A portable photovoltaic cleaning robot

Abstract

The present application relates to photovoltaic operation and maintenance equipment technical field, the present application discloses a kind of portable photovoltaic cleaning robot, including chassis module, cleaning module, centroid adjustment module, perception module and main control module, the centroid adjustment module is driven battery pack along robot longitudinal axis reciprocating movement by linear driver, to dynamically adjust the centroid distribution of whole machine, main control module is based on the attitude angle, motor current and vibration data collected by perception module, identify the slope state and stain working condition type of current photovoltaic module, and calculate the target position of battery pack accordingly, system is switched between cleaning enhancement mode and traction enhancement mode, respectively realize increasing brush down pressure or improve driving wheel adhesion force.The present application utilizes battery pack multiplexing as movable counterweight, effectively solves the dynamics control conflict between high viscosity stain cleaning and large slope antiskid operation under the premise of not increasing the load of whole machine, improves the environmental adaptability of robot.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic operation and maintenance equipment technology, specifically a portable photovoltaic cleaning robot. Background Technology

[0002] With the large-scale development of the photovoltaic power generation industry, dust and dirt accumulation on the surface of photovoltaic modules have become key factors affecting power generation efficiency. To reduce manual maintenance costs and increase cleaning frequency, portable photovoltaic cleaning robots are gradually replacing traditional manual cleaning methods and becoming the mainstream equipment for the operation and maintenance of distributed photovoltaic power plants. These robots typically need to move across the surface of modules at different tilt angles and are capable of removing various types of contaminants, such as dry dust, bird droppings, and sticky adhesives.

[0003] Existing portable photovoltaic cleaning robots primarily employ a tracked or wheeled chassis structure, with a rotating brush mounted at the front for physical cleaning. In terms of structural layout, to ensure the overall compactness and portability, the robot's main weight components, such as the battery pack, drive motor, and control box, are typically fixed in specific locations on the frame. Generally, designers place the robot's center of gravity near its geometric center, or slightly offset towards the cleaning module, to ensure the brush maintains contact with the photovoltaic panels, utilizing the robot's own weight to generate cleaning friction.

[0004] While existing technologies have automated photovoltaic cleaning, several shortcomings remain. Due to their fixed center-of-gravity distribution, robots face significant dynamic challenges when dealing with complex and variable working conditions. Specifically, when facing highly viscous and stubborn stains, the roller brush needs to apply significant normal pressure to break down the stain's adhesion. However, the fixed center of gravity limits the maximum pressure on the brush end, causing the bristles to elastically deform and slip across the stain surface, resulting in incomplete cleaning. Conversely, if the center of gravity is pre-positioned too far forward during the design phase to enhance cleaning power, when the robot travels to components with steep inclines, the downward component of gravity along the slope causes a sharp decrease in the normal load on the rear wheels. This results in the maximum static friction between the drive wheels and the panel being insufficient to overcome the downward component of gravity, leading to slippage, yaw, or even falls. Furthermore, existing control logic often struggles to distinguish between changes in gravity load caused by the slope and changes in friction load caused by the stain, leading to misjudgments of the working conditions by the robot. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a portable photovoltaic cleaning robot. This solves the problem that existing portable photovoltaic cleaning robots, which use a fixed center of mass structure, cannot simultaneously handle the large normal pressure required for cleaning highly viscous stains and the strong driving grip required for walking on steep slopes, resulting in incomplete cleaning or slippage and instability under complex working conditions.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a portable photovoltaic cleaning robot, comprising:

[0007] Chassis module, used to provide driving force along the surface of photovoltaic modules; A cleaning module, located at the front end of the chassis module, includes a rotating roller brush; The center of gravity adjustment module includes a counterweight unit that can move along the longitudinal axis of the robot; A sensing module is used to collect attitude, current and vibration data; and a main control module is connected to the above modules. The main control module is configured to identify the slope and stain conditions based on the data from the sensing module, calculate and drive the counterweight unit to move to the target position, and adjust the normal pressure of the rotating brush and the driving adhesion of the chassis module by changing the overall center of gravity distribution, so as to reconstruct the state between cleaning enhancement, traction enhancement and cruise balance modes.

[0008] Preferably, the centroid adjustment module includes: The longitudinal guide rail is fixedly installed on the top surface of the chassis module frame along the front and rear axis of the robot; A sliding seat is adapted to be installed on the longitudinal guide rail and is used to support the counterweight unit; A linear actuator connects to and drives the sliding block to move linearly along the longitudinal guide rail; The counterweight unit is the robot's power battery pack, which is reused as a dynamically adjustable movable mass block; the linear actuator integrates a position encoder to provide feedback to the main control module on the real-time displacement of the battery pack relative to the robot's geometric center.

[0009] Preferably, the sensing module includes: An attitude sensor, installed at the geometric center of the robot body, is used to collect the robot's pitch and roll angles. A current sensor, integrated into the drive circuit of the cleaning module, is used to collect the original phase current of the rotating brush motor. A vibration sensor is installed on the surface of the mechanical housing of the cleaning module to collect contact vibration signals between the roller brush and the surface of the photovoltaic module. The main control module is configured to calculate the influence of gravity component on current based on the data from the attitude sensor, and to perform frequency domain analysis based on the data from the vibration sensor to extract net vibration energy within the characteristic frequency band.

[0010] Preferably, the main control module has built-in signal decoupling logic, configured to perform the following operations: Based on the real-time angle fed back by the attitude sensor, the change in reference friction current caused by the change in gravity is calculated using a dynamic model. The pure friction current characterizing the viscous resistance of the stain is obtained by subtracting the change in the reference friction current and the inherent mechanical loss current of the motor from the original phase current collected by the current sensor. The main control module uses only the pure frictional current and the net vibration energy as input variables to evaluate the working condition of the stain, thereby eliminating the interference of the robot's own posture changes and mechanical background noise on the identification of the working condition.

[0011] Preferably, the main control module is configured to construct a stain resistance index vector to identify the stain condition type: The stain resistance index vector includes a viscous resistance component that reflects low-frequency mechanical properties and a roughness resistance component that reflects high-frequency contact properties. When the viscous impedance component exceeds a preset current characteristic threshold and the roughness impedance component is lower than a preset energy characteristic threshold, it is determined to be a high-viscous stain condition. When the roughness impedance component exceeds the energy characteristic threshold and the viscous impedance component is lower than the current characteristic threshold, it is determined to be a loose dust accumulation condition. When both components exceed their respective thresholds, it is determined to be a combined pollution condition.

[0012] Preferably, the main control module has a built-in safety priority arbitration mechanism to resolve operating condition conflicts: When the attitude angle data indicates that the current slope exceeds the preset safety limit, regardless of the type of soiling condition, the system prioritizes the traction enhancement mode. The main control module sets the target position to move towards the side of the robot's drive wheel to increase the drive wheel's grip and prevent slippage. The system determines the cleaning enhancement mode only when the current slope does not exceed the safety limit and a high-viscosity stain or complex pollution condition is identified. The main control module sets the target position to move towards the robot cleaning end to increase the downward pressure of the rotating brush by using the lever principle.

[0013] Preferably, the main control module incorporates anti-overturning boundary constraint logic when calculating the target position: The main control module uses a torque balance model to predict the normal support reaction forces at the front and rear support points of the robot based on the current slope data and the proposed target position. If any calculated normal support reaction force is less than the preset safety threshold, it indicates a risk of longitudinal overturning. The main control module performs amplitude limiting correction on the target position, forcibly restricting the counterweight unit to the stroke range that meets the anti-overturning safety constraints.

[0014] Preferably, the main control module is further configured with kinematic interlock logic for controlling the timing of the center of mass adjustment module's actions. When the deviation between the calculated target position and the current position exceeds a safety threshold, the system enters a reconstruction state. In the reconfiguration state, the main control module first controls the chassis module to decelerate to a preset creeping speed or stop running; After confirming that the chassis module speed is below the safety limit, the center of gravity adjustment module is driven to move, and the movement trajectory of the counterweight unit is planned using an S-shaped speed curve to limit the maximum acceleration and prevent the fuselage from shaking due to rapid movement of the center of gravity. The main control module will only release the interlock and restore the normal operating speed of the chassis module after the counterweight unit reaches the target position.

[0015] Preferably, the main control module is further configured with state maintenance and hysteresis exit logic: The main control module has a preset spatial position dead zone threshold. Motion control commands are only generated when the absolute value of the deviation between the target position and the current position is greater than the dead zone threshold. The main control module has a built-in state holding timer. When the detected stain condition or slope condition falls back to the normal range, the counterweight unit is not immediately reset, but a delay countdown is triggered. If a high-load condition is detected again during the countdown, the timer is reset and the current mode is maintained; the counterweight unit can only be controlled to return to the geometric center after the countdown has ended.

[0016] Preferably, the main control module is further configured to execute a power-on self-test calibration strategy: In the initial stage of each robot operation, the rotating brush is controlled to rotate at a low speed and be in a state of minimum pressure. The average value of the motor current at this time is collected as the reference value of the inherent mechanical loss current for this operation. This value is used to correct the zero-point drift caused by mechanical wear or temperature changes, and the corrected reference value is used in subsequent signal decoupling operations.

[0017] This invention provides a portable photovoltaic cleaning robot. It has the following advantages: 1. This invention reuses the power supply battery pack as a movable counterweight unit, achieving active adjustment of the robot's center of gravity position without increasing the overall load. Utilizing the lever principle, the center of gravity shifts forward when highly viscous stains are detected to increase the downward pressure of the roller brush and enhance cleaning ability; when steep terrain is detected, the center of gravity shifts backward to increase the traction of the drive wheels and prevent slippage. This solution effectively resolves the technical contradiction that a single fixed center of gravity structure cannot simultaneously balance cleaning pressure and climbing adhesion.

[0018] 2. This invention constructs a working condition recognition system based on multimodal information fusion. It employs a signal processing algorithm incorporating gravity compensation and spectral subtraction denoising to eliminate interference from vehicle posture changes and mechanical background noise on sensor data. This scheme can decouple and quantify the viscous drag characteristics and surface roughness characteristics of stains, enabling the robot to accurately distinguish between different working conditions such as dry dust accumulation, highly viscous stains, and mixed adhesives, providing reliable decision input for dynamic control.

[0019] 3. This invention introduces a safety priority arbitration mechanism and boundary constraint logic based on torque balance, establishing automatic switching rules between cleaning enhancement mode and traction enhancement mode. The system can prioritize anti-slip requirements while ensuring anti-tipping safety, and perform real-time amplitude limiting on the target position of the counterweight unit. Combined with kinematic interlocking and state-maintaining strategies, it avoids instability caused by rapid changes in the center of gravity or frequent reciprocating motion, ensuring the safety and continuity of the robot's operation in complex terrain. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the overall structural modules of the present invention; Figure 2 This is a perspective view of the present invention; Figure 3 This is a cross-sectional view of the present invention; Figure 4 This is an exploded view of the centroid adjustment module structure of the present invention; Figure 5 This is a hardware connection block diagram of the electrical control system of the present invention; Figure 6 This is a flowchart of the control logic of the present invention; Figure 7 This is a schematic diagram illustrating the signal decoupling principle of the present invention; Figure 8 This is a state transition diagram for mode switching in this invention.

[0021] Among them, 110 is the chassis module; 111 is the walking motor; 120 is the cleaning module; 121 is the roller brush motor; 122 is the rotating roller brush; 130 is the center of gravity adjustment module; 131 is the battery pack; 132 is the linear actuator; 133 is the longitudinal guide rail; 140 is the sensing module; 141 is the attitude sensor; 142 is the current sensor; 143 is the vibration sensor; and 150 is the main control module. Detailed Implementation

[0022] The technical solutions in 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 embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] See attached document Figure 1 Appendix Figure 2 and attached Figure 3 This is a schematic diagram of the overall structure of an adaptive center-of-gravity adjusting photovoltaic cleaning robot according to an embodiment of the present invention. The photovoltaic cleaning robot mainly includes a chassis module 110, a cleaning module 120, a center-of-gravity adjusting module 130, a sensing module 140, and a main control module 150. These modules are integrated into a mechatronic system with autonomous operation capabilities through mechanical or electrical connections, used to perform dust removal and decontamination operations on the surface of photovoltaic modules.

[0024] The chassis module 110 constitutes the robot's basic support platform and motion execution unit. The chassis module 110 includes a main frame, wheelsets mounted on both sides of the main frame, and a motor 111 driving the wheelsets. The main frame is made of lightweight, high-strength aluminum alloy or carbon fiber composite material to reduce the overall weight of the robot. The wheelsets are made of high-friction rubber, with a surface dynamic friction coefficient designed to be greater than 0.6 to ensure sufficient adhesion on the inclined photovoltaic panel surface. The motor 111 is connected to the wheelsets via a reduction gearbox and is used to drive the robot to move along the long or short side of the photovoltaic module array, with a designed climbing ability covering an inclination angle range of 0 to 30 degrees. The specific mechanical assembly method of the chassis module 110, including the selection of bearings and the arrangement of fasteners, can be achieved by those skilled in the art using conventional mechanical design methods and will not be elaborated here.

[0025] The cleaning module 120 is mounted at the front end of the chassis module 110, defining the front of the robot's travel direction as the cleaning end. The cleaning module 120 includes a rotating roller brush 122, a roller brush motor 121, and a cantilever bracket for supporting the roller brush. The rotating roller brush 122 is hinged or fixed to the front side of the chassis module 110 via the cantilever bracket, and its axis is perpendicular to the robot's direction of movement. The roller brush motor 121 drives the rotating roller brush 122 to rotate at high speed via belt drive or direct drive, using the brush bristles to physically brush the surface of the photovoltaic module. In this embodiment, the bristles of the rotating roller brush 122 are made of nylon or modified PBT material, possessing a certain degree of flexibility and wear resistance. The bristle length is set to 40 mm to 80 mm to accommodate the height difference of the photovoltaic module frame.

[0026] The center-of-gravity adjustment module 130 is located above the chassis module 110 and is the core actuator for implementing the adaptive control logic of this invention. The center-of-gravity adjustment module 130 includes a longitudinal guide rail 133, a linear actuator 132, and a battery pack 131 as a movable mass block. The longitudinal guide rail 133 is fixedly mounted on the top surface of the main frame along the robot's front-rear axis, and its length covers the main area from one side of the cleaning module 120 to the side of the rear wheel of the chassis module 110. The battery pack 131 is mounted on a sliding seat adapted to the longitudinal guide rail 133 and can reciprocate linearly along the guide rail. To ensure the effectiveness of the adjustment, the mass of the battery pack 131 is designed to account for 20% to 40% of the total mass of the machine. The linear actuator 132 is connected to and drives the sliding seat carrying the battery pack 131. In this embodiment, the linear driver 132 is specifically implemented as a combination of a ball screw module and a stepper motor. The stepper motor rotates to drive the screw, converting the rotational motion into the linear displacement of the sliding seat. In another embodiment, the linear driver 132 may also adopt a synchronous belt drive mechanism or a gear and rack mechanism.

[0027] Regarding the technical features of center of gravity adjustment, in specific mechanical implementation, it is defined as changing the relative position of the battery pack 131, the main counterweight component, on the longitudinal axis of the robot body, thereby altering the distribution ratio of the overall center of gravity between the front and rear wheels and the contact point of the roller brush. The battery pack 131 not only serves as the robot's energy supply unit but is also reused as a balancing counterweight for dynamic adjustment; this design avoids adding extra load. The linear actuator 132 integrates a position encoder for real-time feedback of the displacement of the battery pack 131 relative to the geometric center of the chassis.

[0028] The sensing modules 140 are distributed at key nodes of the robot to collect environmental information and operational status data. The sensing modules 140 include an attitude sensor 141 (IMU) mounted at the geometric center of the main frame to detect the robot's pitch angle, roll angle, and three-axis acceleration; a current sensor 142 integrated into the drive circuits of the walking motor 111 and the brush motor 121 to monitor the operating current in real time; and a vibration sensor 143 mounted on the surface of the gearbox housing of the cleaning module 120 to collect contact vibration signals. All sensors are connected to the main control module 150 via shielded cables.

[0029] The main control module 150 includes a core control circuit board, a computing unit, and peripheral interface circuits, encapsulated in a dustproof and waterproof electrical box. This electrical box is typically fixed in a non-moving area of ​​the chassis module 110 or directly integrated into the housing of the battery pack 131 (if integrated into the housing of the battery pack 131, flexible cable connections must be considered). The main control module 150 is electrically connected to the walking motor 111, the roller brush motor 121, the linear actuator 132, and the sensing module 140, and is responsible for executing the signal processing, working condition classification, position calculation, and safety control logic described above and in subsequent chapters. Based on the received multi-source sensor data, the main control module 150 dynamically adjusts the action of the linear actuator 132, thereby changing the position of the battery pack 131 and achieving a physical reconstruction of cleaning pressure and traction.

[0030] See attached document Figure 4 , Figure 4 This is a schematic diagram of a signal decoupling and dynamic noise reduction process according to an embodiment of the present invention.

[0031] During operation, the main control module 150 acquires multi-source raw data from the sensing module 140 in real time and performs signal decoupling and dynamic denoising operations on this data. This processing aims to eliminate interference from the gravitational component and background vibration interference generated by the mechanical operation of the robot during operation, thereby extracting physical feature quantities that characterize the contact state between the roller brush and the working surface.

[0032] The sensing module 140 continuously collects motion state data of the robot. The current sensor 142 collects the raw phase current of the brush motor 121 at a fixed sampling frequency. This raw phase current includes a sustaining current to overcome mechanical no-load resistance and a load current to overcome surface friction resistance. To prevent high-frequency electromagnetic noise from interfering with subsequent calculations, the main control module 150 first performs low-pass filtering on the collected raw phase current to obtain a smoothed real-time current value. Simultaneously, the attitude sensor 141 outputs the robot chassis's attitude angle relative to the horizontal plane in real time. Because the photovoltaic modules are installed at an angle, the proportion of gravity in the load of the brush motor 121 varies with the slope; therefore, directly using the raw current cannot accurately characterize the resistance characteristics of the surface dirt.

[0033] The main control module 150 uses a dynamic model to perform gravity compensation processing on the original phase current. The system pre-stores the physical parameters of the roller brush assembly. Unlike existing technologies, this invention modifies the physical model of the influence of gravity on the rotating motor current. Gravity does not directly generate rotational resistance torque, but rather changes the normal pressure of the roller brush assembly on the photovoltaic surface, thereby changing the sliding friction torque between the roller brush bristles and the photovoltaic surface. The main control module 150 calculates the change in roller brush pressure against the ground based on the real-time acquired attitude angle, and then derives the change in reference friction current caused by the change in gravity. The specific decoupling calculation follows the following modified physical model: ; In the formula, The decoupled pure triboelectric current (characterizing the viscous resistance of the stain). The original phase current after acquisition and filtering; For the effective quality of the roller brush assembly; It is the constant of gravitational acceleration; The effective radius of the roller brush; The reference coefficient of dynamic friction between the roller brush bristles and the photovoltaic panel being cleaned; The slope angle is calculated by combining the pitch and roll angles detected by attitude sensor 141, and represents the angle between the gravity vector and the photovoltaic module normal vector. This is the torque constant of the motor; This refers to the inherent mechanical loss current of the motor when it is completely suspended and without contact.

[0034] Through this calculation, the system eliminates fluctuations in the basic frictional torque caused by changes in the robot's posture, ensuring that changes in pure frictional current are only related to the additional resistance caused by the dirt.

[0035] Meanwhile, the main control module 150 performs frequency domain analysis and attenuation denoising on the raw vibration signal collected by the vibration sensor 143. The signal collected by the vibration sensor 143 is a mixture of high-frequency frictional vibration generated by the contact between the roller brush and the dirt, and low-frequency mechanical background noise generated by the operation of the robot's own walking motor 111 and roller brush motor 121. In order to separate the effective contact vibration characteristics, the system adopts a spectral subtraction strategy based on rotational speed mapping.

[0036] The main control module 150 has a pre-installed reference lookup table mapping speed to noise in its internal memory. This lookup table records the inherent background noise power spectral density of the system under different combinations of the travel motor 111 speed and the brush motor 121 speed. To address the noise reference drift problem caused by mechanical aging, the system introduces a dynamic reference update mechanism: when the system determines that it is currently on a smooth and clean surface (i.e., and When the background noise is consistently below the threshold, the main control module 150 will sample the current background noise and update the corresponding value in the lookup table.

[0037] During real-time operation, the main control module 150 first performs a fast Fourier transform on the original vibration signal to obtain its full-frequency spectrum information. Subsequently, the system reads the current rotational speed of the walking motor 111 and the rotational speed of the brush motor 121, and performs indexing and bilinear interpolation in the reference lookup table to obtain the accurate background noise reference value under the current working conditions.

[0038] The main control module 150 calculates the net vibration energy within the selected effective characteristic frequency band. This characteristic frequency band is selected to avoid the motor's fundamental frequency and its low-order harmonics. In this embodiment, the characteristic frequency band is set to [500Hz, 2000Hz]. The net vibration energy is calculated by subtracting the weighted background noise reference from the power spectrum of the original signal. The calculation model is as follows: ; In the formula, The calculated net vibrational energy; and These are the lower and upper limits of the effective characteristic frequency band, respectively. The power spectral density is the result of the original vibration signal after fast Fourier transform. The background noise reference power spectral density is obtained by table lookup and interpolation; The noise suppression coefficient is 1.2 to 1.5 in this embodiment.

[0039] If the calculated difference is negative, the system forces it to zero. The net vibration energy obtained after the above processing eliminates the interference of robot body operating noise and can characterize whether there are particulate stains on the surface of the photovoltaic module.

[0040] See attached document Figure 5 , Figure 5 This is a schematic diagram of the modeling and condition classification logic for stain impedance characteristics according to an embodiment of the present invention. The main control module 150 uses the pure triboelectric current and net vibration energy obtained in the aforementioned steps to construct a stain impedance characteristic model through multimodal data fusion technology, and classifies the working condition of the current working surface in real time accordingly.

[0041] The main control module 150 performs statistical feature extraction on the real-time input pure frictional current data and net vibration energy data. To eliminate the impact of instantaneous signal jumps on the stability of operating condition judgment, the system uses a sliding time window algorithm to smooth the data. The main control module 150 maintains two independent one-dimensional circular queues, which are used to store the pure frictional current values ​​and net vibration energy values ​​of the most recent sampling periods, respectively.

[0042] To address the low automation level caused by reliance on manual calibration in existing technologies, this invention employs a power-on self-test calibration strategy. In the initial phase of each robot operation (e.g., the first 3 seconds), the main control module 150 controls the roller brush to rotate at a low speed and slightly lift it (or uses mechanical limits to keep the roller brush at minimum pressure), automatically collecting the average current value at this time as the calibration result for that operation. This enables automatic correction of zero-point drift.

[0043] The main control module 150 constructs a stain resistance index vector based on the aforementioned statistical characteristics. This vector is a two-dimensional state vector that integrates a time-domain component reflecting low-frequency mechanical properties and a frequency-domain component reflecting high-frequency contact properties. The calculation model for the stain resistance index vector is as follows: ; In the formula, The stain resistance exponent vector; This is the viscous impedance component; This is the roughness impedance component; and These represent the length of the sliding window, typically ranging from 50 to 100. As a normalization scaling factor, in this embodiment, the value range is determined based on the ratio of the sensitivity of the current sensor 142 to the vibration amplitude. to It is used to map the magnitude of vibrational energy to a range comparable to the value of electric current.

[0044] The main control module 150 classifies the current operating condition based on the calculated stain resistance index vector using a preset decision logic tree: 1. High-viscosity stain conditions: and ; 2. Loose dust accumulation condition: and ; 3. Complex pollution conditions: and ; 4. Normal operating conditions: Both components are below the threshold.

[0045] in and These are the current characteristic threshold and the energy characteristic threshold, respectively.

[0046] See attached document Figure 6 , Figure 6 This is a schematic diagram of a dynamic centroid position calculation strategy according to an embodiment of the present invention. The main control module 150 calculates the target position of the centroid adjustment module 130 using an electromechanical coupling model based on the working condition type identified in the aforementioned steps and the real-time slope information.

[0047] To quantify the control objective of the center-of-gravity adjustment module 130, the system defines the longitudinal displacement of the battery pack 131 relative to the geometric center of the chassis as the control variable. Movement towards the cleaning module 120 (i.e., the front end of the unit) is defined as the positive direction, and movement towards the drive wheels of the chassis module 110 (i.e., the rear end of the unit) is defined as the negative direction.

[0048] To resolve control logic conflicts when "highly viscous stains" and "sharp slopes" occur simultaneously, this invention introduces a "safety priority" arbitration mechanism. The main control module 150 first determines the current real-time slope. Has the preset safety limit been exceeded? (e.g., 20 degrees).

[0049] Scenario 1: Slope priority (traction enhancement mode).

[0050] like Regardless of the current stain condition detection results, the system prioritizes anti-slip requirements over stain removal requirements. The main control module 150 sets the target position to move backward to increase the drive wheel traction and prevent the robot from slipping or sliding due to excessive reaction force when removing stubborn stains. The target position is calculated as follows: ; in This is the absolute value of the maximum permissible mechanical travel.

[0051] It is worth noting that for uphill conditions (i.e., the machine nose is facing uphill), the system introduces an anti-overturning priority principle: if the rearward movement target position calculated by the above formula would cause the front support reaction force... If the value is less than the safety threshold (indicating a risk of longitudinal rollover), the system prioritizes satisfying anti-rollover constraints. A truncation process was performed to prevent battery pack 131 from shifting too far back.

[0052] Scenario 2: Cleaning Priority (Enhanced Cleaning Mode).

[0053] like When the system determines the operating condition to be "high-viscosity stain condition" or "compound contamination condition," the system enters the enhanced cleaning mode. The main control module 150 linearly calculates the target position for the battery pack 131 to move forward based on the magnitude of the viscous impedance component. ; in In this embodiment, the position gain coefficient is... The value is set from 50mm / A to 100mm / A, meaning that for every 1 ampere increase in resistance current, the battery pack 131 moves forward by 50mm to 100mm.

[0054] Scenario 3: Cruise Balance Mode.

[0055] When the operating condition is determined to be a normal operating condition, or a loose dust accumulation condition with a small slope, the main control module 150 will reset the target position to zero: .

[0056] Furthermore, the main control module 150 incorporates boundary constraint logic based on torque balance during the calculation process. To correct errors in the physical model, this invention explicitly distinguishes between the mass of the moving parts and the mass of the base. The system estimates the front wheel support reaction force based on the current slope and target position. Reaction force of the rear wheel support The revised mechanical model is as follows: ; In the formula, The mass of the robot body after removing battery pack 131; The battery pack has a mass of 131. The total mass of the machine; This is the horizontal distance from the body's center of mass to the rear wheel axle; This is the fixed horizontal distance from the chassis geometric center to the rear axle (i.e., the offset when the battery pack 131 is at zero position). This is the dynamic displacement of battery pack 131 relative to the geometric center of the chassis (forward is defined as positive). The height of the composite centroid of the entire machine; This is the horizontal distance between the rear axle and the effective support point at the front (this model is set in cleaning operation mode, with the front support point being the center of the brush contact with the ground, ignoring the suspended or lightly loaded front driven wheel).

[0057] If the calculated value of any reaction force is less than the safety threshold (e.g., 5N), the system will... Perform amplitude corrections to ensure that anti-overturning safety constraints are met.

[0058] See attached document Figure 7 , Figure 7 This is a schematic diagram of the kinematic interlock and safety reconfiguration control logic according to an embodiment of the present invention.

[0059] The main control module 150 calculates the position deviation between the current centroid position and the target position in real time.

[0060] When the position deviation is less than the safety threshold (e.g., 10% of the travel), the system determines that it is in a fine-tuning state, allowing movement while in motion.

[0061] When the position deviation value exceeds the safety threshold, the system determines that it is in a reconfiguration state and triggers a safety interlock.

[0062] Regarding the "car emergency lockout" problem, the logic of the "slippage escape exception clause" in this invention is as follows: The main control module 150 first detects slip ratio or abnormal acceleration.

[0063] If slippage is detected (e.g., wheel speed exceeds IMU integral speed by more than 20%), the system immediately bypasses all speed limits and smoothing algorithms (such as S-curve), and drives the battery pack 131 to move to the rearmost position at full speed to restore the normal pressure on the rear wheels in the shortest time. At this time, safety takes precedence over comfort.

[0064] If no slippage is detected, the main control module 150 first sends a deceleration command to reduce the walking speed to a safe creeping speed (e.g., 0.05 m / s), and then starts the linear drive 132. At this time, in order to protect the precision lead screw structure and prevent the machine body from shaking due to sudden changes in the center of gravity, the system uses an S-shaped speed curve generation algorithm to plan the movement trajectory of the battery pack 131.

[0065] Once the displacement encoder provides feedback that the position is correct, the interlock is released, and normal operating speed is resumed. The emergency stop protection logic is consistent with the previous description: when a risk of overturning is detected, the motor power is cut off and the position is locked.

[0066] See attached document Figure 8 , Figure 8 This is a schematic diagram of state maintenance and hysteresis exit logic according to an embodiment of the present invention.

[0067] Main control module 150 performs space hysteresis determination (dead zone control): ; When the deviation is less than the dead zone threshold (2mm to 5mm), no action is performed.

[0068] Main control module 150 execution time hysteresis determination (state maintenance): When the stain resistance index or slope data fed back by the sensing module 140 falls back to the normal operating condition threshold range, the delayed exit mechanism is triggered. The main control module 150 starts a countdown of 2 to 5 seconds.

[0069] If a high-load condition is detected again during this period, the countdown will be immediately reset to zero, and battery pack 131 will remain in the current enhanced position.

[0070] Only after the countdown has completely ended does the main control module 150 control the battery pack 131 to smoothly return to its geometric center. This logic effectively prevents frequent oscillations of the center-of-gravity adjustment mechanism when dealing with discontinuously distributed stains.

[0071] To verify the practical engineering effect of the adaptive centroid-adjusting photovoltaic cleaning robot proposed in this invention, a specific application example is constructed below.

[0072] In this embodiment, a photovoltaic cleaning robot with a total weight of 35 kg was selected. The battery pack 131, serving as a movable counterweight, weighs 10 kg, and the effective linear travel of the center-of-gravity adjustment module 130 is ±150 mm (with the geometric center as 0, and the direction towards the cleaning module 120 as positive). The robot operates in a mixed test field containing a flat photovoltaic array and a 25-degree inclined mountain photovoltaic array. The test field includes three typical areas: Area A is a flat area with dry, accumulated dust on the surface; Area B is a flat area but artificially coated with highly sticky stains simulating bird droppings; and Area C is a 25-degree inclined slope area.

[0073] After the robot is started, the main control module 150 first performs an initialization self-test.

[0074] When the robot is moving in area A (flat dust accumulation area), the data collected by the sensing module 140 shows that the attitude sensor 141 reports a pitch angle of 0.5 degrees, the current sensor 142 reports minimal current fluctuation in the brush motor 121, and both the viscous and roughness components in the calculated stain resistance index vector are below preset thresholds. The system determines this to be a normal operating condition and enters cruise balancing mode. At this time, the main control module 150 controls the linear actuator 132 to maintain the battery pack 131 at a 0mm position, ensuring even weight distribution across the front and rear axles, and performs rapid cleaning at standard power. The current of the walking motor 111 stabilizes at approximately 2.1A.

[0075] When the robot enters area B (highly viscous stains), the roller brush comes into contact with the sticky substance, causing a sudden increase in rotational resistance torque. Current sensor 142 detects a surge in the current of the roller brush motor 121 from 1.5A to 3.8A, while vibration sensor 143 detects a decrease in high-frequency vibration energy. The main control module 150 calculates the viscous resistance component. Significantly exceeds the current characteristic threshold The system determined that it was in enhanced cleaning mode. The main control module 150 calculated the target position based on a linear function relationship and drove the battery pack 131 to move in the positive direction to +120mm. This action significantly increased the normal pressure of the roller brush on the photovoltaic surface by utilizing the lever principle. Although the current of the roller brush motor 121 further increased to 4.5A at this time, the physical scraping ability was significantly enhanced, and the stains were removed in one go without slippage or jamming.

[0076] Subsequently, the robot transitioned from a flat area to area C (a 25-degree slope). Attitude sensor 141 detected that the pitch angle gradually increased to 25 degrees, exceeding the preset slippage threshold of 20 degrees. At this point, the downward component of gravity along the slope reduced the normal pressure on the robot's drive wheels, posing a risk of slippage. The system triggered a traction enhancement mode (anti-slip priority). Regardless of the resistance reported by the cleaning module 120, the main control module 150 forcibly intervened, and the drive battery pack 131 rapidly moved in the negative direction to -140mm (approaching the rearward mechanical limit). This action significantly shifted the robot's center of gravity towards the drive wheels, increasing the static friction limit between the drive wheels and the photovoltaic panels. The robot successfully maintained stable movement on the 25-degree slope and completed the climbing operation.

[0077] Experimental verification and effect comparison.

[0078] To quantitatively evaluate the technical effectiveness of this solution, a comparative experiment was conducted in the aforementioned test field. The experiment included a control group and an experimental group: Control group: A traditional photovoltaic cleaning robot with the same mechanical structure but with the battery pack 131 fixed in position (battery pack 131 is fixed at the geometric center).

[0079] Experimental group: The robot with dynamic centroid adjustment function described in this invention was used.

[0080] The experiment collected key indicators including cleanliness (calculated using image grayscale contrast), maximum slip ratio (the ratio of the difference between encoder speed and IMU integration speed), and average energy consumption. Data from multiple repeated tests are shown in the table below.

[0081] Table 1: Performance Comparison Test Data of Photovoltaic Cleaning Robots under Variable Operating Conditions

[0082] Conclusion Analysis Based on the data in Table 1, a clear technical conclusion can be drawn: Under normal flat operating conditions (T-01, T-02), there was no significant difference in performance between the experimental group and the control group, indicating that under low load conditions, the adaptive system can maintain the basic balance of the system without introducing additional energy consumption burden.

[0083] In cleaning tests targeting highly viscous stains (T-03 to T-06), the cleaning rate of the control group decreased significantly, reaching a minimum of only 58.7%. This was because the robot with a fixed center of mass could not provide sufficient downward pressure on the roller brush, causing the bristles to elastically deform and slide across the stain surface when encountering stubborn stains, failing to effectively break down the stain's adhesion. In contrast, the experimental groups (T-04, T-06) achieved a cleaning rate exceeding 90% by shifting the center of mass forward. Although this resulted in an increase in the current of the roller brush motor (121), indicating increased work, it effectively addressed the industry pain point of "incomplete cleaning."

[0084] In the steep incline tests (T-09 to T-12), the control group exhibited severe slippage, with maximum slippage rates reaching 28.4% and even 35.2%. This is because as the slope increases, the fixed center of gravity causes a sharp decrease in the normal pressure on the rear wheels, and the tire adhesion is insufficient to overcome the downward component of gravity along the slope. The increased slippage rate directly causes the drive motor 111 to idle, resulting in abnormally high current (e.g., 4.85A in T-09), which not only prevents it from completing the task but also poses a safety hazard of slipping backward. The experimental group, by shifting the center of gravity rearward, controlled the slippage rate to around 5% at a 25-degree slope (T-10, T-12), ensuring the traction performance of the drive wheels.

[0085] Especially in the comparison of extreme combined working conditions (sharp slope + high viscosity) between T-11 and T-12, the advantages of this scheme are most obvious. The control group fell into a dual failure state of insufficient traction and insufficient cleaning force (cleaning rate 41.3%, slippage rate 35.2%). While the experimental group, based on the control logic of "anti-tipping priority" and "traction enhancement", sacrificed some of the theoretical maximum cleaning pressure to a certain extent (to prevent excessive forward movement to avoid overturning), it still achieved a high cleaning rate of 88.7% while ensuring that the robot did not slip and fall.

[0086] The above data confirms that the dynamic centroid adjustment technology based on working condition recognition proposed in this invention has successfully solved the adaptability problem of photovoltaic cleaning robots in complex terrain and variable dirt environments, and achieved multi-objective optimization of cleaning efficiency and motion stability.

[0087] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A portable photovoltaic cleaning robot, characterized in that, include: Chassis module, used to provide driving force along the surface of photovoltaic modules; A cleaning module, located at the front end of the chassis module, includes a rotating roller brush; The center of gravity adjustment module includes a counterweight unit that can move along the longitudinal axis of the robot; Sensing modules are used to collect attitude, current, and vibration data; And the main control module, which is connected to the above modules; The main control module is configured to identify the slope and stain conditions based on the data from the sensing module, calculate and drive the counterweight unit to move to the target position, and adjust the normal pressure of the rotating brush and the driving adhesion of the chassis module by changing the overall center of gravity distribution, so as to reconstruct the state between cleaning enhancement, traction enhancement and cruise balance modes.

2. The portable photovoltaic cleaning robot according to claim 1, characterized in that, The centroid adjustment module includes: The longitudinal guide rail is fixedly installed on the top surface of the chassis module frame along the front and rear axis of the robot; A sliding seat, adapted and mounted on the longitudinal guide rail, is used to support the counterweight unit; and A linear actuator connects to and drives the sliding block to move linearly along the longitudinal guide rail; The counterweight unit is the robot's power battery pack, which is reused as a dynamically adjustable movable mass block; the linear actuator integrates a position encoder to provide feedback to the main control module on the real-time displacement of the battery pack relative to the robot's geometric center.

3. The portable photovoltaic cleaning robot according to claim 1, characterized in that, The sensing module includes: An attitude sensor, installed at the geometric center of the robot body, is used to collect the robot's pitch and roll angles. A current sensor, integrated into the drive circuit of the cleaning module, is used to collect the original phase current of the rotating brush motor; and A vibration sensor is installed on the surface of the mechanical housing of the cleaning module to collect contact vibration signals between the roller brush and the surface of the photovoltaic module. The main control module is configured to calculate the influence of gravity component on current based on the data from the attitude sensor, and to perform frequency domain analysis based on the data from the vibration sensor to extract net vibration energy within the characteristic frequency band.

4. The portable photovoltaic cleaning robot according to claim 3, characterized in that, The main control module has built-in signal decoupling logic and is configured to perform the following operations: Based on the real-time angle fed back by the attitude sensor, the change in reference friction current caused by the change in gravity is calculated using a dynamic model. The pure friction current characterizing the viscous resistance of the stain is obtained by subtracting the change in the reference friction current and the inherent mechanical loss current of the motor from the original phase current collected by the current sensor. The main control module uses only the pure frictional current and the net vibration energy as input variables to evaluate the working condition of the stain, thereby eliminating the interference of the robot's own posture changes and mechanical background noise on the identification of the working condition.

5. The portable photovoltaic cleaning robot according to claim 4, characterized in that, The main control module is configured to construct a stain resistance index vector to identify the stain condition type: The stain resistance index vector includes a viscous resistance component that reflects low-frequency mechanical properties and a roughness resistance component that reflects high-frequency contact properties. When the viscous impedance component exceeds a preset current characteristic threshold and the roughness impedance component is lower than a preset energy characteristic threshold, it is determined to be a high-viscous stain condition. When the roughness impedance component exceeds the energy characteristic threshold and the viscous impedance component is lower than the current characteristic threshold, it is determined to be a loose dust accumulation condition. When both components exceed their respective thresholds, it is determined to be a combined pollution condition.

6. The portable photovoltaic cleaning robot according to claim 1, characterized in that, The main control module has a built-in safety priority arbitration mechanism to resolve operating condition conflicts: When the attitude angle data indicates that the current slope exceeds the preset safety limit, regardless of the type of soiling condition, the system prioritizes the traction enhancement mode. The main control module sets the target position to move towards the side of the robot's drive wheel to increase the drive wheel's grip and prevent slippage. The system determines the cleaning enhancement mode only when the current slope does not exceed the safety limit and a high-viscosity stain or complex pollution condition is identified. The main control module sets the target position to move towards the robot cleaning end to increase the downward pressure of the rotating brush by using the lever principle.

7. The portable photovoltaic cleaning robot according to claim 6, characterized in that, The main control module incorporates anti-overturning boundary constraint logic when calculating the target position: The main control module uses a torque balance model to predict the normal support reaction forces at the front and rear support points of the robot based on the current slope data and the proposed target position. If any calculated normal support reaction force is less than the preset safety threshold, it indicates a risk of longitudinal overturning. The main control module performs amplitude limiting correction on the target position, forcibly restricting the counterweight unit to the stroke range that meets the anti-overturning safety constraints.

8. The portable photovoltaic cleaning robot according to claim 1, characterized in that, The main control module is also equipped with kinematic interlock logic to control the timing of the center of mass adjustment module's actions. When the deviation between the calculated target position and the current position exceeds a safety threshold, the system enters a reconstruction state. In the reconfiguration state, the main control module first controls the chassis module to decelerate to a preset creeping speed or stop running; After confirming that the chassis module speed is below the safety limit, the center of gravity adjustment module is driven to move, and the movement trajectory of the counterweight unit is planned using an S-shaped speed curve to limit the maximum acceleration and prevent the fuselage from shaking due to rapid movement of the center of gravity. The main control module will only release the interlock and restore the normal operating speed of the chassis module after the counterweight unit reaches the target position.

9. The portable photovoltaic cleaning robot according to claim 1, characterized in that, The main control module is also configured with state persistence and delayed exit logic: The main control module has a preset spatial position dead zone threshold. Motion control commands are only generated when the absolute value of the deviation between the target position and the current position is greater than the dead zone threshold. The main control module has a built-in state holding timer. When the detected stain condition or slope condition falls back to the normal range, the counterweight unit is not immediately reset, but a delay countdown is triggered. If a high-load condition is detected again during the countdown, the timer is reset and the current mode is maintained; the counterweight unit can only be controlled to return to the geometric center after the countdown has ended.

10. The portable photovoltaic cleaning robot according to claim 1, characterized in that, The main control module is also configured to execute a power-on self-test calibration strategy: In the initial stage of each robot operation, the rotating brush is controlled to rotate at a low speed and be in a state of minimum pressure. The average value of the motor current at this time is collected as the reference value of the inherent mechanical loss current for this operation. This value is used to correct the zero-point drift caused by mechanical wear or temperature changes, and the corrected reference value is used in subsequent signal decoupling operations.

Images