Path planning method of glass curtain wall cleaning robot and glass curtain wall cleaning system

Abstract

The present application relates to glass curtain wall cleaning, in particular to a path planning method of a glass curtain wall cleaning robot and a glass curtain wall cleaning system, global map construction and task initialization, global path planning dominated by a mother warehouse, pose calibration, the robot travels along the theoretical path, the mother warehouse monitors whether the robot appears pose drift in real time, if pose drift appears, a calibration return path is planned, local dynamic obstacle avoidance and intelligent return, the robot detects the front environment in real time during travel, if a dynamic obstacle is found, local obstacle avoidance path planning is started, after the robot completes the obstacle avoidance action, the mother warehouse plans an obstacle avoidance return path, cleaning quality verification and optimization, the mother warehouse detects and judges the cleaning effect, if there is residual stain, a local intensive cleaning path is planned until the visual verification passes, the present application can overcome the obvious defects of the prior art in continuous operation ability, path planning accuracy and adaptability, system cooperation and cleaning quality closed loop control.

Description

Technical Field

[0001] This invention relates to glass curtain wall cleaning, and more specifically to a path planning method for a glass curtain wall cleaning robot and a glass curtain wall cleaning system. Background Technology

[0002] Cleaning the glass curtain walls of high-rise and super high-rise buildings is a crucial, high-risk, and costly task in urban building maintenance. Currently, negative pressure suction cleaning robots have gradually replaced traditional manual suspended platform operations, becoming the mainstream technology in this field and significantly improving operational safety. However, existing robot-based glass curtain wall cleaning technologies still have a series of shortcomings, hindering further improvements in efficiency, reliability, and intelligence. These shortcomings are specifically manifested in the following aspects:

[0003] First, the energy supply model is limited, making long-term continuous operation impossible. Most existing cleaning robots are powered by onboard batteries. Limited by the energy density and weight of these batteries, the single-cycle working time of cleaning robots is typically short (generally 2-4 hours), making it difficult to meet the needs of continuous cleaning of ultra-large glass curtain walls. Frequent interruptions for charging or battery replacement not only significantly reduce work efficiency but also increase operational complexity and safety risks. Although some research has attempted to introduce solar panels for auxiliary power supply, their unstable power output and significant susceptibility to weather conditions prevent them from serving as a primary energy solution.

[0004] Secondly, the low level of intelligence in path planning leads to cleaning blind spots and poor efficiency. Existing cleaning robots employ relatively rudimentary path planning strategies, mostly relying on pre-set fixed programs (such as simple zigzag or spiral paths). This approach lacks intelligent perception and adaptation to the actual glass curtain wall structure, failing to effectively handle complex facade structures (such as dense window frames, irregularly shaped decorations, and billboards), easily creating cleaning blind spots or areas requiring repeated cleaning. More importantly, cleaning robots accumulate positioning errors during movement due to factors such as track slippage, and lack an effective online calibration mechanism. As operating time accumulates, the deviation between the actual trajectory and the planned path increases, ultimately leading to a significant decrease in cleaning coverage and even malfunctions due to collisions with obstacles.

[0005] Third, environmental perception and decision-making capabilities are isolated, resulting in poor system coordination. Existing systems typically use a single robot as the sole perception and decision-making unit, with limited onboard computing resources and the ability to acquire only limited information about the surrounding environment through local sensors (such as ultrasonic sensors and contact sensors). This "island-like" working mode has significant shortcomings: on the one hand, the cleaning robot cannot know its precise pose on the entire glass curtain wall, like "blind men touching an elephant"; on the other hand, it cannot effectively deal with dynamic obstacles not marked on the global map (such as suddenly opening windows, birds, etc.), and after performing local obstacle avoidance, it often relies on its own algorithm to plan a regression path, resulting in a narrow decision-making field of view, easy getting stuck in local optima or generating invalid paths, and low efficiency.

[0006] Fourth, there is a lack of a closed-loop verification and optimization mechanism for cleaning quality. Most current cleaning robots only have the function of "performing cleaning" and cannot objectively evaluate the cleaning effect. After the cleaning work is completed, manual visual inspection is still required to judge the cleaning effect. This prevents the entire process from forming an intelligent closed loop of "perception-decision-execution-verification." Furthermore, for stubborn stains that are not thoroughly cleaned, the cleaning robot cannot autonomously initiate a secondary cleaning operation, affecting the reliability and consistency of service quality.

[0007] In summary, existing glass curtain wall cleaning systems have significant shortcomings in terms of continuous operation capability, accuracy and adaptability of path planning, system synergy, and closed-loop control of cleaning quality. Therefore, there is an urgent need in this field for a novel solution that can overcome these technical bottlenecks and achieve precise, efficient, reliable, and fully automated glass curtain wall cleaning. Summary of the Invention

[0008] In view of the above-mentioned shortcomings of the existing technology, the present invention provides a path planning method and a glass curtain wall cleaning system for a glass curtain wall cleaning robot, which can effectively overcome the obvious deficiencies of the existing technology in terms of continuous operation capability, accuracy and adaptability of path planning, system coordination and closed-loop control of cleaning quality.

[0009] To achieve the above objectives, the present invention provides the following technical solution:

[0010] The path planning method for glass curtain wall cleaning robots includes the following steps:

[0011] S1. Global map construction and task initialization: The mother module scans the glass curtain wall to be cleaned, generates a global map, divides the global map into multiple sub-regions, and assigns cleaning tasks to the robot.

[0012] S2. Global path planning led by the mother warehouse: For each sub-region, a theoretical cleaning path is planned; for transfers between sub-regions, a theoretical movement path is planned and the theoretical path is sent to the corresponding robot.

[0013] S3, Pose Calibration: The robot travels along the theoretical path. The mother module monitors in real time whether the robot has pose drift. If pose drift occurs, a calibration return path is planned and sent to the robot.

[0014] S4. Local dynamic obstacle avoidance and intelligent return: The robot detects the environment in front of it in real time during the movement. If a dynamic obstacle is detected, local obstacle avoidance path planning is initiated. After the robot completes the obstacle avoidance action, the mother warehouse plans the obstacle avoidance return path and sends it to the robot.

[0015] S5. Cleaning quality verification and optimization: After the cleaning work of the sub-area is completed, the mother chamber detects and judges the cleaning effect. If there are residual stains, a local enhanced cleaning path is planned and sent to the robot until the visual verification is passed.

[0016] Preferably, in S1, global map construction and task initialization: the mother module scans the glass curtain wall to be cleaned, generates a global map, divides the global map into multiple sub-regions, and assigns cleaning tasks to the robot, including:

[0017] The mother chamber scans the glass curtain wall to be cleaned and generates a global map that includes cleanable areas, insurmountable obstacles, areas where negative pressure is unreliable, and areas suspected of being stained, using image recognition algorithms.

[0018] Based on the complexity, area, and number of robots in the global map, the mother warehouse uses a grid-based partitioning algorithm to divide the global map into multiple sub-regions and assign cleaning tasks to the robots.

[0019] Preferably, in S2, the global path planning is led by the mother warehouse: for each sub-region, a theoretical cleaning path is planned; for transfers between sub-regions, a theoretical movement path is planned, and the theoretical path is distributed to the corresponding robot, including:

[0020] For each sub-region, the mother warehouse uses an improved bow-shaped path template to plan the theoretical cleaning path, calculates the total path length of all possible starting points and directions, and selects the path with the lowest energy consumption and shortest empty run as the theoretical cleaning path.

[0021] For transfers between sub-regions, the mother warehouse uses the A* algorithm to calculate the optimal safe transfer route as the theoretical movement path;

[0022] The mother warehouse constructs a theoretical path based on the theoretical cleaning path and the theoretical movement path, and then issues it to the corresponding robot.

[0023] Preferably, in S3, pose calibration: the robot travels along the theoretical path, and the mother module monitors in real time whether pose drift occurs. If pose drift occurs, a calibration regression path is planned and sent to the robot, including:

[0024] During the robot's movement, the mother compartment monitors the robot's actual coordinates in real time and compares the actual coordinates with the theoretical path coordinates;

[0025] If the coordinate deviation exceeds the preset threshold, it is determined that the robot has a pose drift. The mother module immediately plans a calibration regression path from the robot's current position to the nearest point on the theoretical path and sends it to the robot to eliminate the accumulated error.

[0026] Preferably, in S4, local dynamic obstacle avoidance and intelligent return: the robot detects the environment ahead in real time during its movement. If a dynamic obstacle is detected, local obstacle avoidance path planning is initiated. After the robot completes the obstacle avoidance maneuver, the mother module plans the obstacle avoidance return path and sends it to the robot, including:

[0027] The robot detects the environment ahead in real time while moving. If it finds a dynamic obstacle that is not marked on the global map, it immediately starts a local planning algorithm based on the Dynamic Window Method (DWA) to plan a local obstacle avoidance path and avoid the obstacle.

[0028] After the robot completes the obstacle avoidance maneuver, it reports its pose information to the mothership. The mothership then plans an obstacle avoidance return path from the post-obstacle position to the nearest point on the theoretical path based on the latest global perspective, and sends it to the robot to guide it to seamlessly return to the theoretical path.

[0029] Preferably, in S5, cleaning quality verification and optimization involves: after completing the cleaning of the sub-area, the mother chamber detects and judges the cleaning effect. If residual stains are found, a localized enhanced cleaning path is planned and sent to the robot until visual verification is passed, including:

[0030] After the cleaning of the sub-area is completed, the mother chamber performs a second scan of the sub-area and uses image difference algorithm or grayscale analysis to detect and judge the cleaning effect.

[0031] If residual stains are detected, the mother compartment plans a high-density, localized, intensive cleaning path and sends it to the robot until visual verification is passed;

[0032] Among them, the localized enhanced cleaning path is a spiral path or a path with multiple round trips.

[0033] A glass curtain wall cleaning system includes a main chamber and at least one negative pressure suction cleaning robot. The main chamber is fixed to a rooftop platform or a dedicated track, serving as the system's control center and energy base. Specifically, it includes:

[0034] The global vision module, which includes a high-definition zoom camera group, an image processor, and a storage module, is used to acquire, process, and store image information of the glass curtain wall to be cleaned, and to monitor the robot's position and posture in real time.

[0035] The core central processing unit is a high-performance computer chip used to perform global map building and task initialization, path planning, pose calibration calculation, cleaning effect detection and judgment, and human-computer interaction logic.

[0036] The communication module establishes a low-latency, highly reliable data communication link with the robot for transmitting control commands, path data, and status information.

[0037] The power and energy management module has a built-in high-capacity energy storage battery or is connected to mains power. It provides continuous or intermittent power to the robot through a wireless power transmitter or a coiled lightweight flexible cable.

[0038] Preferably, the robot specifically includes:

[0039] The moving and adsorption module, comprising a tracked or wheeled walking mechanism, a vacuum fan, and a sealing device, is used for moving and adsorbing on the surface of the glass curtain wall to be cleaned;

[0040] The cleaning execution module includes a water spray device, a cleaning roller brush, and a scraper, which are used to complete the cleaning work.

[0041] The local perception module includes one or more of LiDAR, ultrasonic sensors, and infrared ranging sensors, and is used to detect the environment in front of the robot in real time during the robot's movement.

[0042] The slave control unit receives and executes control commands and path data issued by the mother unit, and at the same time performs local obstacle avoidance path planning based on local perception data to help the robot complete emergency obstacle avoidance.

[0043] The power receiving and communication module receives power from the mother warehouse and maintains bidirectional communication with it.

[0044] Compared with the prior art, the path planning method and glass curtain wall cleaning system of the glass curtain wall cleaning robot provided by the present invention have the following beneficial effects:

[0045] 1) Unlimited battery life: Energy is replenished through the mother chamber, which completely breaks the battery life bottleneck and effectively meets the needs of continuous cleaning of ultra-large glass curtain walls in a single operation.

[0046] 2) Global Precision Control: By utilizing the "overall monitoring perspective" of the mother compartment, the industry problem of cumulative positioning errors caused by factors such as track slippage during robot movement is solved, and the path execution accuracy reaches a new high, achieving true full coverage and no dead angle cleaning;

[0047] 3) Intelligent and efficient: The system has a complete intelligent closed loop of "perception-decision-execution-verification". It can not only plan the path, but also detect and judge the cleaning effect and direct the secondary cleaning, ensuring the reliability and consistency of the cleaning effect;

[0048] 4) Safe and reliable: The mother warehouse can centrally schedule and monitor multiple robots in real time, effectively avoiding the risk of collisions, and can intervene in time when anomalies occur, greatly improving safety;

[0049] 5) Strong collaboration: The system architecture is easy to expand. One master warehouse can manage and schedule multiple robots to work together at the same time, which greatly improves the overall work efficiency, especially suitable for cleaning tasks of glass curtain walls. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0051] Figure 1 This is a schematic diagram of the process of the present invention;

[0052] Figure 2 This is a schematic diagram illustrating the principle of planning the calibration regression path in this invention;

[0053] Figure 3 This is a schematic diagram illustrating the principle of local dynamic obstacle avoidance and intelligent regression in this invention.

[0054] Figure 4 This is a schematic diagram illustrating the principle of cleaning quality verification and optimization in this invention.

[0055] Figure 5 This is a schematic diagram of the mother chamber system in this invention. Detailed Implementation

[0056] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0057] The following describes the specific process of the path planning method for the glass curtain wall cleaning robot provided by this invention, using a concrete example (e.g.) Figure 1(As shown).

[0058] S1. Global Map Construction and Task Initialization: The mother module scans the glass curtain wall to be cleaned, generates a global map, divides the global map into multiple sub-regions, and assigns cleaning tasks to the robot, including:

[0059] The mother chamber scans the glass curtain wall to be cleaned and generates a global map through image recognition algorithms, which includes cleanable areas, insurmountable obstacles (such as window frames, keels, etc.), areas where negative pressure is unreliable (such as seams), and areas suspected of being stained.

[0060] Based on the complexity, area, and number of robots in the global map, the mother warehouse uses a grid-based partitioning algorithm to divide the global map into multiple sub-regions and assign cleaning tasks to the robots.

[0061] S2. Mother Warehouse-led Global Path Planning: For each sub-region, a theoretical cleaning path is planned; for transfers between sub-regions, a theoretical movement path is planned and distributed to the corresponding robots, including:

[0062] For each sub-region, the mother warehouse uses an improved bow-shaped path template to plan the theoretical cleaning path, calculates the total path length of all possible starting points and directions, and selects the path with the lowest energy consumption and shortest empty run as the theoretical cleaning path.

[0063] For transfers between sub-regions, the mother warehouse uses the A* algorithm to calculate the optimal safe transfer route as the theoretical movement path;

[0064] The mother warehouse constructs a theoretical path based on the theoretical cleaning path and the theoretical movement path, and then issues it to the corresponding robot.

[0065] S3, Pose Calibration: The robot travels along the theoretical path. The mother module monitors the robot in real time for pose drift. If pose drift occurs, a calibration regression path is planned and sent to the robot. Figure 2 As shown, it includes:

[0066] During the robot's movement, the mother compartment monitors the robot's actual coordinates in real time and compares the actual coordinates with the theoretical path coordinates;

[0067] If the coordinate deviation exceeds the preset threshold, it is determined that the robot has a pose drift. The mother module immediately plans a calibration regression path from the robot's current position to the nearest point on the theoretical path and sends it to the robot to eliminate the accumulated error.

[0068] S4. Local Dynamic Obstacle Avoidance and Intelligent Return: The robot continuously monitors its surroundings during operation. If a dynamic obstacle is detected (such as a suddenly opened window or a bird), local obstacle avoidance path planning is initiated. After the robot completes the obstacle avoidance maneuver, the mother module plans the obstacle avoidance return path and sends it to the robot. Figure 3 As shown, it includes:

[0069] The robot detects the environment ahead in real time while moving. If it finds a dynamic obstacle that is not marked on the global map, it immediately starts a local planning algorithm based on the Dynamic Window Method (DWA) to plan a local obstacle avoidance path and avoid the obstacle.

[0070] After the robot completes the obstacle avoidance maneuver, it reports its pose information to the mothership. The mothership then plans an obstacle avoidance return path from the post-obstacle position to the nearest point on the theoretical path based on the latest global perspective, and sends it to the robot to guide it to seamlessly return to the theoretical path.

[0071] S5. Cleaning Quality Verification and Optimization: After completing the cleaning of the sub-area, the mother chamber detects and judges the cleaning effect. If residual stains are found, a localized enhanced cleaning path is planned and sent to the robot until visual verification is passed. Figure 4 As shown, it includes:

[0072] After the cleaning of the sub-area is completed, the mother chamber performs a second scan of the sub-area and uses image difference algorithm or grayscale analysis to detect and judge the cleaning effect.

[0073] If residual stains are detected, the mother compartment plans a high-density, localized, intensive cleaning path and sends it to the robot until visual verification is passed;

[0074] Among them, the localized enhanced cleaning path is a spiral path or a path with multiple round trips.

[0075] Based on the path planning method for glass curtain wall cleaning robots disclosed above, this application also discloses a glass curtain wall cleaning system, including a mother chamber and at least one negative pressure suction cleaning robot. The mother chamber is fixed to a rooftop platform or a dedicated track, serving as the system's control center and energy base station. Figure 5 As shown, it specifically includes:

[0076] The global vision module, which includes a high-definition zoom camera group, an image processor, and a storage module, is used to acquire, process, and store image information of the glass curtain wall to be cleaned, and to monitor the robot's position and posture in real time.

[0077] The core central processing unit is a high-performance computer chip used to perform global map building and task initialization, path planning, pose calibration calculation, cleaning effect detection and judgment, and human-computer interaction logic.

[0078] The communication module establishes a low-latency, highly reliable data communication link with the robot for transmitting control commands, path data, and status information.

[0079] The power and energy management module has a built-in high-capacity energy storage battery or is connected to mains power. It provides continuous or intermittent power to the robot through a wireless power transmitter or a coiled lightweight flexible cable.

[0080] Robots specifically include:

[0081] The moving and adsorption module, comprising a tracked or wheeled walking mechanism, a vacuum fan, and a sealing device, is used for moving and adsorbing on the surface of the glass curtain wall to be cleaned;

[0082] The cleaning execution module includes a water spray device, a cleaning roller brush, and a scraper, which are used to complete the cleaning work.

[0083] The local perception module includes one or more of LiDAR, ultrasonic sensors, and infrared ranging sensors, and is used to detect the environment in front of the robot in real time during the robot's movement.

[0084] The slave control unit receives and executes control commands and path data issued by the mother unit, and at the same time performs local obstacle avoidance path planning based on local perception data to help the robot complete emergency obstacle avoidance.

[0085] The power receiving and communication module receives power from the mother warehouse and maintains bidirectional communication with it.

[0086] To better illustrate the technical solution of this application, a specific example will be used for detailed explanation below.

[0087] The mother module is fixedly installed on the rooftop platform of the building, powered by mains electricity, and houses a high-performance industrial computer (core central processing unit) and a 4K high-definition zoom camera (global vision module). The robot uses a tracked walking mechanism and negative pressure adsorption via a turbofan fan.

[0088] Implementation process:

[0089] 1) The mother-house camera scans the glass curtain wall to be cleaned, identifies all window frames and a noticeable bird droppings stain, generates a global map, and divides the global map into 4 sub-areas, which are assigned to robot No. 1 for cleaning in sequence;

[0090] 2) The mother warehouse industrial computer planned and issued the "bow"-shaped theoretical cleaning path for the first sub-area to robot No. 1, and the robot began its work;

[0091] 3) The industrial computer in the mother warehouse discovered through image recognition that the robot's actual coordinates were 15cm off to the east compared to the theoretical path coordinates. It then planned a short-distance calibration and regression path to the west and sent it to the robot to enable it to return accurately.

[0092] 4) During the robot's movement, the ultrasonic sensor detected an unknown protrusion (a small satellite antenna), and the robot then autonomously avoided the obstacle and went around it. After completing the obstacle avoidance action, the robot reported its pose information to the mother compartment.

[0093] 5) After receiving the pose information, the industrial computer in the mother warehouse plans an obstacle avoidance return path from the position after obstacle avoidance to the nearest point on the theoretical path, and sends it to the robot to guide it to seamlessly return to the theoretical path and continue the operation.

[0094] 6) After the cleaning of the first sub-area is completed, the camera in the mother compartment takes pictures of the sub-area for comparison. It finds that there are still bird droppings and stains. The robot is then instructed to go to the point and perform a spiral-shaped local intensive cleaning path with a diameter of 30cm to wipe the key area until the mother compartment visually confirms that it is clean.

[0095] 7) Repeat the above process until all sub-areas are cleaned.

[0096] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A path planning method for a glass curtain wall cleaning robot, characterized in that: It includes a mother chamber and at least one negative pressure suction cleaning robot. The mother chamber is fixed on the top platform of the building or on a dedicated track, serving as the control center and energy base station of the system. The mother chamber includes a global vision module, which contains a high-definition zoom camera group, an image processor and a storage module, used to collect, process and store image information of the glass curtain wall to be cleaned, and to monitor the position and posture of the robot in real time. Path planning methods include the following steps: S1. Global Map Construction and Task Initialization: The mother module scans the glass curtain wall to be cleaned, generates a global map, divides the global map into multiple sub-regions, and assigns cleaning tasks to the robot, including: The mother chamber scans the glass curtain wall to be cleaned and generates a global map containing cleanable areas, insurmountable obstacles, areas where negative pressure is unreliable, and areas suspected of being stained, using image recognition algorithms. Based on the complexity, area, and number of robots in the global map, the mother warehouse uses a grid-based partitioning algorithm to divide the global map into multiple sub-regions and assign cleaning tasks to the robots. S2. Global path planning led by the mother warehouse: For each sub-region, a theoretical cleaning path is planned; for transfers between sub-regions, a theoretical movement path is planned and the theoretical path is sent to the corresponding robot. S3, Pose Calibration: The robot travels along the theoretical path. The mother module monitors in real time whether the robot has pose drift. If pose drift occurs, a calibration return path is planned and sent to the robot. S4. Local Dynamic Obstacle Avoidance and Intelligent Return: The robot continuously detects its surroundings during operation. If a dynamic obstacle is detected, local obstacle avoidance path planning is initiated. After the robot completes the obstacle avoidance maneuver, the mother module plans the obstacle avoidance return path and sends it to the robot, including: The robot detects the environment ahead in real time while moving. If it finds a dynamic obstacle that is not marked on the global map, it immediately starts a local planning algorithm based on the Dynamic Window Method (DWA) to plan a local obstacle avoidance path and avoid the obstacle. After the robot completes the obstacle avoidance maneuver, it reports its pose information to the mother module. The mother module plans an obstacle avoidance return path from the position after obstacle avoidance to the nearest point on the theoretical path based on the latest global perspective, and sends it to the robot to guide it to seamlessly return to the theoretical path. S5. Cleaning quality verification and optimization: After the cleaning work of the sub-area is completed, the mother chamber detects and judges the cleaning effect. If there are residual stains, a local enhanced cleaning path is planned and sent to the robot until the visual verification is passed.

2. The path planning method for the glass curtain wall cleaning robot according to claim 1, characterized in that: In S2, the mother warehouse-led global path planning involves: planning a theoretical cleaning path for each sub-region; planning theoretical movement paths for transfers between sub-regions, and distributing these paths to the corresponding robots, including: For each sub-region, the mother warehouse uses an improved bow-shaped path template to plan the theoretical cleaning path, calculates the total path length of all possible starting points and directions, and selects the path with the lowest energy consumption and shortest empty run as the theoretical cleaning path. For transfers between sub-regions, the mother warehouse uses the A* algorithm to calculate the optimal safe transfer route as the theoretical movement path; The mother warehouse constructs a theoretical path based on the theoretical cleaning path and the theoretical movement path, and then issues it to the corresponding robot.

3. The path planning method for the glass curtain wall cleaning robot according to claim 2, characterized in that: In S3, pose calibration occurs as the robot travels along the theoretical path. The mother module monitors the robot in real time for pose drift. If pose drift occurs, a calibration regression path is planned and sent to the robot, including: During the robot's movement, the mother compartment monitors the robot's actual coordinates in real time and compares the actual coordinates with the theoretical path coordinates; If the coordinate deviation exceeds the preset threshold, it is determined that the robot has a pose drift. The mother warehouse immediately plans a calibration regression path from the robot's current position to the nearest point on the theoretical path and sends it to the robot to eliminate the accumulated error.

4. The path planning method for the glass curtain wall cleaning robot according to claim 1, characterized in that: S5 Cleaning Quality Verification and Optimization: After completing the cleaning of the sub-area, the mother chamber detects and judges the cleaning effect. If residual stains are found, a localized enhanced cleaning path is planned and sent to the robot until visual verification is passed, including: After the cleaning of the sub-area is completed, the mother chamber performs a second scan of the sub-area and uses image difference algorithm or grayscale analysis to detect and judge the cleaning effect. If residual stains are detected, the mother chamber plans a high-density localized intensive cleaning path and sends it to the robot until visual verification is passed; Among them, the localized enhanced cleaning path is a spiral path or a path with multiple round trips.

5. A glass curtain wall cleaning system, used to execute the path planning method of the glass curtain wall cleaning robot according to claim 1, characterized in that: The mother warehouse includes: The global vision module, which includes a high-definition zoom camera group, an image processor, and a storage module, is used to acquire, process, and store image information of the glass curtain wall to be cleaned, and to monitor the robot's position and posture in real time. The core central processing unit is a high-performance computer chip used to perform global map building and task initialization, path planning, pose calibration calculation, cleaning effect detection and judgment, and human-computer interaction logic. The communication module establishes a low-latency, highly reliable data communication link with the robot for transmitting control commands, path data, and status information. The power and energy management module has a built-in high-capacity energy storage battery or is connected to mains power. It provides continuous or intermittent power to the robot through a wireless power transmitter or a coiled lightweight flexible cable.

6. The glass curtain wall cleaning system according to claim 5, characterized in that: The robot specifically includes: The moving and adsorption module, comprising a tracked or wheeled walking mechanism, a vacuum fan, and a sealing device, is used for moving and adsorbing on the surface of the glass curtain wall to be cleaned; The cleaning execution module includes a water spray device, a cleaning roller brush, and a scraper, which are used to complete the cleaning work. The local perception module includes one or more of LiDAR, ultrasonic sensors, and infrared ranging sensors, and is used to detect the environment in front of the robot in real time during the robot's movement. The slave control unit receives and executes control commands and path data issued by the mother unit, and at the same time performs local obstacle avoidance path planning based on local perception data to help the robot complete emergency obstacle avoidance. The power receiving and communication module receives power from the mother warehouse and maintains bidirectional communication with it.

Images