A control system of a window cleaning robot and a path control method

Abstract

This invention relates to the field of window cleaning robot technology, and more particularly to a control system and path control method for a window cleaning robot. The system includes a multi-source sensor module, a weight adjustment module, a state estimation module, a motor drive module, a slip detection module, an adaptive recovery module, a path planning module, a motion decision module, an encrypted transmission module, an anomaly push module, and a client. The state estimation module is connected to the multi-source sensor module, the path planning module is connected to the state estimation module, and the motion decision module is connected to the path planning module. This invention solves the technical problems in existing technologies where sensor data fusion weights are typically configured in a fixed way, making it difficult to adapt to changes in the friction characteristics of different working surfaces. This leads to untimely or misjudged slip detection, and the lack of an adaptive recovery mechanism based on slip severity levels often results in only stopping the robot with an alarm or repeatedly executing preset actions, leading to the robot easily getting stuck or experiencing cleaning interruptions.

Description

Technical Field

[0001] This invention relates to the field of window cleaning robot technology, and in particular to a control system and path control method for a window cleaning robot. Background Technology

[0002] Window cleaning robots are service robots used for automatically cleaning vertical or inclined surfaces such as glass curtain walls and windows of buildings. Because their working environment is at height, window cleaning robots need reliable adsorption capabilities, stable motion control, and autonomous path planning. Currently, window cleaning robots typically use negative pressure adsorption or magnetic adsorption to attach to the glass surface and move via drive wheels or tracks. The control system generates motion control commands by fusing data from multiple sensors, enabling the robot to navigate and perform cleaning operations autonomously. With the widespread use of glass curtain walls in high-rise buildings, higher demands are placed on the intelligence level, operational reliability, and autonomous recovery capabilities of window cleaning robots.

[0003] In existing technologies, the sensor data fusion weights in control methods are usually configured in a fixed way, which makes it difficult to adapt to changes in the friction characteristics of different working surfaces. This can lead to untimely or misjudged slippage detection. Furthermore, the lack of an adaptive recovery mechanism based on the degree of slippage often results in only being able to stop the machine and issue an alarm or repeatedly execute preset actions, which can easily lead to the machine getting stuck or the cleaning process being interrupted. Summary of the Invention

[0004] The purpose of this invention is to provide a control system and path control method for a window cleaning robot, which solves the technical problems in the prior art where the sensor data fusion weights are usually configured in a fixed way, making it difficult to adapt to the changes in the friction characteristics of different working surfaces, thus causing untimely or misjudged slippage detection, and lacking an adaptive recovery mechanism based on the degree of slippage, often only able to stop the machine and alarm or repeatedly execute preset actions, resulting in easy getting stuck or interrupting cleaning.

[0005] To achieve the above objectives, the present invention provides a control system for a window cleaning robot, including a multi-source sensor module, a weight adjustment module, a state estimation module, a motor drive module, a slip detection module, an adaptive recovery module, a path planning module, a motion decision module, an encrypted transmission module, an anomaly push module, and a client. The state estimation module is connected to the multi-source sensor module, the path planning module is connected to the state estimation module, the motion decision module is connected to the path planning module, the adaptive recovery module is connected to both the slip detection module and the path planning module, the weight adjustment module is connected to the state estimation module, the encrypted transmission module is connected to both the motion decision module and the motor drive module, and the anomaly push module is connected to both the motor drive module and the client. The multi-source sensor module is used to collect angular velocity data, acceleration data, motor current data and negative pressure data of the fuselage, and to perform zero-bias calibration and filtering on the collected sensor data. The weight adjustment module is used to adjust the fusion weight coefficients of each sensor data in the multi-source sensor module by adopting an adaptive weight allocation algorithm based on minimum mean square error. The state estimation module is used to receive the data processed by the multi-source sensor module and combine them into a real-time motion state vector. The slippage detection module is used to monitor in real time the deviation between the motor current and the expected current, the abnormal fluctuation of the fused tilt angle, and the difference between the drive wheel encoder pulse and the motion state estimate. It determines the slippage state and slippage level of the machine body through a multi-feature fusion slippage discrimination model. The adaptive recovery module is used to dynamically adjust the current output curve of the drive motor, replan the local path, and adjust the adsorption force parameters of the negative pressure system according to the slippage level output by the slippage detection module, so as to achieve autonomous recovery after slippage. The path planning module is used to generate a global cleaning path and a local obstacle avoidance path based on the real-time motion state vector output by the state estimation module and the preset cleaning map, using a dynamic window method multi-layer path planning strategy. The motion decision module is used to convert the path trajectory generated by the path planning module into a sequence of motor drive commands, including the target speed, direction and running time parameters of each drive motor. The encrypted transmission module is used to encrypt and transmit the motor drive commands generated by the motion decision module and the motion state data output by the state estimation module using a hybrid encryption scheme that combines the SM4 symmetric encryption algorithm and the SM9 identifier cryptography algorithm.

[0006] The control system of the window cleaning robot also includes a distributed storage module and a data mining module. The distributed storage module is connected to the motion decision module, and the data mining module is connected to the distributed storage module. The distributed storage module adopts the Hadoop Distributed File System or Ceph distributed storage architecture to distribute and store the multi-source sensor data, state estimation data, path planning data and motion decision data generated by the window cleaning robot during operation, and establishes multi-level indexes according to the data collection timestamp and data type. The data mining module employs a frequent pattern mining method based on the improved Apriori algorithm to perform association rule mining on the historical operating data stored in the distributed storage module. This identifies the slipping patterns, path deviation patterns, and energy consumption patterns of the window cleaning robot under different window materials, different ambient humidity levels, and different operating postures, and stores the mined association rules into the rule base.

[0007] The control system of the window cleaning robot also includes a model update module, which is connected to the data mining module. The model update module uses a federated learning algorithm to generate model update parameters by obtaining data from the rule base; the model update parameters are then sent to the motion decision module for model update.

[0008] The control system of the window cleaning robot also includes a data caching module, which is connected to the motion decision module. The data caching module is used to cache key status data, exception logs, and temporary path planning information during operation.

[0009] The control system of the window cleaning robot also includes an edge computing module, a fault self-healing module, and an energy management module. The edge computing module is connected to the multi-source sensor module, the fault self-healing module is connected to the motor drive module, and the energy management module is connected to the multi-source sensor module. The edge computing module uses a lightweight neural network model to perform real-time preprocessing and feature extraction of sensor data on the local end of the window cleaning robot, and only transmits key state data and abnormal events to the state estimation module. The fault self-healing module adopts a fault diagnosis algorithm based on a finite state machine. When a fault is detected in the motor drive module, a graded fault recovery strategy is automatically executed. The energy management module uses dynamic voltage and frequency adjustment technology to regulate power consumption according to the actual working conditions.

[0010] The state estimation module includes an inclination angle calculation unit, a heading angle calculation unit, an angular velocity processing unit, and a state vector construction unit. The tilt angle calculation unit is used to: receive the acceleration data processed by the multi-source sensor module, and calculate the fused tilt angle of the body on the surface to be cleaned based on the acceleration components of each axis; The heading angle calculation unit is used to: receive the gyroscope angular velocity data processed by the multi-source sensor module, perform zero-bias compensation on the angular velocity of the coordinate axis perpendicular to the surface to be cleaned, integrate it over time to obtain the real-time heading angle of the fuselage, and map the heading angle to a preset angle range. The angular velocity processing unit is used to: receive the gyroscope angular velocity data processed by the multi-source sensor module and generate the real-time angular velocity of the fuselage; The state vector construction unit is used to combine the tilt angle, real-time heading angle and real-time angular velocity into a real-time motion state vector.

[0011] The multi-source sensor module includes a three-axis gyroscope unit, a three-axis accelerometer unit, a motor current sampling unit, a barometric pressure sampling unit, a zero-bias calibration unit, and a filtering unit. The three-axis gyroscope unit is used to collect the three-axis angular velocity data of the fuselage; The triaxial accelerometer unit is used to collect triaxial acceleration data of the fuselage; The motor current sampling unit collects real-time current data of each drive motor through a current detection circuit composed of a sampling resistor and an operational amplifier. The air pressure sampling unit is used to collect real-time air pressure data of the negative pressure system; The zero-bias calibration unit is used to: execute the zero-bias calibration process when the window cleaning robot starts, continuously collect N triaxial acceleration data and triaxial angular velocity data, and perform mean filtering on the collected data; For coordinate axes parallel to the direction of gravity, the theoretical value of gravitational acceleration and the measured value after mean filtering are compensated to obtain the zero-bias calibration value for each axis; the zero-bias calibration value is stored in non-volatile memory and superimposed each time data is fused. The filtering unit is used to perform moving average filtering and weighted filtering on the sensor data to remove measurement noise.

[0012] This invention also provides a path control method, applied to the control system of the window cleaning robot as described above. include: The multi-source sensor module collects data on the fuselage's angular velocity, acceleration, motor current, and negative pressure, and performs zero-bias calibration and filtering. The state estimation module uses the processed acceleration and angular velocity data to calculate the fused tilt angle, real-time heading angle and real-time angular velocity, and combines them into a real-time motion state vector. The path planning module generates a global cleaning path and a local obstacle avoidance path based on real-time motion state vectors and a preset cleaning map, using a dynamic window method. The motion decision module converts the generated path trajectory into a sequence of motor drive commands, including the target speed, direction, and running time parameters of the drive motor. The slippage detection module monitors the motor current deviation, tilt angle fluctuation, and the difference between encoder pulses and state estimates in real time to determine the slippage state and degree level. If slippage is detected, the adaptive recovery module dynamically adjusts the motor current output curve, replans the local path, and adjusts the negative pressure adsorption force parameters according to the degree of slippage, so as to achieve autonomous recovery. The encrypted transmission module adopts a hybrid encryption scheme of SM4 and SM9 to encrypt and transmit motor drive commands and motion status data to the motor drive module. The motor drive module executes commands to drive the machine body to move, and at the same time, the anomaly push module pushes alarm information to the client when it detects a drive anomaly.

[0013] This invention discloses a control system and path control method for a window cleaning robot. The weight adjustment module employs an adaptive weight allocation algorithm based on minimum mean square error, which adjusts the fusion weights of multi-source sensor data in real time according to changes in the friction characteristics of the working surface. This significantly improves the adaptability and robustness of the state estimation module on different materials such as glass and ceramic tiles. The slip detection module combines multiple features such as motor current deviation, tilt angle fluctuation, and encoder pulse difference for accurate determination of slip state and degree, avoiding misjudgments or missed detections caused by traditional fixed thresholds. The adaptive recovery module dynamically adjusts the drive motor current curve, local path, and negative pressure adsorption force according to the degree of slippage, enabling the robot to autonomously overcome slippage without stopping or repeating ineffective actions, effectively preventing interruptions or jamming during cleaning. Furthermore, the encrypted transmission module uses a hybrid SM4 and SM9 encryption scheme to ensure secure transmission of commands and state data, and the anomaly push module promptly reports faults to the client, further enhancing the system's security and maintainability. This invention achieves stable movement, precise slippage recovery, and efficient cleaning of the window cleaning robot in complex surface environments. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.

[0015] Figure 1 This is a principle block diagram of the first embodiment of the present invention.

[0016] Figure 2 This is a principle block diagram of the second embodiment of the present invention.

[0017] In the diagram: 1-Multi-source sensor module, 2-Weight adjustment module, 3-State estimation module, 4-Motor drive module, 5-Slip detection module, 6-Adaptive recovery module, 7-Path planning module, 8-Motion decision module, 9-Encrypted transmission module, 10-Abnormal push module, 11-Client, 12-Distributed storage module, 13-Data mining module, 14-Model update module, 15-Data caching module, 16-Edge computing module, 17-Fault self-healing module, 18-Energy management module, 19-Tilt angle calculation unit, 20-Heading angle calculation unit, 21-Angular velocity processing unit, 22-State vector construction unit, 23-Three-axis gyroscope unit, 24-Three-axis accelerometer unit, 25-Motor current sampling unit, 26-Air pressure sampling unit, 27-Zero bias calibration unit, 28-Filtering unit, 29-Login module, 30-Access control module; Detailed Implementation

[0018] The embodiments of the present invention are described in detail below. Examples of the embodiments are shown in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, but should not be construed as limiting the present invention.

[0019] First embodiment: Please refer to Figure 1 , Figure 1 This is a principle block diagram of the first embodiment of the present invention. The embodiment provides a control system for a window cleaning robot, including a multi-source sensor module 1, a weight adjustment module 2, a state estimation module 3, a motor drive module 4, a slip detection module 5, an adaptive recovery module 6, a path planning module 7, a motion decision module 8, an encrypted transmission module 9, an anomaly push module 10, and a client 11. The control system also includes a distributed storage module 12, a data mining module 13, a model update module 14, a data caching module 15, an edge computing module 16, a fault self-healing module 17, and an energy consumption management system. The state estimation module 3 includes a tilt angle calculation unit 19, a heading angle calculation unit 20, an angular velocity processing unit 21, and a state vector construction unit 22; the multi-source sensor module 1 includes a three-axis gyroscope unit 23, a three-axis accelerometer unit 24, a motor current sampling unit 25, an air pressure sampling unit 26, a zero-bias calibration unit 27, and a filtering unit 28. The aforementioned solution solves the technical problem that in the prior art, the sensor data fusion weights are usually configured in a fixed way, which makes it difficult to adapt to the changes in the friction characteristics of different working surfaces, thus leading to untimely or misjudged slippage detection. Furthermore, it lacks an adaptive recovery mechanism based on the degree of slippage, often only able to stop the machine and alarm or repeatedly execute preset actions, resulting in easy entrapment or cleaning interruption.

[0020] In this specific embodiment, the multi-source sensor module 1 is used to collect angular velocity data, acceleration data, motor current data and negative pressure data of the fuselage, and to perform zero-bias calibration and filtering on the collected sensor data; The weight adjustment module 2 is used to adjust the fusion weight coefficients of each sensor data in the multi-source sensor module 1 by adopting an adaptive weight allocation algorithm based on minimum mean square error. The state estimation module 3 is used to receive the data processed by the multi-source sensor module 1 and combine it into a real-time motion state vector; The slippage detection module 5 is used to monitor in real time the deviation between the motor current and the expected current, the abnormal fluctuation of the fused tilt angle, and the difference between the drive wheel encoder pulse and the motion state estimate. It determines the slippage state and slippage level of the machine body through a slippage discrimination model with multi-feature fusion. The adaptive recovery module 6 is used to dynamically adjust the current output curve of the drive motor, replan the local path, and adjust the adsorption force parameters of the negative pressure system according to the slippage level output by the slippage detection module 5, so as to achieve autonomous recovery after slippage. The path planning module 7 is used to generate a global cleaning path and a local obstacle avoidance path based on the real-time motion state vector output by the state estimation module 3 and the preset cleaning map, using a dynamic window method multi-layer path planning strategy. The motion decision module 8 is used to convert the path trajectory generated by the path planning module 7 into a sequence of motor drive instructions, including the target speed, direction and running time parameters of each drive motor. The encrypted transmission module 9 is used to encrypt and transmit the motor drive commands generated by the motion decision module 8 and the motion state data output by the state estimation module 3 using a hybrid encryption scheme that combines the SM4 symmetric encryption algorithm and the SM9 identifier cryptography algorithm.

[0021] Specifically, the state estimation module 3 is connected to the multi-source sensor module 1, the path planning module 7 is connected to the state estimation module 3, the motion decision module 8 is connected to the path planning module 7, the adaptive recovery module 6 is connected to both the slip detection module 5 and the path planning module 7, the weight adjustment module 2 is connected to the state estimation module 3, the encrypted transmission module 9 is connected to both the motion decision module 8 and the motor drive module 4, and the anomaly push module 10 is connected to both the motor drive module 4 and the client 11. Secondly, the distributed storage module 12 adopts the Hadoop Distributed File System or Ceph distributed storage architecture to distribute the multi-source sensor data, state estimation data, path planning data and motion decision data generated by the window cleaning robot during operation, and establishes multi-level indexes according to the data collection timestamp and data type. The data mining module 13 uses a frequent pattern mining method based on the improved Apriori algorithm to perform association rule mining on the historical operation data stored in the distributed storage module 12, identify the slipping pattern, path deviation pattern and energy consumption pattern of the window cleaning robot under different window materials, different ambient humidity and different operating postures, and store the mined association rules into the rule base.

[0022] The model update module 14 is connected to the data mining module 13; The model update module 14 uses a federated learning algorithm to generate model update parameters by obtaining data from the rule base; the model update parameters are then sent to the motion decision module 8 for model update.

[0023] Meanwhile, the data caching module 15 is connected to the motion decision module 8; The data caching module 15 is used to cache key status data, exception logs, and temporary path planning information during operation, and automatically deletes the data after it is used.

[0024] Furthermore, the edge computing module 16 is connected to the multi-source sensor module 1, the fault self-healing module 17 is connected to the motor drive module 4, and the energy management module 18 is connected to the multi-source sensor module 1. The edge computing module 16 uses a lightweight neural network model to perform real-time preprocessing and feature extraction of sensor data on the local end of the window cleaning robot, and only transmits key state data and abnormal events to the state estimation module 3. The fault self-healing module 17 adopts a fault diagnosis algorithm based on a finite state machine. When a fault is detected in the motor drive module 4, a graded fault recovery strategy is automatically executed. The energy management module 18 uses dynamic voltage and frequency adjustment technology to adjust power consumption according to the actual working state.

[0025] Furthermore, the tilt angle calculation unit 19 is used to: receive the acceleration data processed by the multi-source sensor module 1, and calculate the fused tilt angle of the body on the surface to be cleaned based on the acceleration components of each axis; The heading angle calculation unit 20 is used to: receive the gyroscope angular velocity data processed by the multi-source sensor module 1, perform zero-bias compensation on the angular velocity of the coordinate axis perpendicular to the surface to be cleaned, integrate it over time to obtain the real-time heading angle of the fuselage, and map the heading angle to a preset angle range. The angular velocity processing unit 21 is used to: receive the gyroscope angular velocity data processed by the multi-source sensor module 1, and generate the real-time angular velocity of the fuselage; The state vector construction unit 22 is used to combine the fused tilt angle, real-time heading angle and real-time angular velocity into a real-time motion state vector.

[0026] The three-axis gyroscope unit 23 is used to collect the three-axis angular velocity data of the fuselage; The triaxial accelerometer unit 24 is used to collect triaxial acceleration data of the fuselage; The motor current sampling unit 25 collects real-time current data of each drive motor through a current detection circuit composed of a sampling resistor and an operational amplifier. The air pressure sampling unit 26 is used to collect real-time air pressure data of the negative pressure system; The zero-bias calibration unit 27 is used to: execute the zero-bias calibration process when the window cleaning robot starts, continuously collect N times of triaxial acceleration data and triaxial angular velocity data, and perform mean filtering on the collected data; For coordinate axes parallel to the direction of gravity, the theoretical value of gravitational acceleration and the measured value after mean filtering are compensated to obtain the zero-bias calibration value for each axis; the zero-bias calibration value is stored in non-volatile memory and superimposed each time data is fused. The filtering unit 28 is used to perform moving average filtering and weighted filtering on the sensor data to remove measurement noise.

[0027] The calculation method of the fused tilt angle is as follows: using accelerometer data to calculate the tilt angle of the body relative to the direction of gravity through trigonometric functions, and at the same time using gyroscope angular velocity data to calculate the body rotation angle through integration. The tilt angle calculated by the accelerometer and the rotation angle obtained by the gyroscope integration are weighted and fused according to a preset weight ratio to obtain the filtered stable tilt angle.

[0028] Second embodiment: Please refer to Figure 2 , Figure 2 This is a schematic diagram of the second embodiment of the present invention. The control system of the window cleaning robot provided in this embodiment of the present invention also includes a login module 29 and an access control module 30.

[0029] The login module 29 is used to log in to the client 11, and the permission control module 30 adopts the role-based access control (RBAC) algorithm to control permissions based on the role of the person logging in to the client 11. Working Principle: The weight adjustment module 2 employs an adaptive weight allocation algorithm based on minimum mean square error, which can adjust the fusion weights of multi-source sensor data in real time according to the changes in the friction characteristics of the working surface, significantly improving the adaptability and robustness of the state estimation module 3 on different materials such as glass or ceramic tiles. The slip detection module 5 combines multiple features such as motor current deviation, tilt angle fluctuation, and encoder pulse difference for accurate determination of slip state and degree, avoiding misjudgment or missed detection caused by traditional fixed thresholds. The adaptive recovery module 6 dynamically adjusts the drive motor current curve, local path, and negative pressure adsorption force according to the degree of slip, enabling the robot to autonomously overcome slip difficulties without stopping or repeating ineffective actions, effectively preventing interruptions or jamming during cleaning. In addition, the encrypted transmission module 9 uses a hybrid encryption scheme of SM4 and SM9 to ensure secure transmission of commands and status data, and the anomaly push module 10 promptly reports faults to the client 11, further improving the system's security and maintainability. This invention achieves stable walking, precise slip recovery, and efficient cleaning of the window cleaning robot in complex surface environments.

[0030] This invention also provides a path control method, applied to the control system of the window cleaning robot as described above. include: The multi-source sensor module 1 collects data on the fuselage's angular velocity, acceleration, motor current, and negative pressure, and performs zero-bias calibration and filtering. The state estimation module 3 uses the processed acceleration and angular velocity data to calculate the fused tilt angle, real-time heading angle and real-time angular velocity, and combines them into a real-time motion state vector. The path planning module 7 generates a global cleaning path and a local obstacle avoidance path based on the real-time motion state vector and the preset cleaning map, using the dynamic window method. The motion decision module 8 converts the generated path trajectory into a sequence of motor drive commands, including the target speed, direction and running time parameters of the drive motor. The slippage detection module 5 monitors the motor current deviation, tilt angle fluctuation, and the difference between encoder pulses and state estimates in real time to determine the slippage state and degree level. If slippage is detected, the adaptive recovery module 6 dynamically adjusts the motor current output curve, replans the local path, and adjusts the negative pressure adsorption force parameters according to the degree of slippage to achieve autonomous recovery. The encrypted transmission module 9 adopts a hybrid encryption scheme of SM4 and SM9 to encrypt and transmit motor drive commands and motion status data to the motor drive module 4; The motor drive module 4 executes commands to drive the body to move, and at the same time, the abnormality push module 10 pushes alarm information to the client 11 when it detects a drive abnormality.

[0031] The above-disclosed embodiments are merely one or more preferred embodiments of this application and should not be construed as limiting the scope of this application. Those skilled in the art can understand that all or part of the processes for implementing the above embodiments and equivalent changes made in accordance with the claims of this application still fall within the scope of this application.

Claims

1. A control system for a window cleaning robot, characterized in that, It includes a multi-source sensor module, a weight adjustment module, a state estimation module, a motor drive module, a slippage detection module, an adaptive recovery module, a path planning module, a motion decision module, an encrypted transmission module, an anomaly push module, and a client. The state estimation module is connected to the multi-source sensor module, the path planning module is connected to the state estimation module, the motion decision module is connected to the path planning module, the adaptive recovery module is connected to both the slip detection module and the path planning module, the weight adjustment module is connected to the state estimation module, the encrypted transmission module is connected to both the motion decision module and the motor drive module, and the anomaly push module is connected to both the motor drive module and the client. The multi-source sensor module is used to collect angular velocity data, acceleration data, motor current data and negative pressure data of the fuselage, and to perform zero-bias calibration and filtering on the collected sensor data. The weight adjustment module is used to adjust the fusion weight coefficients of each sensor data in the multi-source sensor module by adopting an adaptive weight allocation algorithm based on minimum mean square error. The state estimation module is used to receive the data processed by the multi-source sensor module and combine them into a real-time motion state vector. The slippage detection module is used to monitor in real time the deviation between the motor current and the expected current, the abnormal fluctuation of the fused tilt angle, and the difference between the drive wheel encoder pulse and the motion state estimate. It determines the slippage state and slippage level of the machine body through a multi-feature fusion slippage discrimination model. The adaptive recovery module is used to dynamically adjust the current output curve of the drive motor, replan the local path, and adjust the adsorption force parameters of the negative pressure system according to the slippage level output by the slippage detection module, so as to achieve autonomous recovery after slippage. The path planning module is used to generate a global cleaning path and a local obstacle avoidance path based on the real-time motion state vector output by the state estimation module and the preset cleaning map, using a dynamic window method multi-layer path planning strategy. The motion decision module is used to convert the path trajectory generated by the path planning module into a sequence of motor drive commands, including the target speed, direction and running time parameters of each drive motor. The encrypted transmission module is used to encrypt and transmit the motor drive commands generated by the motion decision module and the motion state data output by the state estimation module using a hybrid encryption scheme that combines the SM4 symmetric encryption algorithm and the SM9 identifier cryptography algorithm.

2. The control system of the window cleaning robot as described in claim 1, characterized in that, The control system of the window cleaning robot also includes a distributed storage module and a data mining module. The distributed storage module is connected to the motion decision module, and the data mining module is connected to the distributed storage module. The distributed storage module adopts the Hadoop Distributed File System or Ceph distributed storage architecture to distribute and store the multi-source sensor data, state estimation data, path planning data and motion decision data generated by the window cleaning robot during operation, and establishes multi-level indexes according to the data collection timestamp and data type. The data mining module employs a frequent pattern mining method based on the improved Apriori algorithm to perform association rule mining on the historical operating data stored in the distributed storage module. This identifies the slipping patterns, path deviation patterns, and energy consumption patterns of the window cleaning robot under different window materials, different ambient humidity levels, and different operating postures, and stores the mined association rules into the rule base.

3. The control system of the window cleaning robot as described in claim 2, characterized in that, The control system of the window cleaning robot also includes a model update module, which is connected to the data mining module; The model update module uses a federated learning algorithm to generate model update parameters by obtaining data from the rule base; the model update parameters are then sent to the motion decision module for model update.

4. The control system of the window cleaning robot as described in claim 3, characterized in that, The control system of the window cleaning robot also includes a data caching module, which is connected to the motion decision module; The data caching module is used to cache key status data, exception logs, and temporary path planning information during operation.

5. The control system of the window cleaning robot as described in claim 4, characterized in that, The control system of the window cleaning robot also includes an edge computing module, a fault self-healing module, and an energy management module. The edge computing module is connected to the multi-source sensor module, the fault self-healing module is connected to the motor drive module, and the energy management module is connected to the multi-source sensor module. The edge computing module uses a lightweight neural network model to perform real-time preprocessing and feature extraction of sensor data on the local end of the window cleaning robot, and only transmits key state data and abnormal events to the state estimation module. The fault self-healing module adopts a fault diagnosis algorithm based on a finite state machine. When a fault is detected in the motor drive module, a graded fault recovery strategy is automatically executed. The energy management module uses dynamic voltage and frequency adjustment technology to regulate power consumption according to the actual working conditions.

6. The control system of the window cleaning robot as described in claim 5, characterized in that, The state estimation module includes an inclination angle calculation unit, a heading angle calculation unit, an angular velocity processing unit, and a state vector construction unit; The tilt angle calculation unit is used to: receive the acceleration data processed by the multi-source sensor module, and calculate the fused tilt angle of the body on the surface to be cleaned based on the acceleration components of each axis; The heading angle calculation unit is used to: receive the gyroscope angular velocity data processed by the multi-source sensor module, perform zero-bias compensation on the angular velocity of the coordinate axis perpendicular to the surface to be cleaned, integrate it over time to obtain the real-time heading angle of the fuselage, and map the heading angle to a preset angle range. The angular velocity processing unit is used to: receive the gyroscope angular velocity data processed by the multi-source sensor module and generate the real-time angular velocity of the fuselage; The state vector construction unit is used to combine the tilt angle, real-time heading angle and real-time angular velocity into a real-time motion state vector.

7. The control system of the window cleaning robot as described in claim 6, characterized in that, The multi-source sensor module includes a three-axis gyroscope unit, a three-axis accelerometer unit, a motor current sampling unit, a barometric pressure sampling unit, a zero-bias calibration unit, and a filtering unit; The three-axis gyroscope unit is used to collect the three-axis angular velocity data of the fuselage; The triaxial accelerometer unit is used to collect triaxial acceleration data of the fuselage; The motor current sampling unit collects real-time current data of each drive motor through a current detection circuit composed of a sampling resistor and an operational amplifier. The air pressure sampling unit is used to collect real-time air pressure data of the negative pressure system; The zero-bias calibration unit is used to: execute the zero-bias calibration process when the window cleaning robot starts, continuously collect N triaxial acceleration data and triaxial angular velocity data, and perform mean filtering on the collected data; For coordinate axes parallel to the direction of gravity, the theoretical value of gravitational acceleration is compensated by the measured value after mean filtering to obtain the zero-bias calibration value of each axis; The zero-bias calibration value is stored in non-volatile memory and superimposed during each data fusion. The filtering unit is used to perform moving average filtering and weighted filtering on the sensor data to remove measurement noise.

8. A path control method, applied to the control system of the window cleaning robot as described in claim 7, characterized in that, include: The multi-source sensor module collects data on the fuselage's angular velocity, acceleration, motor current, and negative pressure, and performs zero-bias calibration and filtering. The state estimation module uses the processed acceleration and angular velocity data to calculate the fused tilt angle, real-time heading angle and real-time angular velocity, and combines them into a real-time motion state vector. The path planning module generates a global cleaning path and a local obstacle avoidance path based on real-time motion state vectors and a preset cleaning map, using a dynamic window method. The motion decision module converts the generated path trajectory into a sequence of motor drive commands, including the target speed, direction, and running time parameters of the drive motor. The slippage detection module monitors the motor current deviation, tilt angle fluctuation, and the difference between encoder pulses and state estimates in real time to determine the slippage state and degree level. If slippage is detected, the adaptive recovery module dynamically adjusts the motor current output curve, replans the local path, and adjusts the negative pressure adsorption force parameters according to the degree of slippage, so as to achieve autonomous recovery. The encrypted transmission module adopts a hybrid encryption scheme of SM4 and SM9 to encrypt and transmit motor drive commands and motion status data to the motor drive module. The motor drive module executes commands to drive the machine body to move, and at the same time, the anomaly push module pushes alarm information to the client when it detects a drive anomaly.

Images