Intelligent control method and system for wall painting equipment

Abstract

The invention discloses an intelligent control method and system for wall painting equipment, and relates to the technical field of intelligent control. A plurality of sub wall painting nodes are determined according to recognition of a wall painting path; and determining a wall painting event of each wall painting sub-node based on the node positions of the wall painting sub-nodes, the corresponding wall painting sub-areas and the working state of the wall painting equipment, and determining an intelligent control system of the wall painting equipment according to the wall painting event of each wall painting sub-node, the real-time image of the wall body and the working parameters of the wall painting roller. And the accuracy of the intelligent control system of the wall painting equipment is improved. According to the technical scheme, the multiple sub-regulation items are determined based on the identification of the dynamic regulation event, the autonomous regulation event of the wall painting equipment is determined according to the regulation content of the multiple sub-regulation items, the corresponding regulation priority and the working state of the wall painting equipment, the accuracy of the autonomous regulation event of the wall painting equipment is improved, and the painting surface grade of the wall is dynamically balanced.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology, and in particular to an intelligent control method and system for a wall painting device. Background Technology

[0002] With the development of technology, wall painting equipment has been gradually applied to the construction field to paint walls. The wall painting equipment uses a brushing roller to paint the wall. In the existing technology, multiple images of the wall are collected, and the area to be painted is determined based on these images. The brushing roller then paints the area along a first path. However, this technology ignores the brushing events of each sub-brushing node and the working parameters of the brushing roller, which affects the accuracy of the intelligent control system of the wall painting equipment and results in low accuracy of the autonomous control events of the wall painting equipment. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art. This invention provides an intelligent control method and system for wall painting equipment.

[0004] This invention provides an intelligent control method for a wall painting device, comprising: The camera equipped with the wall painting equipment dynamically captures images of the wall and collects multiple wall images. Based on the recognition of multiple wall images, the corresponding area to be painted is determined. Multiple sub-brushing areas are determined based on the detection of the area to be brushed. The brushing path of the brushing roller is determined according to the location of the multiple sub-brushing areas, the corresponding area shape, and the brushing roller configured in the brushing equipment. Multiple sub-brushing nodes are identified based on the brushing path. The brushing events of each sub-brushing node are determined based on the node position, corresponding sub-brushing area, and working status of the brushing equipment. The intelligent control system of the brushing equipment is determined based on the brushing events of each sub-brushing node, the real-time image of the wall, and the working parameters of the brushing roller. The intelligent control system includes a set of control instructions, which is adjusted according to real-time feedback. In this intelligent control system, the set of working parameters for the wall painting equipment is determined based on the intelligent control system and the real-time position of the wall painting roller. Based on this set of working parameters, the paint supply module of the wall painting equipment, and the wall painting roller, the dynamic control events of the wall painting equipment are determined. The dynamic control events include execution timing, path tracking adjustment instructions, and supply adjustment instructions. Based on the identification of dynamic control events, multiple sub-control items are determined. According to the control content of multiple sub-control items, the corresponding control priority and the working status of the wall painting equipment, the autonomous control events of the wall painting equipment are determined, and the wall painting surface grade is dynamically balanced.

[0005] This invention provides an intelligent control system for a wall painting device, which is applied to the aforementioned intelligent control method for the wall painting device.

[0006] Compared with the prior art, the beneficial effects of the present invention are: (1) Multiple sub-brushing nodes are determined based on the identification of the brushing path. The brushing events of each sub-brushing node are determined based on the node position of the multiple sub-brushing nodes, the corresponding sub-brushing area and the working status of the brushing equipment. The intelligent control system of the brushing equipment is determined based on the brushing events of each sub-brushing node, the real-time image of the wall and the working parameters of the brushing roller. The brushing path is introduced to control the brushing events of each sub-brushing node. It takes into account the overall consideration of the brushing events of each sub-brushing node, the real-time image of the wall and the working parameters of the brushing roller, and improves the accuracy of the intelligent control system of the brushing equipment.

[0007] (2) Based on the identification of dynamic control events, multiple sub-control items are determined. The autonomous control events of the wall painting equipment are determined according to the control content of multiple sub-control items, the corresponding control priority and the working status of the wall painting equipment. The dynamic control events are controlled, realizing the overall consideration of the control content of multiple sub-control items, the corresponding control priority and the working status of the wall painting equipment. This improves the accuracy of the autonomous control events of the wall painting equipment and dynamically balances the wall painting surface grade. Attached Figure Description

[0008] Figure 1 This is a flowchart illustrating the intelligent control method for the wall painting device in an embodiment of the present invention; Figure 2 This is a flowchart illustrating step S11 of the intelligent control method for the wall painting device in this embodiment of the invention. Figure 3 This is a flowchart illustrating step S12 in the intelligent control method of the wall painting device in this embodiment of the invention. Figure 4 This is a flowchart illustrating step S13 in the intelligent control method of the wall painting device in this embodiment of the invention. Figure 5 This is a flowchart illustrating step S14 of the intelligent control method for the wall painting device in this embodiment of the invention. Figure 6 This is a flowchart illustrating step S15 of the intelligent control method for the wall painting device in this embodiment of the invention. Figure 7 This is a schematic diagram of the structural composition of the intelligent control system of the wall painting device in an embodiment of the present invention; Figure 8 This is a schematic diagram of the wall painting device in an embodiment of the present invention. Detailed Implementation

[0009] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0010] Please see Figures 1 to 8 A smart control method for a wall painting device, applied in a smart control scenario; the smart control method for the wall painting device includes: Step S11: The camera configured in the wall painting equipment dynamically captures images of the wall and collects multiple wall images. Based on the recognition of multiple wall images, the corresponding area to be painted is determined. Step S12: Based on the detection of the area to be painted, determine multiple sub-painting areas, and determine the painting path of the painting roller according to the location of the multiple sub-painting areas, the corresponding area shape, and the painting roller configured in the painting equipment. Step S13: Based on the identification of the wall brushing path, determine multiple sub-wall brushing nodes. Based on the node positions of the multiple sub-wall brushing nodes, the corresponding sub-wall brushing areas, and the working status of the wall brushing equipment, determine the wall brushing events of each sub-wall brushing node. Based on the wall brushing events of each sub-wall brushing node, the real-time image of the wall, and the working parameters of the wall brushing roller, determine the intelligent control system of the wall brushing equipment. Step S14: In this intelligent control system, the set of working parameters of the wall painting equipment is determined based on the intelligent control system and the real-time position of the wall painting roller. The dynamic control events of the wall painting equipment are determined based on the set of working parameters, the paint supply module of the wall painting equipment, and the wall painting roller. Step S15: Based on the identification of dynamic control events, determine multiple sub-control items, determine the autonomous control events of the wall painting equipment according to the control content of multiple sub-control items, the corresponding control priority and the working status of the wall painting equipment, and dynamically balance the wall painting surface grade.

[0011] refer to Figure 2 In step S11, the specific steps are as follows: S111: When the wall painting equipment is arranged relative to the wall, the camera configured on the wall painting equipment is facing the wall and dynamically shooting the wall. At this time, the camera shoots the wall at different positions to collect multiple wall images. The multiple wall images are input to the control module of the wall painting equipment. S112: In the control module of the wall painting equipment, a corresponding overall image is constructed based on the synthesis of multiple wall images, multiple wall surface features are determined based on the recognition of the overall image, and the corresponding area to be painted is determined based on the feature position of the multiple wall surface features, the corresponding feature information and the previous wall area.

[0012] In the embodiments of this application, the IMU provides the device's pitch, roll, and yaw angles in real time, ensuring that the angle between the camera's optical axis and the wall normal is within the allowable error range, thereby effectively avoiding image distortion caused by device tilt. At the same time, the wheel encoder accurately records the device's movement distance, providing a crucial positional reference for subsequent image stitching. During device initialization, this module integrates the IMU's angular velocity and the accelerometer's gravity vector to calculate the device's precise attitude relative to the direction of gravity. During device movement, the system measures the displacement increment by counting pulses of the orthogonal encoder and establishes a local Cartesian coordinate system with the initial position as the origin. Each frame of the acquired image is attached with a precise pose stamp containing its three-dimensional coordinates and rotation quaternions.

[0013] The camera itself needs to have automatic exposure and automatic white balance functions to adapt to the complex environment of uneven indoor lighting; the acquired raw image stream will also be immediately preprocessed, such as lens distortion correction, to eliminate barrel or pincushion distortion caused by wide-angle lenses; in specific execution, the module will drive the motion actuator of the equipment (such as a wheeled or tracked chassis) to move along a preset S-shaped or Z-shaped scanning path; during the movement, the industrial camera will capture images at a fixed frame rate or a trigger mode based on displacement increment (e.g., triggering once every 5 centimeters); crucially, the synchronization signal of image acquisition will be strictly timestamped with the pose data stream of the positioning module to ensure that each frame of image can accurately correspond to its spatial pose; the raw image data stream will be denoised, color corrected, and distortion corrected by a real-time image signal processor, and finally generate standardized image data packets.

[0014] By using DMA (Direct Memory Access), the camera's image data is directly written to the device's circular buffer, significantly reducing the CPU load. Each data packet contains an image header (including timestamp, resolution, and encoding format) and a pose header (including coordinates, attitude angle, and positional confidence). The data is transmitted to the receiving queue of the main control computer via Gigabit Ethernet. The daemon process on the main control side is responsible for managing the data packets, storing them in the memory pool, and promptly notifying the upper-layer algorithm (S112) that new data is available. This mechanism ensures the integrity and low latency of the data stream.

[0015] Specifically, the scenario is set as a wall painting device equipped with an industrial camera capable of two-dimensional rotation, a LiDAR, and a main control computer. Its task is to autonomously enter an empty room awaiting renovation, identify and plan the painting paths for the four walls.

[0016] The spatial positioning and attitude alignment module of the wall painting device starts working first; the IMU performs self-calibration, while the lidar quickly constructs a two-dimensional plan of the room through the SLAM algorithm and determines the initial position and orientation of the wall painting device on the map; at the same time, the pan-tilt unit adjusts the camera to the preset initial attitude, such as a pitch angle of 0 degrees, to ensure that the camera's optical axis is perpendicular to the wall it is about to face.

[0017] The multi-viewpoint dynamic scanning and image acquisition module of the wall-painting device begins to take the lead. Based on the preliminary map generated by the LiDAR, it autonomously plans a scanning path that moves along the wall baseline. Specifically, the wall-painting device moves at a constant speed of 0.2 meters per second along the wall edge. During the acquisition process, the camera on the pan-tilt unit takes a picture every 15 degrees in the vertical direction, performing a vertical scan of the wall surface from bottom to top. At the same time, the wall-painting device repeats this vertical scan sequence every 10 centimeters it moves forward. In this way, for a wall that is 3 meters wide and 2.5 meters high, the wall-painting device will acquire approximately 510 images, and each image will accurately include the (x, y) coordinates of the wall-painting device on the map, the (pitch, yaw) angle of the camera, and the pose information such as the timestamp.

[0018] The data transmission and cache management module ensures that this massive amount of data is properly processed; the camera transmits uncompressed RAW format image data to the main control computer on the platform in real time through the GigEVision interface; a real-time operating system runs on the main control computer, a high-priority process is responsible for receiving data and storing it in a huge memory buffer; another low-priority background thread is responsible for organizing this data into a structured dataset according to time sequence and pose information.

[0019] Furthermore, through feature point matching techniques such as SIFT and ORB, and bundle adjustment algorithms, the system can accurately optimize the pose of each image and calculate the 3D point cloud of the scene. With the help of advanced image fusion techniques such as multi-band fusion or Poisson fusion, the textures of images from different perspectives are seamlessly mapped onto the 3D model, ultimately generating a digital wall model with realistic textures.

[0020] The system extracts local invariant features from all input images and performs cross-image matching based on the similarity of feature descriptors. Based on these matching point pairs, the system uses a random sampling consensus algorithm to eliminate mismatches, solves the fundamental matrix or essential matrix, and then recovers the relative poses between cameras. Through incremental or global bundle adjustment, the system jointly optimizes the pose parameters of all cameras and the coordinates of points in 3D space to minimize reprojection errors. Using the optimized poses, the system generates a mesh model through algorithms such as Poisson surface reconstruction, and fuses image information through texture mapping to construct a high-fidelity overall image.

[0021] The system primarily utilizes deep learning models, especially convolutional neural networks, to perform instance segmentation and object detection tasks. The model can accurately distinguish and label corners, doors and windows, switch panels, baseboards, cracks, stains, and areas with different materials. Technically, the system employs a pre-trained semantic segmentation model based on MaskR-CNN or U-Net architectures. The model takes an overall image as input (whether a 2D panoramic view or a rendered view of a 3D model) and outputs a pixel-level semantic label map, where each pixel is classified into predefined categories such as main walls, window frames, doors, obstacles, stained areas, or old paint layers. Simultaneously, the object detection sub-network outputs bounding boxes and confidence scores for locating large objects such as doors and windows.

[0022] Through logical reasoning and rule matching, the engine intelligently filters out areas requiring painting from complex wall environments. It receives structured data from the feature extraction module, such as {object:'door',bounding_box:[x1,y1,x2,y2],confidence:0.98}. The engine matches this data with rules in the knowledge base. For example, rule 1: if object:'window' is detected, its bounding_box area is marked as no_paint_zone; rule 2: if the confidence of object:'stain' is greater than 0.8, its bounding_box area is marked as prime_zone (requiring primer); rule 3: all pixels not marked as no_paint_zone and belonging to the main_wall category constitute base_paint_zone. Through this association and reasoning, the module ultimately outputs one or more areas to be painted, each with specific painting attributes.

[0023] Specifically, the wall painting equipment has acquired 510 pose-captured images covering the entire wall. The multi-view geometry and image fusion module of the wall painting equipment is activated. It utilizes the pose information of all images to construct a precise 3D point cloud of the wall through feature matching and bundle adjustment. The algorithm triangulates the point cloud to form a fine mesh model. The system seamlessly maps the texture information of the 510 images onto the mesh using weighted averaging and Laplacian pyramid fusion techniques. The result is that a rotatable digital 3D wall model corresponding to the real wall surface is generated in the control module of the wall painting equipment.

[0024] The 3D model was rendered into 2D images from several virtual viewpoints and input into its internal Mask R-CNN model. The model output the following results: it detected a window, with the instance segmentation mask accurately outlining the window frame and glass with a confidence level of 99%; it detected a light switch, with the bounding box accurately located with a confidence level of 95%; at the bottom of the wall, it identified a mold stain of about 30x40 cm, which was labeled as "stain" with a confidence level of 92%; the entire wall was labeled as "wall", but the window, switch, and stain area were explicitly excluded; at the same time, the bottom edge of the wall was identified as "skirting board".

[0025] It accesses the built-in prior knowledge base, which contains the rule: the window, switch, and skitting_board areas are no_paint_zone; it also executes another rule: the stain area is prime_zone, with the job parameters being: primer spraying, 2 coats; it applies the core rule: in all wall areas, the parts not designated as no_paint_zone or prime_zone constitute the standard_paint_zone, with the job parameters being: topcoat spraying, 1 coat.

[0026] The control module of the wall painting equipment no longer contains a jumble of images, but a structured set of operation instructions. This set of instructions includes: the area to be painted, 1, a polygon covering the entire wall but excluding windows, switches, baseboards, and stains, marked as standard_paint_zone; the area to be painted, 2, a polygon precisely covering mold stains, marked as prime_zone; and the non-operation area, where the precise location and shape of windows, switches, and baseboards are clearly marked.

[0027] refer to Figure 3 In step S12, the specific steps are as follows: S121: Perform area detection on the area to be painted, determine multiple sub-painting areas based on the area detection of the area to be painted, mark the area position of each sub-painting area, and determine the corresponding area shape based on the shape detection of each sub-painting area. S122: Monitor the wall painting equipment in real time and mark the wall painting rollers configured on the wall painting equipment. Determine the wall painting path of the wall painting rollers based on the current position of the wall painting rollers, the regional positions of multiple sub-wall painting areas and the corresponding regional shapes.

[0028] In the embodiments of this application, the system receives the wall-painting area (such as standard_paint_zone and prime_zone) output by S112, and combines it with the wall surface features (such as door and window, switch, etc. boundaries) identified in S112 to perform advanced semantic segmentation. Its key capability is to distinguish areas that require different processing techniques and instantiate each independent area. In terms of technical implementation, the system adopts a deep learning-based instance segmentation network (such as MaskR-CNN) as its core. Its input is the pixel mask of the wall-painting area and the semantic feature map of the entire wall surface. The network, through convolutional layers and region proposal networks, not only classifies each pixel, but also distinguishes different instances belonging to the same category. For example, it can identify two separate prime_zones (stain areas) as independent instances. Its final output is a set of instantiated masks, each mask representing an independent sub-painting area, and is accompanied by a semantic label (such as standard_paint, prime_paint).

[0029] For each instance mask, a contour-finding algorithm (such as the Suzuki algorithm) is applied to extract its boundary pixel set; this point set is then approximated with a polygon (such as the Douglas-Peucker algorithm) to obtain a simplified polygon boundary; based on this polygon, the system can calculate the area, perimeter, and geometric center; simultaneously, the centroid of the region is determined by calculating the moments of the pixels; the module also calculates the minimum rotational bounding rectangle to determine the main orientation and dimensions of the region; each sub-region is encoded as a structured data object containing semantic labels, a sequence of polygon vertices, centroid coordinates, and area, etc.

[0030] This module includes a morphological analysis engine; it determines whether a region is convex by calculating the convex hull of a polygon and comparing the difference between the original polygon and the convex hull; by analyzing the self-intersection or inclusion relationship of polygons, it can identify complex shapes such as polygons with holes or separated polygons; based on the morphological classification results, the system selects the corresponding template from the preset path strategy library; at the same time, the module queries the process database and assigns the corresponding operation parameters based on the semantic tags of the region.

[0031] Specifically, after step S112 is completed, the device has identified two areas to be painted: a standard_paint_zone that covers most of the wall but excludes windows and switches, and a prime_zone that covers mold and stains. The multimodal area segmentation and instantiation module of the painting device is activated, receiving the two area masks output from S112. For the standard_paint_zone, due to the presence of windows and switches, the area is naturally divided into a main part and several small isolated parts, which the module recognizes as a single, complex instance with holes. For the prime_zone, the module recognizes it as an independent, simple instance. The output is two instantiated sub-regions: SubArea_1, with the semantic label standard_paint, and a shape of a polygon with holes; and SubArea_2, with the semantic label prime_paint, and a shape of a simple convex polygon.

[0032] For SubArea_1, the module extracts its outer contour polygon (defining the working boundary of the wall) and inner contour polygon (defining the avoidance boundary of windows and switches), and calculates its centroid coordinates as (x=1.5m, y=1.2m) and its area as 12.5 square meters; for SubArea_2, the module extracts its outer contour polygon (an approximate rectangle), and calculates its centroid coordinates as (x=0.8m, y=0.5m) and its area as 0.12 square meters; the output results are two structured data objects containing all geometric parameters and precise coordinate information.

[0033] Analyzing the geometric data of SubArea_1, it is determined to be a non-convex polygon with holes. Based on this shape, the system selects a contour-fill hybrid path strategy from the path strategy library. Analyzing SubArea_2, it is determined to be a simple convex polygon, and the system selects a simple reciprocating path strategy for it. Based on semantic tags, the module assigns values ​​to SubArea_1: {paint_type:latex,passes:1,roller_pressure:standard}; Assign values ​​to SubArea_2: {paint_type:oil_based_primer,passes:2,roller_pressure:high}; The control module of the wall painting equipment has a clear and quantitative description of the wall painting task: Sub-painting area 1 is a large area with holes, located in the center of the wall, which needs to be painted with latex paint once, using a complex path strategy; Sub-painting area 2 is a small rectangular area, located below the corner of the wall, which needs to be painted with oil-based primer twice.

[0034] Furthermore, the wall painting equipment is monitored in real time, and the wall painting rollers configured by the wall painting equipment are marked. The wall painting path of the wall painting rollers is determined based on the current position of the wall painting rollers, the regional positions of multiple sub-wall painting areas and the corresponding regional shapes. This takes into account the overall consideration of the current position of the wall painting rollers, the regional positions of multiple sub-wall painting areas and the corresponding regional shapes, ensuring the accuracy of the wall painting path of the wall painting rollers.

[0035] At this point, state estimation algorithms such as extended Kalman filtering or particle filtering are used to construct a system state vector containing the device's two-dimensional coordinates (x, y), heading angle θ, and the vertical lifting height h and tilt angle α of the wall-painting roller. The filter integrates the scan matching pose increment from the lidar, the wheel speed integral from the encoder, and the angular velocity and acceleration data from the IMU through weighted averaging to correct the accumulated error and finally output a pose estimation result. At the same time, the linear displacement of the lifting mechanism (such as a ball screw) is accurately measured by a joint encoder to calibrate the three-dimensional coordinates of the end of the wall-painting roller.

[0036] The global planner employs a sampling-based algorithm to quickly search for a feasible path from the starting point to the ending point within the configuration space formed by sub-regions, primarily addressing the macroscopic problem of how to reach the next region. The local planner and the coverage path generation part, for each sub-region, call different coverage path generation algorithms based on its shape (such as convex polygons or regions with holes as determined by S121). For convex polygons, simple bow-shaped or spiral paths are generated. For regions with holes, paths along the outer contour are generated first, followed by internal filling paths, ensuring a safe distance between the paths and the hole boundaries. The path optimizer smooths the generated paths (e.g., using B-spline interpolation) and introduces a cost function based on a dynamic model to optimize the curvature and velocity variations of the paths.

[0037] It employs time-scaled algorithms such as S-curve acceleration / deceleration planning or trapezoidal velocity curve planning; it uses the arc length of the path as the independent variable to generate a continuous and smooth velocity and position function over time. This process requires full consideration of the kinematic model of the equipment (such as the kinematic constraints of differential drive) and dynamic model (such as the maximum torque limit of the lifting motor); its final output is a high-frequency control command sequence, for example, sending a linear velocity and angular velocity command to the chassis motion controller every 20 milliseconds, and a target position and velocity command to the lifting motor controller.

[0038] Specifically, after step S121 is completed, the device has obtained detailed descriptions of two sub-regions: SubArea_1 (a large area with holes) and SubArea_2 (a small rectangular area). Now, the wall-painting device is positioned 0.5 meters in front of the wall, ready to begin operation. The multi-sensor fusion and pose estimation module of the wall-painting device is activated. The LiDAR scans the wall and the surrounding environment, and the initial coordinates of the wall-painting device in the room map are determined by the SLAM algorithm as (x=0.5, y=0.0, θ=0°). The encoder of the lifting mechanism shows that the wall-painting roller is at its lowest point, with a height of h=0.1m. As the device begins to move, the module continuously fuses the data from various sensors at a frequency of 100Hz, and outputs the three-dimensional coordinates of the end of the wall-painting roller in real time.

[0039] The constraint-based path generation and optimization module begins its work, deciding to first process SubArea_2 (the small area) closest to the starting point. In local planning, since SubArea_2 is a simple convex polygon, the module generates a vertical bow-shaped covering path for it, with the path spacing set to 25cm based on the width of the wall paint roller (e.g., 30cm) to ensure a 10% overlap rate. In global planning, after processing SubArea_2, the module plans a transition path from its endpoint to the starting point of SubArea_1. In local planning, for SubArea_1 (the area with holes), the module calls a complex algorithm to generate a safe contour path along the boundaries of windows and switches, generating a bow-shaped filling path inside it. The algorithm ensures that all path points maintain a safe distance of 15cm from the obstacle boundaries. The optimization module performs B-spline smoothing on all path segments and optimizes the turning radius at the connections to make the device movement smoother.

[0040] The trajectory interpolation module under kinematic and dynamic constraints receives the optimized complete path and parameterizes it in time according to the physical parameters of the wall painting equipment (maximum linear velocity 0.3 m / s, maximum angular velocity 1 rad / s, lifting speed 0.2 m / s). For example, for a straight path, it generates a trapezoidal velocity curve: acceleration-constant speed-deceleration; for a turn, it calculates a smooth angular velocity change curve to ensure that it does not exceed the motor torque limit. Its final output is a high-frequency command stream, for example, at t=0.0s, the command is v=0, ω=0, h_tar When get=0.1m; at t=0.5s, the command is v=0.1m / s, ω=0, h_target=1.2m (the equipment moves forward while the paint roller rises); at t=2.0s, the command is v=0.3m / s, ω=0, h_target=1.2m (the equipment moves at a constant speed along the path of SubArea_2); at t=5.0s, the command is v=0, ω=0.5rad / s, h_target=1.2m (the equipment turns, preparing to enter SubArea_1), and so on, until the work in all areas is completed.

[0041] refer to Figure 4 In step S13, the specific steps are as follows: S131: Collect the wall painting path, determine the corresponding sub-wall painting nodes based on the wall painting path and multiple sub-wall painting areas, and mark the node positions of multiple sub-wall painting nodes; at the same time, collect multiple working data of the wall painting equipment. S132: Determine the working status of the wall-painting equipment based on the identification of multiple working data of the wall-painting equipment; determine the wall-painting events of each sub-wall-painting node based on the node positions of multiple sub-wall-painting nodes, the corresponding sub-wall-painting areas, and the working status of the wall-painting equipment. S133: Obtain real-time images of the wall, determine the first level of intelligent control content based on the real-time images of the wall and the wall-painting events of each sub-wall-painting node, determine the second level of intelligent control content based on the real-time images of the wall and the working parameters of the wall-painting roller, and determine the intelligent control system of the wall-painting equipment based on the first level of intelligent control content and the second level of intelligent control content.

[0042] In the embodiments of this application, the wall painting path is collected, and the corresponding sub-wall painting nodes are determined based on the wall painting path and multiple sub-wall painting areas to mark the node positions of multiple sub-wall painting nodes; at the same time, multiple working data of the wall painting device are collected, which takes into account the overall consideration of the wall painting path and multiple sub-wall painting areas, and ensures the accuracy of the corresponding sub-wall painting nodes.

[0043] At this point, for straight segments in the path, uniform sampling based on distance is used (for example, setting a node every 80% of the radius of the brush roller); for curved segments or turning points, adaptive sampling based on curvature is used, increasing the node density in areas with high curvature to ensure the accuracy of path tracking; each generated node object contains a unique ID, a precise three-dimensional spatial coordinate (x, y, z), and a pointer or index to the corresponding sub-region data structure in S121; all nodes are organized into a spatial index data structure.

[0044] The device subscribes to data streams from various sensors via an internal fieldbus (such as CAN bus or EtherCAT). The data sources cover the power system (current, voltage, temperature, and speed of the chassis drive motor; position, speed, and output torque of the lifting servo motor), the execution system (speed and torque of the paint roller drive motor; speed and outlet pressure of the paint pump motor), the supply system (level sensor readings, viscometer readings, and temperature sensor readings of the paint tank), and the sensing system (three-axis acceleration and angular velocity output from the IMU). The module uses a high-precision hardware clock as the time reference, accurately timestamps all acquired data packets, and manages them through a circular buffer to ensure the real-time performance and integrity of the data.

[0045] The spatialized nodes are associated with time-aligned device status data to form a structured, self-contained data packet, providing a complete context for the event definition of S132. Whenever the path-node mapping module generates a new sub-painting node, the module immediately extracts the closest complete data snapshot to the current moment from the circular buffer of the multi-source data acquisition module. It creates a composite data structure, which we call the node status snapshot. This snapshot object contains the node's unique identifier (NodeID), three-dimensional coordinates (NodePosition), a pointer to the associated sub-area (AssociatedSubArea), a data snapshot timestamp (DeviceStateTimestamp), and nested objects (ChassisStatus, ActuatorStatus, SupplyStatus) that contain all parameters of the chassis, actuator, and supply system, respectively.

[0046] Specifically, after step S122 is completed, the device has planned a complete wall painting path, including a detailed path in SubArea_2 (the moldy stain area). The path-node mapping and spatial indexing module of the wall painting device begins to process the path of SubArea_2. Since this is a rectangular area that requires fine-grained operation, the module uses distance-based dense sampling, generating a node every 3 centimeters. The output result is that the module generates 40 nodes, from Node_201 to Node_240. Each node is precisely marked with its position on the wall and is associated with the data structure of SubArea_2. For example, the coordinates of Node_205 are (x=0.85m, y=0.55m, z=0.6m), and it knows that it belongs to SubArea_2.

[0047] The multi-source heterogeneous data acquisition and synchronization module continuously acquires data at a frequency of 200Hz. At the instant the wall painting device moves to the corresponding position of Node_205 (assuming t=10.5s), the module captures the following data snapshots: the current of the chassis left wheel motor is 3.2A and the speed is 120rpm; the position of the lifting motor is 0.6m and the torque is 1.5Nm; the pressure of the paint pump is 2.1MPa and the speed is 300rpm; the paint tank level is 78% and the viscosity is 85KU; all of these data are stamped with a precise timestamp of t=10.500s.

[0048] The node-state association and data encapsulation module is triggered because the device is about to execute Node_205; it immediately creates a node state snapshot object for Node_205; the encapsulation result is a structured data packet containing the NodeID "Node_205", the 3D coordinates of NodePosition, the associated SubArea_2, the timestamp "10.500s", and a complete nested object containing ChassisStatus (such as the current of the left and right wheel motors), ActuatorStatus (such as the position and torque of the lifting motor, pump pressure), and SupplyStatus (such as the liquid level and viscosity).

[0049] Furthermore, the working status of the wall-painting equipment is determined based on the identification of multiple working data of the wall-painting equipment; the wall-painting events of each sub-wall-painting node are determined based on the node positions of multiple sub-wall-painting nodes, the corresponding sub-wall-painting areas, and the working status of the wall-painting equipment. This comprehensive consideration of the node positions of multiple sub-wall-painting nodes, the corresponding sub-wall-painting areas, and the working status of the wall-painting equipment ensures the accuracy of the wall-painting events of each sub-wall-painting node.

[0050] At this point, the massive amount of operational data collected by S131 is comprehensively analyzed to determine the overall operating status of the equipment. It doesn't simply list the data, but rather uses data fusion and a rule engine to arrive at a highly generalized status conclusion, such as normal, standby, fault, or performance degradation. Technically, this module contains a status evaluation machine that receives snapshots of the node status from S131 and applies a series of predefined evaluation rules. For example, the power system rule is: IF(motor_current>rated_current×1.2)AND(motor_temperature>80°C)THENstatus=overload_warning; the supply system rule is: IF(pump_pressure...<setpoint×0.8)OR(viscosity> upper_limit)THENstatus=supply_anomaly; The comprehensive status rule will further determine: IF(any_subsystem_status==fault)THENglobal_status=faultELSEIF(any_subsystem_status==warning)THENglobal_status=degradedELSEglobal_status=nominal; Through the cascading judgment of these rules, the module finally outputs a structured working status object.

[0051] Using each sub-painting node as a processing unit, and combining the node's spatial information, task information (from S121), and the equipment's current operating status (from the previous module), a specific, executable painting event is generated. This event serves as the bridge connecting planning and execution. Technically, the engine employs a template- and rule-based event generation model. For each sub-painting node, it queries the context, obtaining the node's NodePosition and AssociatedSubArea, and extracts process requirements from the AssociatedSubArea, such as {paint_type:primer,passes:2,targ}. et_thickness:80μm}; It acquires the status, reads the global_status output by the multi-dimensional status fusion and evaluation module and the status of each subsystem; The engine searches for matching rules in the event rule base, and the entries in the rule base are in the form of: IF(context==pattern)AND(status==pattern)THENevent=template; The engine fills the parameters in the template with specific values ​​to generate a final wall-painting event object; For example, for a node located in prime_zone and with a device status of nominal, it will generate an event containing specific execution parameters.

[0052] Specifically, after step S131 generates a complete node state snapshot for Node_205, the multi-dimensional state fusion and evaluation module receives the data snapshot of Node_205 and begins to execute the evaluation rules. It detects that the chassis motor current is 3.2A, which is lower than 120% of the rated current, and the temperature is normal, so the status is determined to be nominal. The lifting motor position is 0.6m, and the torque is 1.5Nm, which is within the normal operating range, so the status is nominal. The paint pump pressure is 2.1MPa, which is higher than the preset 1.8MPa, but lower than the overpressure threshold, so the status is nominal. The paint tank level is 78%, and the viscosity is 85KU, both of which are within the normal range, so the status is nominal. Combining the status of all subsystems, the module finally outputs a working status object: {global_status:nominal,chassis_status:nominal,actuator_status:nominal,supply_status:nominal}.

[0053] The context-aware event generation engine begins working for Node_205; it queries the context to obtain SubArea_2 associated with Node_205, whose process requirements are {paint_type:primer,passes:2,target_thickness:80μm,roller_pressure:high,linear_speed:slow}; the engine obtains the status and reads that the device's global_status is nominal; the engine finds a matching rule in the event rule base: IF(sub_area.paint_type==primer)AND(global_status==nominal) THENgenerate_event(primer_coating_normal); The engine calls the primer_coating_normal template and fills it with specific parameters to generate the final painting event E205. This event object contains EventID, NodeID, triggering conditions, a detailed sequence of actions (such as opening the primer valve, setting the pump pressure, setting the roller speed and pressure, and setting the chassis speed), expected results (such as coating thickness and surface quality), and interruption conditions (such as pump pressure failure or roller motor stall). For each node on the path, the painting device no longer only has position and status data, but also a complete painting event containing specific action instructions, expected results, and interruption conditions.

[0054] Therefore, by acquiring real-time images of the wall and determining the first level of intelligent control based on these images and the brushing events of each sub-brushing node, and then determining the second level of intelligent control based on the real-time images of the wall and the working parameters of the brushing roller, the intelligent control system of the brushing equipment is determined. This system considers both levels of intelligent control, ensuring the accuracy of the intelligent control system. Furthermore, by introducing a brushing path to control the brushing events of each sub-brushing node, the system considers the brushing events of each sub-brushing node, the real-time images of the wall, and the working parameters of the brushing roller, further improving the accuracy of the intelligent control system.

[0055] At this point, a real-time image of the wall surface that the wall roller has just applied is acquired using a high-resolution industrial camera and a structured light sensor. It uses image processing algorithms (such as photometric stereo method and laser triangulation) to reconstruct the micro-three-dimensional morphology of the coating, thereby quantitatively evaluating the coating thickness, uniformity, gloss, and whether there are defects such as sagging, orange peel, and missed coating. These quantitative indicators are compared in real time with the ExpectedOutcome (such as {coating_thickness:80μm}) in the S132 wall painting event. If a thickness deviation exceeding ±10μm or obvious defects are detected, the module will generate the first level of control content, such as: {control_type:quality_feedback,adjustment:{pump_flow_rate:+5%,roller_speed:-10%}}.

[0056] The system reads various operating parameters of the wall painting roller in real time, including the speed and torque of the roller drive motor, the contact pressure between the roller and the wall (inferred from the pressure sensor or motor torque), and the amount of paint adhering to the roller surface (inferred from the high-frequency vibration sensor or optical sensor). It compares these parameters with the target values ​​set in the ActionSequence of the S132 wall painting event. For example, if the roller torque is detected to be continuously higher than the threshold, indicating excessive pressure, the module will generate a second layer of control: {control_type:process_feedback,adjustment:{roller_pressure_actuator:-15%}}. If excessive speed fluctuations are detected, instructions to adjust the motor PID parameters will be generated.

[0057] An arbitration mechanism based on priority and weighting is adopted. It receives control content from the first two modules and merges it according to preset rules. In priority arbitration, the second level of control (proprioception) which involves equipment safety and process stability usually has higher priority than the first level of control (vision) which involves quality optimization. For example, if the proprioception module requests to reduce pressure to prevent overload, while the vision module requests to increase pressure to improve coverage, the arbitrator will prioritize executing the pressure reduction instruction. In weighted fusion, when the two control contents do not conflict but are in the same direction (such as both requesting to reduce speed), the arbitrator will perform weighted calculations based on their respective deviations to obtain a more accurate adjustment amount. The arbitrator encapsulates the merged instructions, along with the original wall-painting event, into a complete intelligent control system object. The intelligent control system object contains a set of control instructions, which is adjusted according to real-time feedback and directly sent to the underlying motion controller and actuator.

[0058] Specifically, when the wall painting device moves to the position of Node_205 and S132 generates a wall painting event E205 for it, the wall painting device begins to execute the event. The camera of the vision-guided quality control module is aimed at the work area corresponding to Node_205. By analyzing the real-time image, the module finds that due to slightly higher local water absorption of the wall surface, the current coating thickness is 72μm, which is lower than the expected 80μm for the E205 event. Therefore, the module generates the first layer of control content: {control_type:quality_feedback,target_node:Node_205,deviation:thickness_low,adjustment:{pump_flow_rate:+8%}}.

[0059] The module reads the real-time operating parameters of the wall-painting roller; it finds that the torque of the roller drive motor is 1.8 Nm, slightly higher than the upper limit of the normal range in event E205, indicating that the contact pressure is too high; therefore, the module generates a second layer of control: {control_type:process_feedback,target_node:Node_205,deviation:torque_high,adjustment:{roller_pressure_actuator:-10%}}.

[0060] During the arbitration process, the module determined that excessive torque was a process stability issue affecting equipment lifespan, while insufficient thickness was a quality problem affecting the final result. Based on preset rules, process stability had a slightly higher priority. In the fusion decision, the module decided to immediately execute the pressure reduction command of the second level of control to stabilize the process. Simultaneously, it also incorporated the flow increase command of the first level of control into the plan, but slightly reduced its adjustment range (e.g., from +8% to +5%) to prevent excessive flow increase after pressure reduction from causing new problems (such as splashing). The module generated a dynamic control system based on the E205 event, but with two added real-time correction rules: Rule 1 (Process): IF roller_torque>1.7NmTHENdecreaseroller_pressureby10%; Rule 2 (Quality): IF coating_thickness<75μmTHENincreasepump_flow_rateby5%. This dynamic system was issued in real-time to guide the operation of the wall-painting equipment at Node_205 and subsequent nodes.

[0061] refer to Figure 5 In step S14, the specific steps are as follows: S141: Real-time monitoring intelligent control system, determine the paint supply module according to the overall form of the intelligent control system and the wall painting equipment, mark the current working mode of the paint supply module, and determine the set of working parameters of the wall painting equipment according to the current working mode of the paint supply module, the real-time position of the wall painting roller and the intelligent control system. S142: Determine the working parameters of the wall brush roller and the paint supply module based on the matching of the working parameter set, the wall brush roller and the paint supply module, and determine the first level of dynamic control content according to the working parameters of the wall brush roller and the wall brushing path; S143: Determine the second level of dynamic control content based on the working parameters of the paint supply module and the wall painting path, and determine the dynamic control event of the wall painting equipment based on the first level of dynamic control content and the second level of dynamic control content.

[0062] In the embodiments of this application, a real-time monitoring intelligent control system is used to determine the paint supply module based on the overall form of the intelligent control system and the wall painting equipment, and to mark the current working mode of the paint supply module. Based on the current working mode of the paint supply module, the real-time position of the wall painting roller, and the intelligent control system, the set of working parameters for the wall painting equipment is determined. This approach takes into account the current working mode of the paint supply module, the real-time position of the wall painting roller, and the overall considerations of the intelligent control system, ensuring the accuracy of the set of working parameters for the wall painting equipment.

[0063] At this point, it is necessary to accurately obtain the overall shape of the equipment at the current moment. This is not just a position, but a complete spatial configuration that includes the poses of all movable parts (such as lifting arm, telescopic sleeve, and gimbal). By subscribing to the output of S13 through a high-speed internal bus, it is ensured that any updates to the control system (such as new correction rules) can be obtained in real time. At the same time, it uses the forward kinematics model of the equipment and integrates the data of each joint encoder (lifting, rotation, telescopic) to calculate in real time the precise pose of the end of the wall brush roller in the global coordinate system, as well as the shape vector of the entire mechanical structure of the equipment. This shape vector can be represented as: M=[base_pose,lift_extension,arm_rotation,boom_angle,…].

[0064] Based on the current task requirements (from the control system) and the physical form of the equipment (from form perception), a complex rule engine selects the optimal working mode for the paint supply module from a variety of preset working modes. Technically, this module internally implements a multi-condition decision tree or fuzzy logic reasoning system. Its inputs include task context (extracted from the control system, such as paint_type (primer / topcoat), application_method (spraying / rolling), target_area (ceiling / wall)) and form context (extracted from form vectors, such as lift_extension (height) and arm_angle (tilt angle)). The decision rule base contains the following rules: IF(paint_type==primer)AND(application_method==spray)THENmode=HIGH_PRESSURE_AIRLESS; IF(lift_extension>3.0m)THENmode=PRESSURE_COMPENSATED; The module, through inference, ultimately outputs a definite operating mode identifier, such as Mode_HP_Airless_Comp.

[0065] Based on the newly determined operating mode of the supply module and combined with the macro-instructions of the intelligent control system, a global, preliminary set of operating parameters is generated. This set serves as the benchmark and constraint for subsequent refined adjustments in S142 and S143. This module maintains a mode-parameter mapping database. Once the decision module determines the mode Mode_HP_Airless_Comp, the module queries this database to obtain a set of basic parameter templates, such as: {pump_base_pressure:180bar,fluid_heater:on,agitator_speed:medium}; It integrates macro-level commands from the S13 control system. For example, if the control system requires an overall deceleration of 5%, this module will adjust the speed-related parameters accordingly. It generates a structured set of operating parameters, which provides target operating points for all subsystems (chassis, lifting, paint roller, pump, etc.), ensuring global coordination of equipment behavior.

[0066] Specifically, when the wall-painting device reaches Node_205, after S13 generates a dynamic control system for it, the control system activation and shape perception module immediately loads the dynamic control system (containing two correction rules) generated by S13 for Node_205; simultaneously, it reads the encoder data of each joint of the device and calculates the current overall shape vector through the kinematic model: M=[base_pose(x=0.85,y=0.55),lift_extension=1.2m,arm_angle=0°,…].

[0067] The rule-based supply module pattern decision module begins operation; it extracts the task context from the control system: paint_type:primer, application_method:spray; it extracts the morphological context from the morphological vector: lift_extension=1.2m (belonging to the normal operation height); the module matches the highest priority rule in the rule base: IF(paint_type==primer)AND(application_method==spray)AND(lift_extension<2.5m)THENmode=HIGH_PRESSURE_AIRLESS_STANDARD; The module determines that the current working mode of the paint supply module is Mode_HP_Airless_Std.

[0068] The parameter set generation and coordination module queries the parameter database based on the Mode_HP_Airless_Std mode to obtain the basic template: {pump_base_pressure:200bar,fluid_heater:off,agitator_speed:low}; It integrates the macroscopic commands of the S13 control system; the quality feedback rules in the control system require an increase in flow rate of 5%, which this module converts into pump pressure adjustment parameters; the module generates a complete set of operating parameters and sends them to S142 and S143 as a baseline, which includes the supply module mode, global constraints (such as maximum chassis speed), and target parameters for each subsystem (pump system, roller system, chassis system).

[0069] Furthermore, the working parameters of the wall brush roller and the paint supply module are determined based on the matching of the working parameter set, the wall brush roller and the paint supply module. The first level of dynamic control content is determined according to the working parameters of the wall brush roller and the wall brushing path, which takes into account the overall consideration of the working parameters of the wall brush roller and the wall brushing path, and ensures the accuracy of the first level of dynamic control content.

[0070] At this point, the system contains a matching engine based on a physical model or empirical data (such as a response surface fitted through a large number of experiments); its inputs are parameters such as {pump_base_pressure, roller_target_rpm} in the S141 parameter set; the engine ensures that the supplied paint flow rate (determined by pump pressure) matches the coating capacity of the roller (determined by rotation speed, pressure, and roller material) by solving a set of coupled equations, avoiding oversupply (leading to dripping and splashing) or undersupply (leading to uneven coating and dry brushing); its output is a set of calibrated and coordinated execution parameters, such as {roller_calibrated_rpm:58, pump_calibrated_pressure:2.05MPa}.

[0071] The system employs a model predictive control algorithm; internally, it contains a predictive model based on the kinematics and dynamics equations of the wall-painting device. In each control cycle (e.g., every 20 milliseconds), the module performs the following steps: acquiring the current pose (from real-time monitoring in S131) ​​and velocity of the wall-painting roller; predicting the device's trajectory with the current control input within a short future time window (e.g., the next second); comparing the predicted trajectory with the planned path in S12, calculating the future position, velocity, and acceleration errors; solving an optimization problem online, aiming to minimize future errors, control input variations, and energy consumption, while satisfying physical constraints (e.g., maximum acceleration, motor torque limits); the solution result is an optimal control input sequence (e.g., the driving torque of the left and right wheels).

[0072] Specifically, after the wall painting device generates a global set of working parameters for the execution of Node_205 via S141, the parameter matching and calibration module receives the parameter set from S141, which includes {pump_base_pressure:205, roller_target_rpm:55}. The module queries its internal primer-air spraying matching model, which takes into account the current ambient temperature of 25°C, the primer viscosity of 85KU, and the specific type of velvet roller used by the wall painting device. After calculation, the model determines that at a pressure of 205 bar, a rotation speed of 55 rpm is slightly high to achieve uniform coating, resulting in slight splashing. To achieve the best match, calibration is required. The module outputs the calibrated parameters: {roller_calibrated_rpm:52, pump_calibrated_pressure:205}. This new set of parameters will serve as the benchmark for path tracking control.

[0073] The path tracking prediction control module activates its MPC controller, preparing to guide the wall-painting device into the path segment corresponding to Node_205. In the status acquisition phase, the module reads the current wall-painting roller pose as (x=0.848, y=0.551), speed as 0.15 m / s, and heading angle as 1°. In the trajectory prediction phase, the module predicts that if the current control input is maintained, the device's position after 0.5 seconds will be (x=0.923, y=0.559). In the error calculation phase, the module queries the rules in S12. The path was drawn, and the target position after 0.5 seconds was found to be (x=0.925, y=0.560). The predicted position had a 1mm deviation on the Y-axis. More importantly, the path had a left turn after 1 second, and the current heading angle was insufficient to smoothly enter the turn. In the optimization solution stage, the MPC optimizer started working, with the goals of: 1) eliminating the 1mm lateral deviation; 2) smoothly adjusting the heading angle to prepare for the turn; 3) minimizing the disturbance to the linear velocity; and 4) ensuring that the motor torque does not exceed the safety limit.

[0074] After the optimizer solves the problem, it outputs the first level of dynamic control: a fine control command that includes fine-tuning of the torque of the left and right wheels of the chassis (e.g., left wheel +0.2Nm, right wheel -0.2Nm) and fine-tuning of the roller speed (e.g., -1rpm). This command will cause the equipment to have a slight tendency to turn to the left, while slightly reducing the roller speed to match the upcoming turn, thereby achieving smooth path tracking.

[0075] Therefore, the second level of dynamic control content is determined based on the working parameters of the paint supply module and the wall painting path. The dynamic control events of the wall painting equipment are determined based on the first and second levels of dynamic control content. This approach takes into account both the first and second levels of dynamic control content, ensuring the accuracy of the dynamic control events of the wall painting equipment.

[0076] At this point, the system receives in real-time the wall-painting path generated by S12 (as a feedforward signal) and the wall-painting roller operating parameters calibrated by S142 (such as roller_calibrated_rpm and linear velocity). Through a kinematic-hydrodynamic coupled model, it can calculate the paint demand under different path conditions: in straight sections, the demand flow rate Q = k × v × w (where k is the paint transfer coefficient, v is the linear velocity, and w is the roller width); in turning sections, the linear velocity of the outer roller increases and the inner roller decreases, causing a change in the total demand flow rate. The model can accurately calculate the dynamic change in demand ΔQ_turn based on the radius of curvature R of the path; in acceleration and deceleration sections, speed changes cause a lag in flow demand, and the model predicts this transient response through a transfer function; the module finally outputs a demand flow rate prediction curve for a future time window.

[0077] The feedforward control section directly uses the output Q_predicted from the demand forecasting module as the setpoint. It calculates the required target pressure P_target using the pump's flow model (e.g., pressure-flow characteristic curve) and sends it directly to the pump's driver. This allows the system to adjust in advance before demand changes, significantly improving response speed. Its feedback control section includes a closed-loop feedback loop to eliminate model errors and external disturbances (such as changes in paint viscosity). It reads the actual flow rate Q_actual in real time using a flow meter installed on the pipeline and compares it with Q_predicted. A PID controller then fine-tunes P_target. The module outputs a second level of dynamic control, for example: {control_type:supply_feedforward,adjustment:{pump_pressure_setpoint:P_target+ΔP_feedback}}.

[0078] The arbitrator receives control commands from S142 and itself, and processes them according to the following principles: Time synchronization, ensuring that the two control commands are precisely aligned in time. For example, the deceleration command for path tracking must occur synchronously with the pressure reduction command of the supply system; Conflict detection and resolution, detecting whether there are logical conflicts. For example, S142 requires acceleration to improve efficiency, while the supply module predicts that acceleration will lead to insufficient flow, thus requiring acceleration to be limited. The arbitrator makes a decision based on a preset priority (usually quality over efficiency) and rejects the acceleration command; Coordination enhancement, if the two commands are coordinated (such as both requiring deceleration), the arbitrator will perform weighted fusion to generate a better adjustment amount; The arbitrator encapsulates all the fused commands, along with triggering conditions, timestamps, and other information, into a complete DynamicRegulationEvent object.

[0079] Specifically, when the wall-painting equipment is at Node_205, after S142 has generated the first level of dynamic control (fine-tuning in preparation for turning), the supply and demand prediction module based on path characteristics analyzes that the wall-painting equipment will enter a left turn with a radius of 0.5m in 0.5 seconds. Based on the roller_calibrated_rpm:52 calibrated by S142 and the current linear velocity, the module calculates the demand flow rate for the current straight segment as Q_line=150ml / min. Through its turning model, the module predicts that during the turning process, due to the increase in the linear velocity of the outer roller, the total demand flow rate will instantaneously increase by 8%, reaching Q_turn=162ml / min. The module generates a demand flow rate prediction curve, showing a step increase in flow rate demand at 0.5 seconds.

[0080] The feedforward-feedback composite controller of the supply system receives the predicted curve. In the feedforward stage, based on Q_turn = 162 ml / min, the controller queries the pump's characteristic curve, calculates that the target pressure needs to be increased from the current 2.05 MPa to 2.15 MPa, and immediately generates a feedforward command. In the feedback stage, the flow meter detects an actual flow rate of 148 ml / min, which deviates from the current target value by 2 ml / min. The PID controller calculates a small pressure compensation amount ΔP_feedback = +0.02 MPa. The module generates the second level of dynamic control: {control_type:supply_feedforward,adjustment:{pump_pressure_setpoint:2.15+0.02=2.17MPa,timing:execute_in_0.45s}}.

[0081] The multi-channel event arbitration and fusion module receives the first level of control content from S142 (for chassis fine-tuning during turns) and the second level of control content from S143 itself (pre-pressurization). In the time synchronization stage, the module precisely aligns the execution time of the chassis fine-tuning with the execute_in_0.45s of the pressurization command. In the conflict detection stage, no conflict was found, and the two commands are coordinated actions, both aimed at ensuring the quality of operation during turns. In the event encapsulation stage, the module merges the two commands to generate the final dynamic control event E_Final_205, which includes precise execution timing, path tracking adjustment commands (such as fine-tuning of left and right wheel torque), and supply adjustment commands (such as setting the pump pressure to 2.17MPa at a specific time).

[0082] refer to Figure 6 In step S15, the specific steps are as follows: S151: Identify dynamic control events and determine multiple sub-control items based on the identification of dynamic control events, and determine the control priority of multiple sub-control items based on the multiple sub-control items and the current working task of the wall painting equipment; S152: Determine the first autonomous control coefficient based on the control priority of multiple sub-control items and the working status of the wall painting equipment; determine the second autonomous control coefficient based on the control content of multiple sub-control items and the working status of the wall painting equipment; and determine the autonomous control event of the wall painting equipment based on the mapping relationship between the first autonomous control coefficient, the second autonomous control coefficient, and the autonomous control event. S153: Collect surface images of the wall, determine surface compensation measures based on the surface images of the wall and the autonomous control events of the wall painting equipment, optimize the surface painting grade of the wall based on the surface compensation measures, the wall painting roller and the surface images of the wall, so as to dynamically balance the surface painting grade of the wall.

[0083] In the embodiments of this application, dynamic control events are identified, and multiple sub-control items are determined based on the identification of dynamic control events. The control priority of multiple sub-control items is determined based on the multiple sub-control items and the current working task of the wall painting equipment. This approach takes into account the overall consideration of multiple sub-control items and the current working task of the wall painting equipment, ensuring the accuracy of the control priority of multiple sub-control items.

[0084] At this point, the dynamic adjustment events output by S14 are deeply analyzed. The concurrent instructions contained in the events, targeting different subsystems, are decomposed into the most basic, indivisible atomic operation units. This process ensures the refinement and accuracy of subsequent decisions. In terms of technical implementation, an event parser reads the JSON or similar structured data of the dynamic adjustment events. It decomposes the compound instructions according to a predefined executor mapping table. For example, a chassis_adjustment instruction that includes torque adjustment for both the left and right wheel motors will be decomposed into two independent sub-items: adjust_left_wheel_motor_torque and adjust_right_wheel_motor_torque. Similarly, supply_adjustment will be decomposed into adjust_pump_pressure_setpoint, control_agitator_speed, etc. Finally, a list containing all sub-adjustment items is output.

[0085] For each atomized sub-control item, a quantified priority weight is calculated. This weight is not fixed but dynamically generated, based on the context of the currently executing task, such as whether the task is a rapid large-area primer coating or a fine local topcoat coating. Technically, this module implements a multi-attribute decision model, such as the analytic hierarchy process (AHP) or a fuzzy logic system. It defines multiple evaluation attributes for each sub-item, such as the impact on quality (the degree of influence of the control on final coating thickness, uniformity, gloss, etc.), the impact on efficiency (the degree of influence of the control on overall operation speed, energy consumption, etc.), and the impact on safety (whether the control involves equipment stability, structural safety, etc.). The module assigns different weights to these attributes according to the current task type; for example, in a fine topcoat task, the weight for the impact on quality is the highest; in a rapid primer task, the weight for the impact on efficiency will be correspondingly increased. Finally, through weighted summation, a priority weight value between 0 and 1 is output for each sub-item.

[0086] The system receives all sub-control items and their corresponding priority weights, sorts them from highest to lowest weight, and generates an ordered execution sequence. This sequence serves as the input for S152 to calculate the autonomous control coefficient. The system uses a simple sorting algorithm (such as quicksort or heapsort) to rearrange the list of sub-control items. The sorting key is the priority weight calculated in the previous step. For items with the same weight, secondary sorting rules can be introduced, such as executing safety-related items first, then quality-related items, and finally efficiency-related items. The final output is a clear queue of sub-control items arranged in descending order of importance.

[0087] Furthermore, a first autonomous control coefficient is determined based on the control priority of multiple sub-control items and the working status of the wall painting equipment. A second autonomous control coefficient is determined based on the control content of multiple sub-control items and the working status of the wall painting equipment. The autonomous control event of the wall painting equipment is determined based on the mapping relationship between the first autonomous control coefficient, the second autonomous control coefficient, and the autonomous control event. This approach takes into account the overall consideration of the mapping relationship between the first autonomous control coefficient, the second autonomous control coefficient, and the autonomous control event, ensuring the accuracy of the autonomous control event of the wall painting equipment.

[0088] At this time, the key operating status parameters of the wall painting equipment are monitored in real time, such as the state of charge of the power battery, the temperature and vibration spectrum of the key drive motor and pump motor, and the stress sensor readings of the mechanical structure. These parameters are input into a health function H(state), which outputs a health index between 0 and 1. At the same time, the control priority P in S151 is introduced as a weight. The first autonomous control coefficient α1 is calculated through a weighting function, for example: α1=f(H(state),P). A typical implementation is: α1=H(state)×(0.5+0.5×P), which means that when the equipment is in poor condition, α1 will decrease, and the instructions will be discounted. At the same time, for tasks with higher priority, α1 is closer to 1, and the equipment is more willing to do its best.

[0089] The analysis focuses on the content of the sub-control items, particularly their adjustment range. For example, a command to instantly increase pump pressure by 20% carries a higher risk than a command to increase it by 5%. The engine, considering the current state of the equipment, quantifies the risk using a risk function R(content,state). For instance, when the motor temperature is already high, a command to increase torque carries a higher risk. Simultaneously, the module queries a historical database to obtain the historical success rate S of similar control content. The second autonomous control coefficient α2 is calculated using a risk-reward function, for example: α2=g(R(content,state),S). A typical implementation is: α2=(1-R)×S, which means that the higher the risk or the lower the historical success rate, the smaller α2, and the more conservative the equipment is when executing the command.

[0090] The module receives the sorted list of sub-control items from S151, along with α1 and α2 calculated for each item. It calculates a total autonomous control coefficient α_total using a fusion function (usually a weighted multiplication), for example: α_total = w1 × α1 + w2 × α2 (where w1 and w2 are preset weights, usually w1 > w2, because device safety is more important). For each sub-control item, the module multiplies its original adjustment amount by α_total to obtain the final, autonomously discretionary adjustment amount. The module repackages all adjusted sub-items to generate the final autonomous control event.

[0091] Therefore, by acquiring surface images of the wall and determining surface compensation measures based on these images and the autonomous control events of the painting equipment, the wall surface grade is optimized using these compensation measures, the painting roller, and the wall surface images. This dynamic balance of the wall surface grade takes into account both the wall surface images and the autonomous control events of the painting equipment, ensuring the accuracy of the surface compensation measures. Simultaneously, by controlling the dynamic control events, the overall consideration of the control content, corresponding control priorities, and working status of multiple sub-control items is achieved, improving the accuracy of the autonomous control events of the painting equipment and dynamically balancing the wall surface grade.

[0092] At this point, a multimodal imaging system is integrated, including a high-resolution 2D camera and a structured light or laser profilometer. After executing an autonomous control event, the system immediately scans the work area. Using computer vision algorithms, such as photometric stereo or phase-shift profilometry, the micro-3D topography of the wall surface is reconstructed. A deep learning-based defect segmentation network (such as a variant of U-Net++) is applied to the reconstructed surface data and 2D images to identify and segment various defects at the pixel level, such as pinholes, runs, orange peel, and uneven thickness. The module outputs a structured quality report containing quantitative data such as the type, location, size, and severity (e.g., depth, area) of the defects.

[0093] The defect report output by the high-precision quality inspection module is subjected to spatiotemporal correlation analysis with the detailed parameters of the recently executed autonomous control event (such as pump pressure, roller speed, and travel speed). Through a pre-trained defect-cause mapping model (which is a decision tree or gradient boosting model), the root cause can be inferred. For example, when the coating thickness of a region is detected to be below the threshold, the correlation analysis finds that the region corresponds to the instant when the pump_pressure_adjustment command is executed. The causal inference model will then infer that the cause is that the pump pressure response lags behind the increase in travel speed, and a precise compensation measure will be generated accordingly.

[0094] The compensation measures are transformed into one or more compensation control events and inserted into the upcoming S15 processing flow, or directly triggered an immediate fine-tuning. This process forms an iterative optimization loop: execution > detection > analysis > compensation > re-execution. The module maintains a global surface quality map and updates the grade of each area in real time. Through a state estimation algorithm based on Kalman filtering or particle filtering, it can predict and smooth the quality distribution of the entire wall surface, dynamically adjust the control strategy of subsequent areas to eliminate quality differences between areas, and ultimately achieve dynamic balance and overall improvement of the wall painting surface grade.

[0095] For wall painting equipment, when the ball screw is rotating, the lifting seat rises and falls with the rotation of the ball screw, and the wall painting roller dynamically paints different positions on the wall as the lifting seat rises and falls; the paint mixing rod rotates synchronously under the drive of the linkage module and the ball screw to dynamically mix the paint in the paint storage tank, thus realizing the linkage between the wall painting roller and the paint mixing rod. This allows the wall painting roller and the paint mixing rod to work together to mix the paint in the paint storage tank while painting the wall, avoiding the wall painting component and the paint mixing component being independent of each other, and improving the linkage of the wall painting equipment.

[0096] Please see Figure 7 , Figure 7 This is a schematic diagram of the structural composition of the intelligent control system of the wall painting device in this embodiment of the invention; the intelligent control system of the wall painting device includes: The wall-to-be-painted area module 21 is used by the camera configured in the wall-painting device to dynamically capture images of the wall and collect multiple wall images, and to determine the corresponding wall-to-be-painted area based on the recognition of multiple wall images. The wall brushing path module 22 is used to determine multiple sub-wall brushing areas based on the detection of the area to be brushed, and to determine the wall brushing path of the wall brushing roller according to the area location, corresponding area shape and wall brushing roller configured in the wall brushing device. The intelligent control system module 23 is used to determine multiple sub-brushing nodes based on the identification of the brushing path, determine the brushing events of each sub-brushing node based on the node position of the multiple sub-brushing nodes, the corresponding sub-brushing area and the working status of the brushing equipment, and determine the intelligent control system of the brushing equipment based on the brushing events of each sub-brushing node, the real-time image of the wall and the working parameters of the brushing roller. The dynamic control event module 24 is used to determine the set of working parameters of the wall painting equipment based on the intelligent control system and the real-time position of the wall painting roller in the intelligent control system, and to determine the dynamic control events of the wall painting equipment based on the set of working parameters, the paint supply module of the wall painting equipment and the wall painting roller. The autonomous control module 25 is used to identify multiple sub-control items based on the recognition of dynamic control events. It determines the autonomous control events of the wall painting equipment according to the control content of multiple sub-control items, the corresponding control priority and the working status of the wall painting equipment, and dynamically balances the wall painting surface grade.

[0097] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A smart control method for a wall painting device, characterized in that, include: The camera equipped with the wall painting equipment dynamically captures images of the wall and collects multiple wall images. Based on the recognition of multiple wall images, the corresponding area to be painted is determined. Multiple sub-brushing areas are determined based on the detection of the area to be brushed. The brushing path of the brushing roller is determined according to the location of the multiple sub-brushing areas, the corresponding area shape, and the brushing roller configured in the brushing equipment. Multiple sub-brushing nodes are identified based on the brushing path. The brushing events of each sub-brushing node are determined based on the node position, corresponding sub-brushing area, and working status of the brushing equipment. The intelligent control system of the brushing equipment is determined based on the brushing events of each sub-brushing node, the real-time image of the wall, and the working parameters of the brushing roller. The intelligent control system includes a set of control instructions, which is adjusted according to real-time feedback. In this intelligent control system, the set of working parameters for the wall painting equipment is determined based on the intelligent control system and the real-time position of the wall painting roller. Based on this set of working parameters, the paint supply module of the wall painting equipment, and the wall painting roller, the dynamic control events of the wall painting equipment are determined. The dynamic control events include execution timing, path tracking adjustment instructions, and supply adjustment instructions. Based on the identification of dynamic control events, multiple sub-control items are determined. According to the control content of multiple sub-control items, the corresponding control priority and the working status of the wall painting equipment, the autonomous control events of the wall painting equipment are determined, and the wall painting surface grade is dynamically balanced.

2. The intelligent control method for the wall painting equipment according to claim 1, characterized in that, The wall painting device is equipped with a camera that dynamically captures images of the wall and obtains multiple wall images. Based on the recognition of these multiple wall images, the corresponding area to be painted is determined, including: When the wall painting equipment is positioned relative to the wall, the camera equipped with the wall painting equipment faces the wall and dynamically captures images of the wall. At this time, the camera captures images of the wall at different positions to collect multiple wall images, which are then input to the control module of the wall painting equipment. In the control module of the wall painting equipment, a corresponding overall image is constructed based on the synthesis of multiple wall images. Multiple wall surface features are determined based on the recognition of the overall image. The corresponding area to be painted is determined based on the feature position of the multiple wall surface features, the corresponding feature information, and the previous wall area.

3. The intelligent control method for the wall painting equipment according to claim 1, characterized in that, The process of determining multiple sub-painting areas based on the detection of the area to be painted, and determining the painting path of the painting rollers according to the location, corresponding shape, and painting roller configuration of the painting equipment, includes: The area to be painted is detected, and multiple sub-painting areas are determined based on the area detection. The location of each sub-painting area is marked, and the corresponding area shape is determined based on the shape detection of each sub-painting area. The system monitors the wall painting equipment in real time and marks the wall painting rollers configured on the equipment. Based on the current position of the wall painting rollers, the regional positions of multiple sub-wall painting areas, and the corresponding regional shapes, the system determines the wall painting path of the wall painting rollers.

4. The intelligent control method for the wall painting equipment according to claim 1, characterized in that, The process involves identifying multiple sub-brushing nodes based on the brushing path, determining brushing events for each sub-brushing node based on its node position, corresponding sub-brushing area, and the operating status of the brushing equipment, and establishing an intelligent control system for the brushing equipment based on these events, real-time images of the wall, and the operating parameters of the brushing roller. This includes: The wall painting path is collected, and the corresponding sub-wall painting nodes are determined based on the wall painting path and multiple sub-wall painting areas to mark the node positions of multiple sub-wall painting nodes; at the same time, multiple working data of the wall painting equipment are collected. The working status of the wall-painting equipment is determined by identifying multiple working data of the wall-painting equipment; the wall-painting events of each sub-wall-painting node are determined by the node positions of multiple sub-wall-painting nodes, the corresponding sub-wall-painting areas, and the working status of the wall-painting equipment.

5. The intelligent control method for the wall painting equipment according to claim 4, characterized in that, The process of identifying multiple sub-brushing nodes based on the brushing path, determining brushing events for each sub-brushing node based on its node position, corresponding sub-brushing area, and the working status of the brushing equipment, and determining an intelligent control system for the brushing equipment based on the brushing events of each sub-brushing node, real-time images of the wall, and working parameters of the brushing roller, further includes: The system acquires real-time images of the wall, determines the first level of intelligent control based on the real-time images and the brushing events of each sub-brushing node, determines the second level of intelligent control based on the real-time images of the wall and the working parameters of the brushing roller, and determines the intelligent control system of the brushing equipment based on the first and second levels of intelligent control.

6. The intelligent control method for the wall painting equipment according to claim 1, characterized in that, In this intelligent control system, the set of operating parameters for the wall-painting equipment is determined based on the intelligent control system and the real-time position of the wall-painting roller. Based on this set of operating parameters, the paint supply module of the wall-painting equipment, and the wall-painting roller, dynamic control events for the wall-painting equipment are determined, including: The system monitors the intelligent control system in real time, determines the paint supply module based on the overall shape of the intelligent control system and the wall painting equipment, marks the current working mode of the paint supply module, and determines the set of working parameters of the wall painting equipment based on the current working mode of the paint supply module, the real-time position of the wall painting roller, and the intelligent control system.

7. The intelligent control method for the wall painting equipment according to claim 6, characterized in that, In this intelligent control system, the set of working parameters for the wall-painting equipment is determined based on the intelligent control system and the real-time position of the wall-painting roller. Based on this set of working parameters, the paint supply module of the wall-painting equipment, and the wall-painting roller, dynamic control events for the wall-painting equipment are determined. The system also includes: The working parameters of the wall brush roller and the paint supply module are determined based on the matching of the working parameter set, the wall brush roller and the paint supply module. The first level of dynamic control content is determined according to the working parameters of the wall brush roller and the wall brushing path. The second level of dynamic control content is determined based on the working parameters of the paint supply module and the wall painting path. The dynamic control events of the wall painting equipment are then determined based on the first and second levels of dynamic control content.

8. The intelligent control method for the wall painting equipment according to claim 1, characterized in that, The process involves identifying multiple sub-control items based on dynamic control events, determining the autonomous control events of the wall painting equipment based on the control content, corresponding control priorities, and working status of the wall painting equipment, and dynamically balancing the wall painting surface grade, including: The system identifies dynamic control events and determines multiple sub-control items based on these events. The control priority of these sub-control items is then determined based on the current working tasks of the wall-painting equipment.

9. The intelligent control method for the wall painting equipment according to claim 8, characterized in that, The method of identifying multiple sub-control items based on the recognition of dynamic control events, determining the autonomous control events of the wall painting equipment according to the control content of the multiple sub-control items, their corresponding control priorities, and the working status of the wall painting equipment, and dynamically balancing the wall painting surface grade, also includes: The first autonomous control coefficient is determined based on the control priority of multiple sub-control projects and the working status of the wall painting equipment. The second autonomous control coefficient is determined based on the control content of multiple sub-control projects and the working status of the wall painting equipment. The autonomous control event of the wall painting equipment is determined based on the mapping relationship between the first autonomous control coefficient, the second autonomous control coefficient and the autonomous control event. The surface image of the wall is acquired, and the surface compensation measures for wall painting are determined based on the surface image of the wall and the autonomous control events of the wall painting equipment. The surface painting grade of the wall is optimized based on the surface compensation measures, the wall painting roller and the surface image of the wall to dynamically balance the surface painting grade of the wall.

10. An intelligent control system for a wall painting device, characterized in that, The intelligent control system of the wall painting equipment is applied to the intelligent control method of the wall painting equipment as described in any one of claims 1-9.

Images