units, as well as the loading / unloading operations of materials within each processing unit.

Description

Technical Field

[0001] This invention relates to the field of automated production technology, and in particular to a composite robot collaborative processing system. Background Technology

[0002] As the manufacturing industry shifts towards small-batch, multi-variety, and customized models, especially in discrete manufacturing fields such as medical devices, precision molds, and aerospace, stringent requirements are placed on the flexibility and operational efficiency of production systems. In existing technologies, the handling of materials between processing units (such as CNC machine tools) and the loading and unloading of materials within each unit are all done manually. This manual operation increases labor intensity, increases the risk of workplace injuries, and is relatively inefficient. Furthermore, products handled manually lack sufficient quality stability and cannot meet the demands of mass production. Therefore, a novel composite robot collaborative processing system is urgently needed to realize the handling of materials between processing units and the loading and unloading of materials within each unit. Summary of the Invention

[0003] The main objective of this invention is to propose a composite robot collaborative processing system, which aims to realize the handling and movement of materials and products between processing units, as well as the loading and unloading operations of materials and products in each processing unit.

[0004] To achieve the above objectives, the present invention proposes a composite robot collaborative processing system, comprising: At least one processing unit is provided for processing materials and products; wherein each processing unit is equipped with a loading and unloading execution unit for loading and unloading the materials and products into and out of the processing unit. A transport unit is used to transport the material products between the processing units. The control and management unit is electrically connected to the processing unit, the loading and unloading execution unit, and the handling unit.

[0005] In one embodiment, the transport unit includes: Mobile base; A first multi-degree-of-freedom robotic arm is mounted on the mobile base; a first gripper is mounted at the end of the first multi-degree-of-freedom robotic arm, and the first gripper is used to perform gripping operations on the material product; The path navigation module is used to generate a travel path from the starting point to the ending point for the mobile base through a map server and / or a global and local planner and / or a cost map. The autonomous driving module uses lidar and IMU technology to drive the mobile base to drive autonomously along the driving path.

[0006] In one embodiment, the transport unit includes an intelligent obstacle avoidance module, which is configured as follows: Obtain the current travel speed of the mobile base; The driving path is divided into multiple active protection zones based on the current driving speed, wherein each protection zone corresponds to a different speed range and has a different zone size; Detect whether there are obstacles within the active protection area; When the obstacle is present within the active protection area, the mobile base is driven to enter the protection stop state; When the obstacle within the active protection area is removed, the mobile base is driven to exit the protection stop state.

[0007] In one embodiment, the intelligent obstacle avoidance module includes a laser sensor and / or an ultrasonic radar sensor, the laser sensor and / or the ultrasonic radar sensor being used to detect whether the obstacle exists within the active protection area.

[0008] In one embodiment, the loading / unloading execution unit includes: Fixed base; A second multi-degree-of-freedom robotic arm is mounted on the fixed base; The quick-change fixture module is installed at the end effector of the second multi-degree-of-freedom robotic arm; the quick-change fixture module is equipped with at least two second fixtures; A deep learning vision module is installed at the end effector of the second multi-degree-of-freedom robotic arm, and its field of view points to the gripping area of ​​the second gripper. The deep learning vision module is configured as follows: Acquire multimodal data of the material product, wherein the multimodal data includes first two-dimensional image data and three-dimensional point cloud data of the material product; The first two-dimensional image data and the three-dimensional point cloud data are fused to form a high-dimensional feature vector; The high-dimensional feature vector is matched with the standard feature vector stored in the standard database to identify the specification and model information of the material product; The quick-change fixture module switches to the corresponding second fixture according to the specifications and model information of the material product.

[0009] In one embodiment, the step of fusing features from the first two-dimensional image data and the three-dimensional point cloud data to form a high-dimensional feature vector includes the following steps: A first set of geometric features is extracted from the first two-dimensional image data, wherein the first set of geometric features includes the shape contour, aspect ratio and hole distribution features of the product to be inspected; A second set of geometric features is extracted from the three-dimensional point cloud data. The second set of geometric features includes the overall size, volume and surface area features of the product to be inspected. The first set of geometric features and the second set of geometric features are concatenated and fused to form the high-dimensional feature vector.

[0010] In one embodiment, the loading / unloading execution unit includes a material buffer module, which is disposed in the junction area between the loading / unloading execution unit and the handling unit; the material buffer module is used to temporarily place the material products to be processed transferred from the handling unit and the processed material products transferred from the processing unit.

[0011] In one embodiment, the material buffer module is configured as a Go board structure or a chessboard structure.

[0012] In one embodiment, the end effector of the second multi-degree-of-freedom robotic arm is equipped with a quick-change fixture module, which has at least two second fixtures installed. The quick-change fixture module is used to change the corresponding second fixture according to the material properties of the material product.

[0013] In one embodiment, the processing unit includes: An optical transmission trigger probe module is used for online measurement and alignment of the material product; The wired five-axis tool setter module is used to measure the length, diameter, and runout of machining tools; A pneumatic material clamping module is used to pneumatically clamp and release the material product; wherein the pneumatic material clamping module can adjust its clamping force according to the material properties and process properties of the material product.

[0014] In one embodiment, the control and management unit adopts an MES system module combined with an IoT system module, which can be used to integrate, analyze and process data from the processing unit, the loading and unloading execution unit and the handling unit and production process.

[0015] The technical solution of this invention sets up a handling unit to be responsible for the handling of materials and products between processing units, and equips each processing unit with a loading / unloading execution unit to be responsible for the loading / unloading operations of materials and products. Simultaneously, a control and management unit centrally schedules and coordinates the processing units, loading / unloading execution units, and handling units. This achieves full automation of material and product flow and processing clamping, fundamentally eliminating the labor burden and workplace injury risks associated with manual operation. Secondly, the handling and loading / unloading operations are functionally independent, can be executed in parallel in time, and can be spatially separated, forming a decoupled architecture. Specifically, the handling unit and the loading / unloading execution unit each perform their own functions and do not depend on each other. In the time dimension, when the handling unit performs its flow task, the loading / unloading execution units at each processing unit can simultaneously unload completed materials and products and load materials to be processed into the processing unit, working in parallel without obstruction. In the spatial dimension, the main travel path of the handling unit is separated from the working area of ​​the loading / unloading execution unit, with material handover only occurring at temporary buffer locations or designated interaction points, avoiding the interference risks that may arise from simultaneous operations in the same spatial area. When the material handling unit malfunctions, the loading / unloading execution unit can still perform loading / unloading operations on the same machine, and vice versa, achieving system-level fault tolerance and high availability. This decoupled design significantly improves the system's modularity, operational efficiency, ease of maintenance, and overall flexibility of the production and processing system. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a schematic diagram of an embodiment of the composite robot collaborative processing system provided by the present invention.

[0018] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0019] The technical solutions of this invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a portion of the embodiments of this invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0020] It should be noted that if the embodiments of the present invention involve directional indicators (such as up, down, left, right, front, back, etc.), the directional indicators are only used to explain the relative positional relationship and movement of the components in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indicators will also change accordingly.

[0021] Furthermore, it should be noted that the descriptions involving "first," "second," etc., in this invention are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature. Additionally, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or impossible to implement, such a combination of technical solutions should be considered non-existent and not within the scope of protection claimed by this invention.

[0022] As the manufacturing industry shifts towards small-batch, multi-variety, and customized models, especially in discrete manufacturing fields such as medical devices, precision molds, and aerospace, stringent requirements are placed on the flexibility and operational efficiency of production systems. In existing technologies, the handling of materials and products between processing units (such as CNC machine tools) and the loading and unloading of materials and products within each processing unit are all done manually. On the one hand, manual labor increases labor intensity, increases the risk of workplace injuries, and results in relatively low efficiency. On the other hand, products handled manually lack sufficient quality stability and cannot meet the demands of mass production.

[0023] To address the aforementioned technical problems, this invention proposes a composite robot collaborative processing system.

[0024] Please see Figure 1 In one embodiment of the present invention, the composite robot collaborative processing system includes: At least one processing unit is provided for processing materials and products; wherein each processing unit is equipped with a loading and unloading execution unit for loading and unloading the materials and products into and out of the processing unit. A transport unit is used to transport the material products between the processing units. The control and management unit is electrically connected to the processing unit, the loading and unloading execution unit, and the handling unit.

[0025] The technical solution of this invention sets up a handling unit to be responsible for the handling of materials and products between processing units, and equips each processing unit with a loading / unloading execution unit to be responsible for the loading / unloading operations of materials and products. Simultaneously, a control and management unit centrally schedules and coordinates the processing units, loading / unloading execution units, and handling units. This achieves full automation of material and product flow and processing clamping, fundamentally eliminating the labor burden and workplace injury risks associated with manual operation. Secondly, the handling and loading / unloading operations are functionally independent, can be executed in parallel in time, and can be spatially separated, forming a decoupled architecture. Specifically, the handling unit and the loading / unloading execution unit each perform their own functions and do not depend on each other. In the time dimension, when the handling unit performs its flow task, the loading / unloading execution units at each processing unit can simultaneously unload completed materials and products and load materials to be processed into the processing unit, working in parallel without obstruction. In the spatial dimension, the main travel path of the handling unit is separated from the working area of ​​the loading / unloading execution unit, with material handover only occurring at temporary buffer locations or designated interaction points, avoiding the interference risks that may arise from simultaneous operations in the same spatial area. When the material handling unit malfunctions, the loading / unloading execution unit can still perform loading / unloading operations on the same machine, and vice versa, achieving system-level fault tolerance and high availability. This decoupled design significantly improves the system's modularity, operational efficiency, ease of maintenance, and overall flexibility of the production and processing system.

[0026] As a preferred embodiment of the above embodiments, the transport unit includes: Mobile base; A first multi-degree-of-freedom robotic arm is mounted on the mobile base; a first gripper is mounted at the end of the first multi-degree-of-freedom robotic arm, and the first gripper is used to perform gripping operations on the material product; The path navigation module is used to generate a travel path from the starting point to the ending point for the mobile base through a map server and / or a global and local planner and / or a cost map. The autonomous driving module uses lidar and IMU technology to drive the mobile base to drive autonomously along the driving path.

[0027] With this configuration, the mobile base serves as the mobile carrier of the entire handling unit, enabling it to move autonomously within the workshop environment. A first multi-degree-of-freedom robotic arm is mounted on the mobile base, with a first gripper at its end effector. This first multi-degree-of-freedom robotic arm can perform multi-axis movements to adjust the spatial position and attitude of the first gripper, thereby achieving precise gripping and placement of materials. This allows the handling unit to not only transport materials but also proactively complete the loading, unloading, and handover of materials. The path navigation module, by calling the pre-stored environmental map in the map server and comprehensively utilizing global and local planners as well as cost map algorithms, automatically plans a path that avoids obstacles and is optimal in terms of distance or time, based on the current starting point and target endpoint. The autonomous driving module uses lidar for environmental perception and positioning, and IMU (Inertial Measurement Unit) for attitude and heading calculations, driving the mobile base to autonomously drive along the path planned by the path navigation module, without the need for manual remote control or track laying. The four modules work together to enable the handling unit to move autonomously from the starting point to the destination, and can independently grab and release materials, thus achieving full automation of the handling operation without human intervention.

[0028] The map server is a software module or hardware device used to store, manage, and provide environmental map data. Its working principle is as follows: During the initial deployment or operation of the system, the material handling unit scans the environment using sensors such as LiDAR and binocular vision. It then uses simultaneous localization and mapping (SLAM) algorithms to build a real-time map of the workshop environment, including the location information of fixed obstacles such as walls, equipment, racks, and processing units. The completed map is stored in the map server in the form of a raster map, topological map, or feature map. Each map cell records information such as whether there are obstacles or whether the location is a feasible area. When the material handling unit needs to perform a material handling task, the path navigation module requests map data of the current work area from the map server. The server returns the corresponding map information as the basis for subsequent path planning. This component enables centralized storage and reuse of the environmental map, avoiding the need for re-scanning and mapping for each task, thus saving computing resources and time.

[0029] The global planner is responsible for calculating the macroscopic path from the starting point to the destination based on the complete environmental map. Its working principle is as follows: it obtains the global environmental map provided by the map server, as well as the starting and ending coordinates specified for the current task. It then uses a graph search algorithm (such as A algorithm, Dijkstra's algorithm, D algorithm, etc.) to find the optimal path in the map. The algorithm discretizes the map into a node network, where each node represents a location point, and the edges between nodes represent walkable path segments. Each edge can be assigned a cost (such as distance, time, energy consumption, etc.). The search is gradually expanded from the starting point to the destination, and the node sequence with the minimum cumulative cost is selected as the output path. This path is a coarse path at the global level, not considering local changes such as dynamic obstacles. The local planner is responsible for micro-level path adjustments based on real-time sensor information, building upon the global path. Its working principle is as follows: a dynamic window is constructed centered on the current position of the transport unit, considering only environmental information within this window to reduce computational complexity. The linear and angular velocities of the transport unit are sampled, and the possible trajectories formed within future time windows under each set of velocity parameters are simulated. A multi-dimensional cost evaluation is performed on each candidate trajectory (evaluation indicators include whether there is a collision with an obstacle, whether it deviates from the global planned path, whether the movement direction points to the target point, and whether the travel speed is smooth). The trajectory with the lowest evaluation cost is selected as the execution trajectory for the current control cycle, driving the transport unit to move along that trajectory. Working together, the global planner ensures that the transport unit can find a feasible path connecting the starting and ending points from a macroscopic perspective, while the local planner enables the transport unit to respond in real-time to dynamic obstacles in the environment (such as personnel and other mobile devices), achieving obstacle avoidance and path fine-tuning, thus compensating for the global planner's inability to perceive the dynamic environment.

[0030] The cost map is the core data structure in path planning, representing the "cost" or "difficulty" of traversing different areas of the environment for the mobile base. Its working principle is as follows: The cost map is typically divided into multiple layers, including a static layer composed of fixed obstacles from a map server, an obstacle layer composed of real-time obstacle detection from sensors such as LiDAR, and an expansion layer consisting of safety buffer zones around obstacles. Each map grid is assigned a cost value, typically ranging from 0 to 255. A cost value of 0 represents a completely feasible area (free space), a cost value of 255 represents a completely infeasible area (grids occupied by obstacles), and intermediate values ​​represent different levels of danger zones or high-cost traversal areas (such as areas near obstacles, slippery surfaces, etc.). The obstacle layer and expansion layer in the cost map are dynamically updated frequently based on real-time sensor feedback, ensuring the map reflects the latest state of the environment. Both the global planner and the local planner perform path search and trajectory evaluation based on the cost map, tending to choose areas with low cost values ​​during the search process and actively avoiding obstacles and nearby areas. This component provides a unified quantitative expression of geometric information and safety constraints in the environment, offering a cost space that can be directly used for path search by the planner, effectively balancing the optimality and safety of the path.

[0031] Furthermore, the transport unit includes an intelligent obstacle avoidance module, which is configured to: acquire the current travel speed of the mobile base; divide the travel path into multiple active protection zones based on the current travel speed, wherein each protection zone corresponds to a different speed range and has a different zone size; detect whether there are obstacles within the active protection zones; when an obstacle exists within the active protection zone, drive the mobile base into a protection stop state; and when the obstacle is eliminated within the active protection zone, drive the mobile base out of the protection stop state. Specifically, during the movement of the mobile base along the travel path, a specific area extending outward from the mobile base as its center is defined as the active protection zone. The size of this zone is not fixed but dynamically determined according to the real-time speed of the mobile base. The higher the speed, the longer the braking distance of the mobile base, so a larger active protection zone is needed to ensure sufficient distance to complete safe braking when an obstacle is detected. The lower the speed, the shorter the required braking distance, and the smaller the active protection zone is accordingly to avoid frequent unnecessary shutdowns due to an excessively large area (for example, a protection zone larger than 35cm is selected when the speed is less than 0.1m / s, a protection zone larger than 40cm is selected when the speed is between 0.1 and 0.3m / s, and a protection zone larger than 135cm is selected when the speed is greater than 0.9m / s). Once the size of the active protection zone is determined, the intelligent obstacle avoidance module continuously monitors the area for obstacles using its configured laser sensors and / or ultrasonic radar sensors. Upon detecting an obstacle, it drives the mobile base to brake and enter a protective stop state, bringing the mobile base to a complete stop before contacting the obstacle. Subsequently, the system continuously monitors the status of the active protection zone. Only when the obstacle is removed from the zone or the mobile base replans a path to avoid the obstacle, and the zone's condition is deemed cleared, can the mobile base exit the protective stop state and resume normal operation. The beneficial effect of this technical solution is that by dynamically linking the active protection zone with the real-time speed of the mobile base, speed-based graded safety protection is achieved—providing sufficient safe reaction distance to ensure reliable braking at high speeds, and avoiding oversensitivity leading to frequent shutdowns at low speeds, thereby maximizing the operating efficiency of the handling unit while ensuring safety.

[0032] Furthermore, the intelligent obstacle avoidance module includes a laser sensor and / or an ultrasonic radar sensor, which are used to detect whether the obstacle exists within the active protection area. Specifically, the laser sensor accurately calculates the distance between the sensor and the obstacle by emitting a laser beam and measuring its reflection time (i.e., the time of flight method) or by using triangulation. Its advantages include high detection accuracy, strong directionality, and high angular resolution, enabling accurate perception of the outline and position of obstacles over long distances. However, its detection effect on transparent objects (such as glass doors and windows) or highly reflective surfaces is limited. The ultrasonic radar sensor, on the other hand, emits ultrasonic pulses and receives the echoes, calculating the distance to the obstacle based on the propagation speed and time difference of sound waves in the medium. Its advantages include strong adaptability to transparent objects, light-sensitive environments, and dusty environments, and it is unaffected by the color or surface reflectivity of the target object. However, its detection accuracy is relatively low and its beam angle is relatively large. The intelligent obstacle avoidance module of this invention can be configured with either a laser sensor or an ultrasonic radar sensor individually, or both can be configured simultaneously to form a multi-sensor fusion detection architecture. When configured simultaneously, the laser sensor and the ultrasonic radar sensor serve as redundant backups for each other. A data fusion algorithm is used to comprehensively judge the detection results from both: for transparent obstacles that are difficult for the laser sensor to detect, the ultrasonic radar sensor provides supplementary detection; for scenarios where the ultrasonic radar sensor's accuracy is insufficient, the laser sensor provides precise distance and contour information. This fusion configuration of the two sensors effectively eliminates the perception blind spots that may exist with a single sensor, ensuring that regardless of the type of obstacle present in the workshop environment (including personnel, equipment, transparent partitions, dusty areas, etc.), it can be reliably detected and timely triggers protective shutdown, thereby significantly improving the operational safety and environmental adaptability of the handling unit in complex industrial environments.

[0033] As a preferred embodiment of the above embodiments, the loading and unloading execution unit includes: Fixed base; A second multi-degree-of-freedom robotic arm is mounted on the fixed base; The quick-change fixture module is installed at the end effector of the second multi-degree-of-freedom robotic arm; the quick-change fixture module is equipped with at least two second fixtures; A deep learning vision module is installed at the end effector of the second multi-degree-of-freedom robotic arm, and its field of view points to the gripping area of ​​the second gripper. The deep learning vision module is configured as follows: Acquire multimodal data of the material product, wherein the multimodal data includes first two-dimensional image data and three-dimensional point cloud data of the material product; The first two-dimensional image data and the three-dimensional point cloud data are fused to form a high-dimensional feature vector; The high-dimensional feature vector is matched with the standard feature vector stored in the standard database to identify the specification and model information of the material product; The quick-change fixture module switches to the corresponding second fixture according to the specifications and model information of the material product.

[0034] With this configuration, the loading / unloading execution unit of the present invention has a second multi-degree-of-freedom robotic arm mounted on its fixed base. The end effector of this robotic arm is simultaneously equipped with a quick-change fixture module and a deep learning vision module. The quick-change fixture module is equipped with at least two different types of second fixtures. The field of view of the deep learning vision module points to the gripping area of ​​the second fixture. The two work together to automatically change the corresponding fixture according to the specifications and model of the material product through the following collaborative workflow: First, the deep learning vision module acquires multimodal data of the material product. This multimodal data includes first-dimensional image data of the material product (color or grayscale images captured by a high-definition industrial camera, used to provide appearance features such as color, texture, and edge contours) and third-dimensional point cloud data (a set of spatial coordinate points collected by a 3D laser scanner or binocular stereo vision sensor, used to provide depth information, 3D dimensions, surface shape, and spatial pose information of the material product). Second, the deep learning vision module uses a multimodal feature fusion algorithm based on a deep neural network to combine the appearance features extracted from the two-dimensional image data with… The geometric features extracted from the 3D point cloud data are fused at the feature level to form a unified high-dimensional feature vector. This high-dimensional feature vector simultaneously contains the apparent attributes and spatial geometric attributes of the material product, and has stronger descriptive and discriminative capabilities compared to single-modal data. Next, the deep learning vision module performs similarity matching calculations on this high-dimensional feature vector and multiple standard feature vectors pre-stored in the standard database. The standard database pre-stores standard feature vectors corresponding to different specifications and models of material products and their associated fixture type information. Once a match is successful, the specification and model information of the current material product is identified. Finally, the deep learning vision module sends the identified specification and model information to the control and management unit. The control and management unit determines the type of fixture required for the current material product based on this information and sends a replacement command to the fixture quick-change module. The second multi-degree-of-freedom robotic arm automatically moves to the fixture replacement station, places the fixture currently installed at the end effector on the fixture frame, aligns it with the target fixture, and completes the automatic gripping and locking of the new fixture through the locking mechanism of the quick-change module, thus completing the automatic fixture switching. The beneficial effects of this technical solution are as follows: by integrating multimodal data of two-dimensional images and three-dimensional point clouds, it achieves accurate identification of material product specifications and models, overcoming the problem of decreased recognition rate of a single vision sensor under conditions of lighting changes, texture loss, or occlusion; at the same time, by linking the visual recognition results with the quick-change fixture module, the system can automatically select the most suitable fixture for gripping after identifying the material, without the need for manual preset or manual clamp changing, realizing fully automatic flexible adaptation of small batches and multiple varieties of materials, which can be identified and changed immediately upon arrival.

[0035] Furthermore, the step of fusing features from the first two-dimensional image data and the three-dimensional point cloud data to form a high-dimensional feature vector includes the following steps: A first set of geometric features is extracted from the first two-dimensional image data, wherein the first set of geometric features includes the shape contour, aspect ratio and hole distribution features of the product to be inspected; A second set of geometric features is extracted from the three-dimensional point cloud data. The second set of geometric features includes the overall size, volume and surface area features of the product to be inspected. The first set of geometric features and the second set of geometric features are concatenated and fused to form the high-dimensional feature vector.

[0036] This setup, involving feature fusion of the first two-dimensional image data and the three-dimensional point cloud data to form a high-dimensional feature vector, specifically includes the following sub-steps: First, extracting a first geometric feature set from the first two-dimensional image data. This first geometric feature set includes the shape contour of the product to be inspected (i.e., the outer boundary curve of the product on the two-dimensional plane, used to characterize the overall shape category of the product, such as round, square, irregular, etc.), aspect ratio (i.e., the ratio of the long side to the short side of the product's circumscribed rectangle, used to distinguish between slender or flat products), and hole distribution features (i.e., whether the product surface has through holes or blind holes, the number of holes, the size of the holes, and the relative position distribution of the holes, used to distinguish products with different functional structures); Second, extracting features from the three-dimensional point cloud data... The first geometric feature set is extracted from the image. This second geometric feature set includes the overall dimensions of the product under inspection (i.e., the actual physical dimensions of the product in three-dimensional space: length, width, and height), volume (i.e., the total volume of the product in three-dimensional space, used to distinguish solid parts from hollow parts or to estimate product weight), and surface area features (i.e., the total area of ​​the product's outer surface in three-dimensional space, used to characterize the complexity of the product). Finally, the first and second geometric feature sets are concatenated and fused. That is, the apparent geometric features such as contours, aspect ratios, and hole distribution extracted from the two-dimensional image are combined with the spatial geometric features such as dimensions, volume, and surface area extracted from the three-dimensional point cloud in a series along the feature vector dimension to form a unified high-dimensional feature vector. The beneficial effects of this technical solution are: two-dimensional image data can provide high-resolution texture and edge information, but is not sensitive to depth and scale information; three-dimensional point cloud data can provide accurate spatial geometric information, but has relatively low resolution and lacks apparent information. The two have complementary information. By fusing features such as shape contours, aspect ratios, and hole distribution extracted from 2D images with features such as overall size, volume, and surface area extracted from 3D point clouds, the resulting high-dimensional feature vector simultaneously encompasses both the apparent geometric attributes and spatial geometric attributes of the material product. This preserves the sensitivity of 2D images to local details while supplementing the ability of 3D data to accurately describe the global spatial scale. As a result, the feature vector provides a more comprehensive, robust, and discriminative description of the material product, effectively improving the accuracy and reliability of subsequent matching with standard feature vectors in a standard database. In particular, when faced with material products that are similar in appearance but different in size or similar in size but different in structure, multimodal fusion features can significantly outperform single-modal data features in terms of recognition performance.

[0037] Furthermore, the loading / unloading execution unit includes a material buffer module, which is located in the handover area between the loading / unloading execution unit and the handling unit. The material buffer module is used to temporarily hold the material products to be processed transferred from the handling unit, as well as the processed material products transferred from the processing unit. Specifically, in traditional automated production lines, the handling unit and the loading / unloading execution unit typically require strict time synchronization. The handling unit must wait for the loading / unloading execution unit to complete the loading and unloading of the current material product before it can hand over the next material, or the loading / unloading execution unit must wait for the handling unit to arrive before it can perform loading / unloading operations. This tightly coupled working mode easily leads to equipment idleness and wasted time due to one party waiting for the other. This invention introduces a material buffer module: when the handling unit transports a batch of material products to be processed to the vicinity of the processing unit, it does not need to wait for the loading / unloading execution unit to immediately process these material products. Instead, it can place the material products on the designated workstation of the material buffer module and leave to continue performing other handling tasks. Subsequently, the loading / unloading execution unit can calmly grab the material products to be processed one by one from the material buffer module for loading operations according to its own work rhythm. Similarly, after the loading / unloading execution unit completes workpiece processing, it does not need to wait for the transport unit to arrive immediately. The processed material can be placed in the material buffer module, and the transport unit can retrieve it after completing other tasks. This technical solution decouples the transport and loading / unloading operations in the time dimension, allowing them to operate independently at their optimal pace. The transport unit does not need to wait for loading / unloading to complete at the processing unit, and the loading / unloading execution unit does not need to stop due to delays in the transport unit, thus effectively avoiding accumulated waiting time and significantly improving the overall operating efficiency and equipment utilization of the production line. At the same time, the material buffer module, as a standardized handover interface, clarifies the functional boundaries of the transport unit and the loading / unloading execution unit. The transport unit only needs to have the ability to place materials at the designated position in the buffer module, without needing to master the complex positioning requirements of direct docking with the processing unit; the loading / unloading execution unit only needs to retrieve materials from the buffer module, without needing to have the ability to dynamically hand over materials with the moving base, thus achieving further decoupling between the two in terms of function.

[0038] Furthermore, the material buffer module is configured as a Go board structure or a chessboard structure. This configuration, where the material buffer module is configured as a Go board structure or a chessboard structure, means that the workpiece placement area of ​​the buffer module is designed as a grid layout with regular array characteristics. The Go board structure corresponds to a matrix grid layout (i.e., horizontally and vertically spaced equally spaced rows and columns arranged alternately), while the chessboard structure corresponds to an alternating array layout (i.e., adjacent rows of workpiece placement points are staggered by half a grid interval, similar to the initial arrangement of chess pieces on a chessboard). Specifically, in the Go board structure, the workpiece placement stations are evenly distributed in a rectangular grid of M rows × N columns according to a Cartesian coordinate system. Each station has a clear and unique world coordinate (X_i, Y_j). The biggest advantage of this layout is its high compatibility with the recognition algorithm of the deep learning vision module. The vision module can quickly traverse all stations by following a fixed row and column scanning order. The theoretical position of each station can be calculated by simple linear interpolation. The vision system only needs to perform a binary classification judgment on whether a workpiece exists at each station, which greatly reduces the complexity of image processing and the consumption of computing resources. In the chessboard structure, the workpiece placement points in odd-numbered rows and even-numbered rows are staggered by half a station spacing in the horizontal direction. This staggered arrangement increases the effective spacing between adjacent workpieces in the same buffer area (i.e., the minimum distance between the center points of any two adjacent stations is increased), thereby reducing the edge overlap or occlusion interference problems that may occur when adjacent workpieces are visually recognized. It is particularly suitable for buffering scenarios of workpieces with large dimensions or irregular shapes. Whether it's a Go board structure or a chessboard structure, the robotic arm of the loading and unloading execution unit can directly plan its motion trajectory to the target workstation according to the preset coordinate mapping relationship, without having to perform complex real-time position calculations and visual guidance one by one. This significantly shortens the time for the robotic arm to move and position between multiple workstations, improves the storage and retrieval efficiency of buffered materials, and the regular arrangement also enables the handling unit to complete the material placement and retrieval operations with a standardized sequence of actions, reducing the complexity of system integration.

[0039] As a preferred embodiment of the above embodiments, the processing unit includes: An optical transmission trigger probe module is used for online measurement and alignment of the material product; The wired five-axis tool setter module is used to measure the length, diameter, and runout of machining tools; A pneumatic material clamping module is used to pneumatically clamp and release the material product; wherein the pneumatic material clamping module can adjust its clamping force according to the material properties and process properties of the material product.

[0040] With this setup, the optical transmission trigger probe module, acting as a movable precision measuring unit, automatically measures the actual position of the material on the machine tool table and compensates for deviations before processing by contacting the material's preset reference point. During processing, it performs in-sequence measurements of key dimensions of the material to adjust parameters for subsequent processes. After processing, it automatically detects the finished product dimensions and uploads the results to the MES system for quality traceability, thus achieving full-process online measurement and automatic alignment of the material. The wired five-axis tool setter module supports trigger measurements in five directions: ±X, ±Y, and +Z. After tool replacement, it automatically measures the tool length and writes it into the CNC system's tool compensation table. It can also measure the tool diameter for radial compensation and detect the radial runout of the tool during rotation to determine the tool's concentricity. When the deviation between the measured value and the preset value exceeds the threshold, a tool breakage alarm is automatically triggered, thereby achieving precise measurement and management of the length, diameter, and runout of the machining tool. The material pneumatic clamping module, as the execution unit for fixing material products, uses pneumatic drive to achieve automatic clamping and release of material products. Its core technical feature is that the clamping force can be dynamically adjusted according to the material properties of the material product (such as the workpiece material being steel, aluminum alloy, or plastic, and the workpiece shape being thin-walled or thick-walled) and process properties (such as rough machining requiring high clamping force to resist large cutting forces, and finish machining requiring appropriately reduced clamping force to reduce workpiece deformation). For rough machining of steel parts, the maximum clamping force can be adjusted to 30kN to provide sufficient workpiece stability. For finish machining of thin-walled aluminum alloy parts, it can be precisely controlled at a lower level to prevent workpiece deformation under pressure. The beneficial effects of the three modules working together are as follows: the optical transmission trigger probe module saves auxiliary time and ensures machining accuracy; the wired five-axis tool setter module realizes automatic compensation of machining tool parameters and tool breakage detection; and the material pneumatic clamping module realizes compatible machining of various materials and processes through adjustable clamping force. Together, the three enable the machining unit to operate stably under unattended or minimally manned conditions, significantly improving the level of machining automation and flexible adaptability.

[0041] As a preferred embodiment of the above, the control and management unit adopts a combination of an MES system module and an IoT system module. This allows for the integration, analysis, and processing of data from the processing unit, the loading / unloading execution unit, the handling unit, and the production process. In this configuration, the MES system module (Manufacturing Execution System) focuses on macro-level scheduling and control at the production management level, primarily responsible for creating and issuing production orders, managing product bills of materials (BOMs) and process routes, dispatching and reporting production work orders, developing quality inspection plans and tracing inspection records, managing warehouse inbound and outbound operations, and compiling equipment ledgers and maintenance plans—functions at the manufacturing execution level. The IoT system module (Industrial Internet of Things system) focuses on data acquisition and real-time monitoring at the equipment level. It uses a DNC communication module to remotely transmit and manage CNC programs, and an MDA equipment data acquisition module to acquire real-time data on the processing unit's equipment status (running, idle, alarm, shutdown), spindle load and speed, feed rate and ratio, processing quantity and processing time, tool number, and alarm information. Simultaneously, it collects the operating status, task progress, and fault information of the loading / unloading execution unit and the handling unit. The two system modules do not operate independently but are deeply integrated to form a collaborative closed loop: real-time equipment data collected by the IoT system module (such as whether a processing unit is currently idle, whether the loading / unloading execution unit has completed loading, and whether the transport unit is moving towards the target location) is pushed to the MES system module in real time. The MES system module then performs dynamic production scheduling based on this data—for example, when it detects that a processing unit is about to complete the processing of its current workpiece, it schedules the transport unit to go to that unit in advance to pick up the material; when it detects that a piece of equipment has triggered an alarm and stopped, it automatically reassigns the work orders for that equipment to other idle equipment. At the same time, production plans, process parameters, and other information in the MES system module can also be distributed to each execution unit through the IoT system module, achieving end-to-end data flow from order to execution. The beneficial effects of this technical solution are as follows: it realizes full-process digital control from the placement of production orders to the output of finished products, making the production process transparent and visible, and enabling managers to grasp the operating status of each unit in real time and make rapid decisions; the real-time data collected by IoT is fed back to MES for dynamic adjustment of production plans, forming a closed-loop control of perception-analysis-decision-execution, which significantly improves the intelligence level and response speed of the production system, while providing complete data support for quality traceability, equipment efficiency analysis (OEE), and production bottleneck identification.

[0042] It should be noted that other aspects of the composite robot collaborative processing system disclosed in this invention are existing technologies and will not be described in detail here.

[0043] The above are merely optional embodiments of the present invention and do not limit the patent scope of the present invention. Any application of the present invention directly or indirectly in other related technical fields is included within the patent protection scope of the present invention.

Claims

1. A composite robotic collaborative machining system, characterized by, include: At least one processing unit is provided for processing materials and products; wherein each processing unit is equipped with a loading and unloading execution unit for loading and unloading the materials and products into and out of the processing unit. A transport unit is used to transport the material products between the processing units. The control and management unit is electrically connected to the processing unit, the loading and unloading execution unit, and the handling unit.

2. The composite robotic collaborative machining system of claim 1, wherein: The transport unit includes: Mobile base; A first multi-degree-of-freedom robotic arm is mounted on the mobile base; a first gripper is mounted at the end of the first multi-degree-of-freedom robotic arm, and the first gripper is used to perform gripping operations on the material product; The path navigation module is used to generate a travel path from the starting point to the ending point for the mobile base through a map server and / or a global and local planner and / or a cost map. The autonomous driving module uses lidar and IMU technology to drive the mobile base to drive autonomously along the driving path.

3. The composite robotic collaborative machining system of claim 2, wherein: The transport unit includes an intelligent obstacle avoidance module, which is configured as follows: Obtain the current travel speed of the mobile base; The driving path is divided into multiple active protection zones based on the current driving speed, wherein each protection zone corresponds to a different speed range and has a different zone size; Detect whether there are obstacles within the active protection area; When the obstacle is present within the active protection area, the mobile base is driven to enter the protection stop state; When the obstacle within the active protection area is removed, the mobile base is driven to exit the protection stop state.

4. The composite robotic collaborative machining system of claim 3, wherein: The intelligent obstacle avoidance module includes a laser sensor and / or an ultrasonic radar sensor, which are used to detect whether the obstacle exists within the active protection area.

5. The composite robot collaborative processing system as described in claim 1, characterized in that: The loading and unloading execution unit includes: Fixed base; A second multi-degree-of-freedom robotic arm is mounted on the fixed base; The quick-change fixture module is installed at the end effector of the second multi-degree-of-freedom robotic arm; the quick-change fixture module is equipped with at least two second fixtures; A deep learning vision module is installed at the end effector of the second multi-degree-of-freedom robotic arm, and its field of view points to the gripping area of ​​the second gripper. The deep learning vision module is configured as follows: Acquire multimodal data of the material product, wherein the multimodal data includes first two-dimensional image data and three-dimensional point cloud data of the material product; The first two-dimensional image data and the three-dimensional point cloud data are fused to form a high-dimensional feature vector; The high-dimensional feature vector is matched with the standard feature vector stored in the standard database to identify the specification and model information of the material product; The quick-change fixture module switches to the corresponding second fixture according to the specifications and model information of the material product.

6. The composite robot collaborative processing system as described in claim 5, characterized in that: The step of fusing features from the first two-dimensional image data and the three-dimensional point cloud data to form a high-dimensional feature vector includes the following steps: A first set of geometric features is extracted from the first two-dimensional image data, wherein the first set of geometric features includes the shape contour, aspect ratio and hole distribution features of the product to be inspected; A second set of geometric features is extracted from the three-dimensional point cloud data. The second set of geometric features includes the overall size, volume and surface area features of the product to be inspected. The first set of geometric features and the second set of geometric features are concatenated and fused to form the high-dimensional feature vector.

7. The composite robot collaborative processing system as described in claim 5, characterized in that: The loading and unloading execution unit includes a material buffer module, which is located in the junction area between the loading and unloading execution unit and the handling unit. The material buffer module is used to temporarily place the material products to be processed transferred from the handling unit and the processed material products transferred from the processing unit.

8. The composite robot collaborative processing system as described in claim 7, characterized in that: The material caching module is configured as a Go board structure or a chessboard structure.

9. The composite robot collaborative processing system as described in claim 1, characterized in that: The processing unit includes: An optical transmission trigger probe module is used for online measurement and alignment of the material product; The wired five-axis tool setter module is used to measure the length, diameter, and runout of machining tools; A pneumatic material clamping module is used to pneumatically clamp and release the material product; wherein the pneumatic material clamping module can adjust its clamping force according to the material properties and process properties of the material product.

10. The composite robot collaborative processing system as described in any one of claims 1 to 9, characterized in that: The control and management unit adopts an MES system module combined with an IoT system module, which can be used to integrate, analyze and process data from the processing unit, the loading and unloading execution unit and the handling unit, as well as data from the production process.

Images