Dynamic balance adjusting system for underwater cleaning robot and control method of dynamic balance adjusting system

Abstract

The invention relates to a dynamic balance adjusting system for an underwater cleaning robot and a control method of the dynamic balance adjusting system, and the dynamic balance adjusting system comprises an inclination detection module which comprises a plurality of liquid level sensors located on the same transverse plane; the dynamic leveling module comprises a plurality of water spraying pumps which are located on the same transverse plane; the central control unit is electrically connected with the inclination detection module and the dynamic leveling module; the central control unit judges whether the robot body of the cleaning robot inclines or not and judges the position where the robot body inclines on the basis of distance data measured by the liquid level sensors. And when it is judged that the robot body of the cleaning robot tilts, the central control unit controls the water spraying pump located at the tilting position to start water spraying downwards, so that the balance posture of the cleaning robot body is corrected by generating upward water flow thrust. The technical problems that in the prior art, due to inclination or rollover of a robot, the water surface operation efficiency is reduced, and even the robot loses the autonomous working capacity are solved.

Description

Technical Field

[0001] This invention relates to the field of communication and control technology for cleaning robots, and in particular to a dynamic balance adjustment system and control method for underwater cleaning robots. Background Technology

[0002] Underwater cleaning robots have evolved to have dual-mode functionality, capable of both underwater sediment removal and surface debris cleaning. Their workflow typically involves first sucking in and collecting impurities, fallen leaves, and other debris underwater using a built-in filtration system. After completing the underwater task, they then surface to perform surface debris cleaning.

[0003] However, existing technologies present a common technical problem when switching from underwater to surface mode. The diverse shapes and densities of the debris collected underwater often lead to uneven distribution within the internal filters or debris bins, causing a shift in the robot's overall center of gravity. While underwater, the buoyancy of the water can mask this problem; however, when the robot surfaces, part of its volume leaves the water, altering its buoyancy. This shift in the center of gravity results in an imbalance of torque, causing the robot to tilt, with one side sinking and the other tilting.

[0004] This tilted posture will bring a series of adverse effects:

[0005] 1. Impact on surface operation performance. Surface cleaning requires the robot's suction nozzle to fit tightly against the water surface to create an effective adsorption flow field. Tilting the robot body disrupts this fit, resulting in a reduced adsorption area, suction leakage, and significantly reduced efficiency in collecting floating debris.

[0006] 2. Impact on autonomous operation capability. If the center of gravity shifts significantly, the robot may even capsize when surfacing, completely losing its autonomous operation capability. In this case, manual intervention is required to reset it, which contradicts the original design intention of the equipment's autonomy.

[0007] Therefore, effectively solving the problem of unbalanced buoyancy caused by uneven internal load when underwater cleaning robots are under load, and ensuring that they can smoothly and reliably transition from underwater mode to surface mode, has become a technical challenge to improve the reliability and practicality of such robots' autonomous operation throughout the entire process. Summary of the Invention

[0008] To address the problems existing in the prior art as much as possible, this invention provides a dynamic balance adjustment system and control method for underwater cleaning robots. The system aims to automatically detect and actively correct the posture imbalance caused by uneven internal load distribution when the underwater cleaning robot floats to the surface, thereby solving the technical problem in the prior art where robots reduce their efficiency on the water surface or even lose their autonomous working ability due to tilting or overturning.

[0009] This invention provides a dynamic balance adjustment system for an underwater cleaning robot, comprising:

[0010] The tilt detection module includes multiple liquid level sensors all located on the same horizontal plane. Each liquid level sensor is positioned on the top of the cleaning robot and is distributed at least at the four diagonal positions of the cleaning robot. Each liquid level sensor is used to measure the vertical distance from its respective position to the water surface in real time to obtain distance data characterizing the robot's posture.

[0011] The dynamic leveling module includes multiple water pumps all located on the same horizontal plane. Each water pump is installed at the bottom of the cleaning robot and corresponds vertically to each of the liquid level sensors.

[0012] The central control unit is electrically connected to both the tilt detection module and the dynamic leveling module.

[0013] The central control unit determines whether the body of the cleaning robot is tilted and the location of the tilt based on the distance data measured by each of the liquid level sensors.

[0014] When the cleaning robot is detected to be tilted, the central control unit controls the water pump located at the tilted position to start spraying water downwards, so as to correct the balance posture of the cleaning robot body by generating an upward water flow thrust.

[0015] In summary, the adjustment system of this invention structurally includes a tilt detection module, a dynamic leveling module, and a central control unit. The tilt detection module includes multiple liquid level sensors positioned on the same horizontal plane at the top of the robot. These sensors are distributed at least four diagonally opposite positions to measure their vertical distance to the water surface in real time, thereby obtaining distance data characterizing the robot's posture. The dynamic leveling module includes multiple water pumps positioned on the same horizontal plane at the bottom of the robot, their layout corresponding vertically to the positions of the top liquid level sensors. The central control unit interacts with the above two modules via electrical connection for data and command exchange. The system's workflow is as follows: The central control unit receives and analyzes the real-time distance data collected by the tilt detection module. By comparing the differences in the distance values, it determines whether the robot is tilted and determines the specific location of the tilt based on the position with the smallest value. If the robot is determined to be tilted, the central control unit instructs the water pumps located below the tilted position to spray water downwards, generating an upward water flow thrust. This thrust forms a corrective torque opposite to the tilting torque, thereby restoring the robot's balanced posture.

[0016] The present invention also provides a control method executed by the above-described dynamic balance adjustment system for underwater cleaning robots, comprising the following steps:

[0017] S1. Distance data acquisition steps: When the cleaning robot floats to the water surface, the vertical distance data from each sensor to the water surface, which represents the posture of the cleaning robot, is acquired in real time through the various liquid level sensors set on the top of the cleaning robot.

[0018] S2. Tilt determination step: Based on the difference between each distance value in the vertical distance data group, determine whether the cleaning robot has tilted and determine the position of the tilt.

[0019] S3. Balance adjustment step: If it is determined that the body of the cleaning robot is tilted, control one or more water pumps located at the bottom of the cleaning robot corresponding to the tilt position of the body to spray water downwards, so as to generate an upward water flow thrust and correct the balance posture of the cleaning robot.

[0020] During the correction process, steps S1, S2 and S3 are continuously executed in a loop until the difference between the distance values ​​in the vertical distance data set is less than a preset tilt threshold.

[0021] Specifically, the system operation logic of the control method of the present invention is as follows: First, in the distance data acquisition step (S1), when the robot completes underwater operations and rises to the surface, the system is activated, and the liquid level sensors at the top begin to continuously collect their vertical distances to the water surface, forming a real-time updated vertical distance data set. Next, in the tilt judgment step (S2), the central control unit receives this data set, quantifies the robot's tilt degree by calculating the difference between the distance values ​​within the data set, and determines the specific tilt position of the robot body based on the distribution of the differences. Subsequently, in the balance adjustment step (S3), if the tilt degree is determined to exceed a preset range, the control system starts the bottom water pump corresponding to the determined tilt position, generating upward thrust by spraying water downwards to correct the attitude. A key technical effect of this method is the realization of a closed-loop feedback control mechanism, that is, during the correction process, the system continuously and cyclically executes the entire process from data acquisition (S1), tilt judgment (S2) to balance adjustment (S3). As the water pump operates, the robot body attitude is gradually corrected, and the distance differences collected by the liquid level sensors also decrease accordingly. The central control unit continuously monitors this difference until it decreases to below a preset tilt threshold, at which point it determines that the robot has returned to a horizontal position and stops adjusting. This cyclical feedback ensures the accuracy of the adjustment and the self-terminating nature of the process.

[0022] According to the aforementioned technical solution of the present invention, in step S1, the number of liquid level sensors is set to four and distributed at the four diagonal positions on the top of the cleaning robot; in step S2, the number of water pumps is set to four and corresponding vertically to the four liquid level sensors. In summary, in this solution, the tilt detection module includes four liquid level sensors respectively installed in the four corner areas of the robot's top shell. Correspondingly, the dynamic leveling module includes four water pumps installed at the four corners of the robot's chassis. The position of each water pump in the vertical projection direction precisely corresponds one-to-one with the liquid level sensor on the top, establishing a direct mapping relationship between perception and execution. When any corner sinks, the distance value measured by the liquid level sensor at that position is the minimum, and the central control unit can directly locate the corresponding bottom water pump that requires a correction force. The technical advantages of this solution are: 1. This layout can comprehensively monitor the tilt of the robot in both pitch and roll dimensions; 2. The precise correspondence of the positions ensures that the application point of the correction thrust and the attitude sinking point are precisely matched, making the application efficiency of the correction torque the highest and the effect the most direct, and avoiding secondary disturbances caused by the deviation of the torque application point.

[0023] According to the aforementioned technical solution of the present invention, step S2 further includes: obtaining a distance difference by calculating the maximum and minimum distance values ​​in the vertical distance data set; and comparing the distance difference with the preset tilt threshold to determine whether the cleaning robot has tilted. In summary, in this solution, the central control unit processes the received vertical distance data set: first, it traverses the data set to identify the maximum and minimum distance values, and then calculates the distance difference between them. This distance difference physically and intuitively reflects the maximum tilt height difference of the robot's top plane. Subsequently, the algorithm compares the distance difference with a preset tilt threshold, and the comparison result serves as the basis for determining whether to trigger the balance adjustment program. The technical effects of this implementation are: 1. It provides a simple and physically meaningful method for quantifying tilt degree; 2. By simplifying multi-dimensional posture information into a single key scalar—the maximum distance difference—it simplifies the judgment logic of the control system, enabling the system to quickly and reliably determine whether the posture imbalance has reached a level requiring intervention, thereby improving the real-time performance and reliability of control decisions.

[0024] According to the aforementioned technical solution of the present invention, in step S3, the water spraying power of the water pump is proportional to the portion of the distance difference exceeding the tilt threshold. In summary, in this solution, when the robot is determined to have tilted, the output power of the activated water pump is dynamically correlated with the severity of the tilt. Specifically, the water spraying power of the water pump is proportional to the portion of the calculated distance difference exceeding a preset tilt threshold. Therefore, the more severe the tilt, the larger the portion exceeding the threshold, and the corresponding increase in the control signal (such as the duty cycle of the PWM signal) output by the controller to the water pump driver, thereby generating a stronger reverse thrust. This proportional control-based adjustment scheme achieves adaptive adjustment of the correction force. When the tilt amplitude is large, the system can output a larger power to achieve rapid and powerful correction; as the posture gradually returns to level, the distance difference decreases, and the correction force also smoothly weakens. This effectively avoids overshoot or oscillation caused by excessive force when approaching the equilibrium point, thus ensuring that the leveling process is both fast and stable, achieving a smooth transition from strong correction to fine-tuning.

[0025] According to the aforementioned technical solution of the present invention, the control method of the water spray pump in step S3 includes: when it is determined that the cleaning robot is tilted at a single corner, the water spray pump corresponding to that corner is activated individually; and when it is determined that the cleaning robot is tilted at two corners on the same side, the water spray pumps corresponding to the two corners on that side are activated simultaneously. In summary, the control logic of this solution aims to identify and distinguish between two tilting scenarios: one is single-corner sinking, where the distance value of the liquid level sensor at only one corner is significantly smaller than the other three; the other is same-side sinking, where the distance values ​​of the liquid level sensors at the two corners on the same side of the robot are both significantly smaller than the other side. The control scheme differs for these two situations: in single-corner sinking, only the single water spray pump corresponding to the sinking corner is activated; while in same-side sinking, both water spray pumps on the sinking side are activated simultaneously. The technical effect of this differentiated control method is that it enables the shape of the correction torque to accurately match the rectangular state of the tilting force causing the imbalance. For single-angle sinking caused by single-point center of gravity shift, applying force at a single point is the most direct and effective method; while for side sinking caused by excessive load on the entire side, applying a uniform lifting force along the entire side can avoid introducing new torsional torque during the correction process, thereby ensuring the stability of attitude correction and preventing secondary instability.

[0026] According to the aforementioned technical solution of the present invention, before step S1, an underwater attitude pre-calibration step is further included. This step includes: controlling the cleaning robot to perform a hovering action at a preset depth underwater; controlling each water pump to simultaneously perform a preset pulse water spraying action to apply an upward pulse thrust to the robot body; during the application of the pulse thrust, acquiring the attitude angle change data of the robot body through the inertial measurement unit; if the attitude angle change data of the robot body exceeds a preset stability threshold, then performing preventive calibration intervention on the cleaning robot underwater until the robot body attitude angle recovers to the stability threshold range, and then continuing to rise. In summary, this solution adds an underwater attitude pre-calibration step before the robot fully surfaces, and its specific process is as follows:

[0027] 1. Control the robot to hover at a preset safe depth (e.g., 0.5 meters underwater);

[0028] 2. Instruct all bottom water pumps to synchronously execute a brief and low-power pulse water spray action to apply an upward pulse thrust to the fuselage simulating the initial force of buoyancy, thereby stimulating the potential tilting tendency;

[0029] 3. During the application of pulse thrust, the robot's attitude angle (such as roll angle and pitch angle) change data are collected at high frequency by the internal inertial measurement unit (IMU);

[0030] 4. Compare the maximum change in attitude angle collected with a preset stability threshold. If it exceeds the threshold, perform preventive calibration intervention underwater and repeat this pre-calibration step until the attitude angle change data is within the stability threshold range before allowing the robot to continue to surface.

[0031] This method shifts the timing of attitude adjustment from a passive response at the water surface to active prevention underwater. Its technical advantage lies in its ability to anticipate and address imbalance risks in advance through a pre-surfacing self-check control scheme. This active prevention scheme helps avoid the possibility of the robot tilting or even capsizing violently at the hydrodynamically unstable air-water interface, thereby improving the smoothness and safety of the robot's transition from underwater to surface mode and significantly enhancing the reliability of its fully autonomous operation.

[0032] According to the aforementioned technical solution of the present invention, in step S3, the method of controlling one or more water pumps to spray water downwards is as follows: pulse spraying is adopted; and the pulse width of the pulse spraying is related to the difference between the distance values ​​in the vertical distance data set. The larger the difference, the longer the pulse width of the pulse spraying. In summary, in this solution, the central control unit applies thrust through a high-frequency "spray-stop" cycle, i.e., pulse spraying. In each working cycle, the pulse width (or duty cycle) of the water pump is dynamically adjusted according to the real-time monitored tilt degree (i.e., the magnitude of the distance difference). The adjustment logic is: the greater the tilt degree, the longer the pulse width, and the greater the impulse applied by a single pulse. The technical effect of this control method is that, through dynamic modulation of the pulse width, a fine and stepless adjustment of the correction torque from strong to weak is achieved. When the tilt is severe, a longer pulse width provides a strong corrective force for rapid response; as the attitude tends to be level, the pulse width shortens accordingly, making the corrective force gentler, thus effectively suppressing overshoot and oscillations that may be caused by system inertia. Compared to linear proportional control, this pulse control makes the leveling process smoother, more stable, and with higher attitude accuracy.

[0033] According to the aforementioned technical solution of the present invention, step S3 further includes: calculating the tilt axis characterizing the overall tilt direction and degree of the cleaning robot body based on the vertical distance data set; and assigning differentiated pulse widths to multiple water pumps located on the side of the tilt axis that are lower than the tilt axis, based on the distance from the position of each water pump to the tilt axis, wherein the larger the distance, the longer the assigned pulse width.

[0034] In summary, the specific implementation of this method in the balance adjustment step is as follows:

[0035] 1. Calculate the tilt axis: Based on the three-dimensional spatial coordinate data provided by four liquid level sensors, a spatial plane representing the current top posture of the robot is fitted by an algorithm (such as the least squares method), and the tilt axis vector that can accurately characterize the overall tilt direction and degree of the robot body is calculated.

[0036] 2. Assign Differentiated Pulse Width: Identify all water pumps located on the downward side of the inclined axis, and assign a differentiated pulse width to each water pump in the cooperative array based on its vertical distance from the inclined axis. The allocation principle is: the greater the distance (i.e., the greater the downward contribution), the longer the assigned pulse width.

[0037] The technical advantage of this method lies in elevating attitude correction from a point or line level to a vector level. By assigning weighted and differentiated thrust to multiple water pumps within the sinking area, the resulting multiple counter-thrusts can be synthesized into a vector lifting moment that is precisely opposite in direction and dynamically matched in magnitude to the tilting moment. This allows the leveling action to efficiently correct more complex tilt attitudes, while avoiding secondary tilting or attitude oscillations that may be caused by point-by-point or edge-by-edge adjustments, thereby effectively improving the efficiency, stability, and accuracy of leveling.

[0038] According to the aforementioned technical solution of the present invention, the method further includes: recording and storing tilt events occurring during the balance adjustment step, wherein the tilt events include robot task information and tilt feature information at the time of triggering the adjustment; performing correlation analysis on historical tilt events to establish correlation rules between specific tasks and specific tilt patterns; and, when the robot performs a specific task matching the correlation rules, predicting the position where the robot body will tilt according to the correlation rules, and controlling one or more water pumps corresponding to the predicted position in advance to perform preset compensating pulse spray.

[0039] In summary, this method can be summarized into the following control logic:

[0040] 1. Record tilt events: The system's built-in event log module records the tilt event in structured data after each successful balance adjustment, including the task information and tilt characteristic information when the adjustment was triggered.

[0041] 2. Establish association rules: During idle periods, the system performs association analysis on accumulated historical event data to uncover high-probability causal relationships between specific work tasks and specific skew patterns, and generates association rules accordingly.

[0042] 3. Predictive compensation: In subsequent work, when the system detects that the task currently being performed by the robot matches the established association rules, it predicts the position where the robot body will tilt based on the rules, and controls the water pump at that position to perform a preset, weak compensatory pulse spray in advance.

[0043] The technical advantage of this method lies in its ability to empower robots with self-optimization capabilities based on their own work experience through a closed loop of "memory-analysis-prediction-prevention." This allows the robot to proactively anticipate and counteract gradual shifts in its center of gravity caused by specific behavioral patterns using a small, continuous torque. This method helps to prevent severe tilting at its source, reduces the frequency of triggering high-intensity balance adjustment programs, and makes the robot's autonomous operation smoother and more stable, thereby elevating its intelligence level and environmental adaptability to a new level.

[0044] The technical advantages of the dynamic balance adjustment system and control method for underwater cleaning robots of the present invention are as follows:

[0045] 1. Precise attitude perception. By placing liquid level sensors at the four corners of the robot's top to directly measure the distance to the water surface, the differences in sensor readings can directly and accurately quantify the direction and degree of the robot's tilt in both roll and pitch dimensions, providing a reliable data basis for subsequent adjustments.

[0046] 2. Highly efficient torque correction. The one-to-one spatial correspondence between the water pump and the liquid level sensor in the dynamic leveling module ensures that the correction torque can be precisely applied to the sinking position that needs to be raised, achieving precise control, thereby improving adjustment efficiency and avoiding the occurrence of new imbalances.

[0047] 3. Stable closed-loop control. The entire control process forms a real-time, dynamic closed-loop feedback system that can dynamically adjust the behavior according to real-time changes in attitude and automatically stop after the attitude returns to a horizontal state, ensuring rapid response, smooth transition, and final attitude stability during the adjustment process.

[0048] In summary, by enabling the robot to quickly and smoothly recover its posture, this invention ensures the effective working area and suction power of the suction port during water surface operations. While significantly improving cleaning efficiency, it also effectively avoids the risk of tipping over due to severe imbalance, thereby enhancing the robot's autonomous working ability and environmental adaptability. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0050] Figure 1 This is a flowchart of the core control method of the present invention (regarding steps S1 and S2);

[0051] Figure 2 This is the flowchart of the core control method of the present invention (regarding step S3).

[0052] Figure label:

[0053] none. Detailed Implementation

[0054] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0055] Example 1

[0056] This embodiment provides a dynamic balance adjustment system and control method for an underwater cleaning robot, aiming to solve the technical problem that existing underwater cleaning robots, after collecting garbage, experience tilting or even tipping over when floating up, due to uneven internal load distribution, which affects their work stability and efficiency.

[0057] The dynamic balance adjustment system of this embodiment is integrated into the body of an underwater cleaning robot. This robot possesses conventional mobility drive devices (such as tracks or wheels) and a main circulation filtration system (including a main water pump and a waste collection bin). The adjustment system of this invention, as a modular component of the robot's functionality, works in conjunction with the robot's central processing core. Its system composition mainly includes a tilt detection module, a dynamic leveling module, and a central control unit. Specifically:

[0058] 1. Tilt detection module

[0059] This module is used to acquire the robot's attitude information on the water surface in real time and with high accuracy. In this embodiment, the module includes four independent liquid level sensors (labeled C1, C2, C3, and C4), all arranged on the same horizontal plane of the robot's top shell, and located in four diagonal areas (e.g., C1 is the front left corner, C2 is the front right corner, C3 is the rear left corner, and C4 is the rear right corner). To avoid physical interference from debris or floating objects, non-contact sensors, such as ultrasonic liquid level sensors or laser rangefinders, are preferred. These sensors calculate the vertical distance to the water surface by measuring the round-trip time of the signal, offering advantages such as high accuracy and high reliability. Each liquid level sensor is electrically connected to the central control unit to transmit real-time data.

[0060] Its working principle is as follows: When the robot floats to the water surface, four liquid level sensors work synchronously to measure the vertical distance from itself to the water surface, thus obtaining a set of real-time distance data (D1, D2, D3, D4). This data set can quantitatively characterize the robot's current posture. Under ideal horizontal conditions, the values ​​of D1 to D4 should be basically equal. If tilting occurs due to uneven load (such as the left front corner sinking), D1 will decrease significantly, while D4 at its diagonal position may increase accordingly. By comparing and analyzing these four distance values ​​in real time, the central control unit can accurately determine whether the robot is tilted and calculate the specific direction and degree of tilt.

[0061] 2. Dynamic leveling module

[0062] This module is the attitude correction execution unit. Based on instructions from the central control unit, it generates a controllable counterforce to restore the tilted robot body to a horizontal position. In this embodiment, the module consists of four independent miniature water pumps (labeled P1, P2, P3, and P4), mounted on the same transverse plane of the robot chassis. Their mounting positions correspond precisely to the positions of the four liquid level sensors on the top, meaning each water pump is vertically aligned with its corresponding liquid level sensor (e.g., P1 is directly below C1). This arrangement ensures a high degree of spatial coupling between the application point of the leveling force and the sensing point of the attitude change, achieving precise local attitude control.

[0063] Each water pump has a nozzle pointing vertically downwards. Upon startup, the downward-spraying water generates an upward reaction thrust on the robot body. The pumps can be DC brushless pumps, whose speed and power can be precisely and linearly adjusted via pulse-width modulation (PWM) signals. Each water pump is connected to an independent drive circuit that receives PWM control signals from the central control unit, enabling independent and precise control of the water spray intensity of each pump. This lays the foundation for subsequent refined leveling strategies.

[0064] 3. Central control unit

[0065] This unit is the control core and data processing center of the system, implemented by a microcontroller (MCU) or embedded processor integrated on the robot's main control board, which runs the core control algorithm of this invention. The central control unit is electrically connected to the four liquid level sensors of the tilt detection module and the four water pump drive circuits of the dynamic leveling module.

[0066] Its core functions are: to receive and analyze distance data measured by each liquid level sensor in real time; to determine whether the robot body is tilted and the specific location of the tilt based on this data; if tilting is detected, to generate and send control commands, which precisely control one or more water pumps located at the tilted position to spray water downwards through the corresponding drive circuit. The upward thrust generated by the water spray forms a corrective torque to counteract the tilting torque caused by uneven load, thereby actively correcting the robot's balance posture. Throughout the correction process, the central control unit continuously monitors the posture data, forming a closed-loop feedback control until the robot's posture returns to stability.

[0067] like Figure 1 and Figure 2 As shown, based on the above system architecture, the core of the control method disclosed in this embodiment lies in constructing a closed-loop feedback control process to achieve automated and precise adjustment of the robot's water surface attitude, as detailed below:

[0068] Once the underwater cleaning robot completes its underwater task and surfaces, with its top shell protruding above the water, the central control unit will automatically activate the dynamic balance adjustment system. This control method includes the following steps:

[0069] S1. Distance Data Acquisition Steps

[0070] Four liquid level sensors (C1-C4) mounted on top of the robot collect data at a preset sampling frequency (e.g., 10Hz), measuring the vertical distance from their respective positions to the water surface in real time. This yields a set of vertical distance data containing four distance values ​​(D1, D2, D3, D4). This data set provides a precise quantitative representation of the robot's current instantaneous attitude and is continuously transmitted to the central control unit.

[0071] S2, Steps for determining tilt

[0072] After receiving the vertical distance data set, the central control unit calculates the maximum distance value D in the data set in real time. max minimum distance value D min And obtain a distance difference ΔD = D max -D min This distance difference visually quantifies the robot's current maximum tilt. The central control unit then compares this distance difference ΔD with a preset tilt threshold ΔD. threshold Compare them.

[0073] In an optional embodiment, the tilt threshold ΔD threshold The following method can be used to determine the level: First, under the baseline state of the robot being unloaded and balanced, collect N sets (e.g., N=100) of liquid level sensor data in a still water surface, and calculate the inherent measurement and fluctuation noise level σ of the system. noise Secondly, based on the requirements for water surface cleaning operations, determine the maximum allowable height H of the suction port deviating from the water surface. max This height typically does not exceed 1 / 4 of the designed height of the suction port to ensure the effectiveness of the adsorption flow field. Taking all factors into consideration, this tilt threshold ΔD... threshold Can be set to 3σ noise To H max Between, for example, the possible values ​​are max(3σ) noise 0.5H max This ensures that the adjustment action can filter out the minute noise from the sensor and water surface fluctuations, while guaranteeing that the robot's core operational performance is not affected. Optionally, this embodiment sets the tilt threshold ΔD. threshold The adjustment distance is 2 cm to ensure that the adjustment action can filter out minor noise and guarantee operational performance.

[0074] The system response to the comparison results is as follows:

[0075] If ΔD≤ΔD threshold If the robot's posture is within an acceptable range, no adjustment is needed, and the system will continue to maintain the monitoring state in step S1.

[0076] Conversely, if ΔD > ΔD threshold If this is detected, it is determined that the robot has tilted in a posture requiring intervention. At this point, subsequent balance adjustment steps are triggered, and the result is recorded with respect to D. min The corresponding sensor locations are used to determine the areas of most severe subsidence.

[0077] S3, Balance Adjustment Steps

[0078] If the system determines that the fuselage is tilting, it immediately proceeds to this step. The central control unit first determines the tilt based on D... min The location is used to precisely pinpoint the sinking corner or side, and then the target output power P of the water pump is dynamically calculated based on the tilt degree ΔD. This calculation follows a proportional control function, the mathematical formula of which is:

[0079] P=k*(ΔD-ΔD threshold )

[0080] Where k is a preset proportionality coefficient, the value of which determines the strength of the correction response. This coefficient can be calibrated through the following experiment:

[0081] A1. Place a counterweight of known mass m at a single corner of the robot (e.g., the front left corner) to generate a defined tilting torque;

[0082] A2. After the robot stabilizes, measure the distance difference ΔD at this point;

[0083] A3. Gradually increase the output power P of the water pump at the corresponding corner (front left corner) until the distance difference ΔD detected by the liquid level sensor recovers to less than the tilt threshold.

[0084] A4. Record the power P in this equilibrium state. By repeating the above steps using multiple sets of counterweights with different masses m, a series of values ​​related to P = k*(ΔD - ΔD) can be obtained. threshold The data pairs are P and P. Linear regression analysis is performed on these data pairs, and the slope of the resulting straight line is the initial value of the proportionality coefficient k. This initial value can be further fine-tuned in subsequent dynamic tests to achieve the optimal balance between response speed and system stability.

[0085] After determining the target power P, the central control unit precisely controls the water pump based on the tilt position, specifically including the following two scenarios:

[0086] 1. Single Corner Tilting: An algorithm is used to determine if it is a single corner sinking. For example, four distance data points are sorted from smallest to largest to obtain D1, D2, D3, and D4. If the condition D2 / D1 > 2.5 (hypothetical value) is met, it is determined to be a single point with significant tilting. In this case, only the bottom water pump corresponding to the sinking corner is activated.

[0087] 2. Tilt of two corners on the same side: When it is determined that two corners on the same side of the fuselage sink at the same time (for example, D1 and D3 are significantly smaller than D2 and D4), in order to avoid generating new torsion, the central control unit simultaneously starts the water pumps corresponding to the two corners on that side and sends the same power control command to them, so that they work together to smoothly lift the fuselage on that side.

[0088] The entire balance adjustment is a continuous closed-loop feedback process. During the correction period, the system continuously executes S1, S2, and S3 in a loop. The central control unit recalculates ΔD′ in real time based on the latest attitude data and dynamically adjusts the output power of the water pump. As the attitude is corrected and ΔD′ decreases, the controller smoothly reduces the power accordingly, achieving a smooth transition from strong correction to fine-tuning. This cycle continues until ΔD′ is detected to be less than or equal to ΔD. threshold At that point, the system determined that the robot had returned to a horizontal position, and immediately stopped all water pumps, thus completing one adjustment process.

[0089] Example 2

[0090] To further improve the robot's autonomous operation stability and avoid adjustment difficulties caused by adjustment delays or excessive tilting after surfacing, this embodiment, based on Embodiment 1, provides a preferred control method that incorporates pre-surfacing prediction and calibration. Its core lies in advancing the timing of balance adjustment to the stage when the robot is still submerged underwater, thereby proactively preventing potential imbalance risks. The specific implementation is as follows:

[0091] First, regarding the hardware configuration: In terms of hardware structure, in addition to the tilt detection module and dynamic leveling module described in Embodiment 1, the system of this embodiment further integrates an inertial measurement unit (IMU) inside or near the robot's central control unit. This IMU includes sensing elements such as a three-axis accelerometer and a three-axis gyroscope, enabling it to measure the robot's attitude angle data (such as roll and pitch) in three-dimensional space at high frequency, and its performance is not limited by water depth. To ensure the accuracy of the measurement data, the IMU is rigidly fixed inside the robot's shell, preferably installed near its geometric center or center of mass.

[0092] Then, regarding the underwater attitude pre-calibration step: Based on the above system configuration, this control method adds an underwater attitude pre-calibration step before step S1, the specific process of which is as follows:

[0093] 1. Hovering Self-Check: After completing the underwater cleaning task and receiving the surfacing command, the robot does not immediately and continuously surface. Instead, it first surfaces to a preset self-check depth (e.g., 0.3 to 1 meter below the surface, preferably 0.5 meters) and enters a static hovering state. This depth must ensure that the robot is completely submerged underwater to eliminate external interference such as surface waves.

[0094] 2. Pulse Excitation: Upon entering self-test mode, the central control unit instructs all bottom water pumps (P1-P4) to synchronously execute a preset pulse water jet action, applying an upward pulse thrust to the fuselage. This action aims to simulate the initial force change during ascent through a gentle and purely vertical upward thrust disturbance, thereby "exciting" the potential tilting tendency caused by the internal center of gravity shift. The pulse parameters are pre-calibrated to ensure that the applied thrust is sufficient to induce an attitude change detectable by the IMU, but does not cause excessive vertical displacement.

[0095] Regarding the calibration method: When hovering underwater, start with low-energy pulses and gradually increase them to find the minimum pulse energy combination that makes the standard deviation of the IMU attitude angle change significantly higher than the static noise (e.g., more than 5 times) while the robot's vertical displacement is less than a predetermined value (e.g., 1 cm). This combination is then used as the pulse parameter for self-testing.

[0096] Regarding parameter examples: In this embodiment, the pulse duration can be set to 100 milliseconds to 300 milliseconds, and the water spray power can be set to 10% to 20% of the rated maximum power of each pump.

[0097] 3. Attitude Data Acquisition: At the instant the pulse thrust is applied and within a very short time window thereafter (e.g., 500 milliseconds), the central control unit continuously acquires the fuselage's attitude angle change data via the IMU at a high sampling frequency (e.g., 200Hz). To ensure measurement accuracy, before issuing the pulse command, the system first acquires a segment of static IMU data and calculates its average value as a "zero-point reference." All subsequently acquired dynamic data is compared with this reference to accurately calculate the pure attitude angle offset caused by the pulse excitation.

[0098] 4. Judgment and Intervention: The central control unit analyzes the collected attitude angle change data and compares its maximum deviation with a preset stability threshold (e.g., the maximum angle change of roll or pitch angle is less than 2°).

[0099] If the offset does not exceed the threshold, the robot's center of gravity is considered balanced, the self-check passes, and the normal ascent procedure continues. If the offset of any attitude angle exceeds the threshold, the robot is deemed to have a risk of ascent imbalance, the system will stop ascent, and immediately perform preventative calibration intervention at the current underwater depth.

[0100] Regarding intervention methods: One method is that the central control unit activates the corresponding sinking-side water pump based on the precise tilt direction fed back by the IMU, directly correcting the attitude underwater. Another method is to execute a set of preset disturbance actions (such as "forward-sudden stop-reverse-sudden stop" or rapid rotation in place), using inertia and water flow to redistribute the internal debris, thereby improving the center of gravity balance.

[0101] 5. Closed-loop calibration: After performing a calibration intervention, the system returns to the self-test action of pulse excitation to conduct a new round of attitude stability assessment. This closed-loop process of "self-test-calibration-re-self-test" will continue until the latest self-test result meets the stability threshold requirement. Only when the attitude is finally confirmed to be stable will the robot exit the self-test mode and execute the complete ascent procedure.

[0102] This embodiment moves the attitude balance control node forward by probing before surfacing, changing passive adjustment to active prevention. This reduces the probability of the robot tilting or capsizing severely when it surfaces, ensuring a smooth and stable transition from underwater to surface mode and significantly enhancing the reliability of its fully autonomous operation.

[0103] Example 3

[0104] To further optimize the linear proportional control scheme described in Embodiment 1, which may cause overshoot or oscillation problems under certain operating conditions (especially when the system response has inertia or the initial tilt amplitude is large), this embodiment provides an optimized control method to achieve more stable and precise attitude adjustment. Specifically:

[0105] The system hardware configuration of this embodiment is the same as that of Embodiment 1. Its core innovation lies in the control algorithm running inside the central control unit. Specifically, in the balance adjustment step (S3), the method of controlling the water pump is optimized to use pulse spraying.

[0106] The implementation steps of this control method are as follows:

[0107] The tilt detection and leveling triggering process (steps S1 and S2) is basically the same as in Example 1. When ΔD > ΔD threshold At this time, the system triggers a dynamic leveling procedure. Unlike continuously running the water pump and linearly adjusting the power, in this embodiment, the central control unit instructs one or more water pumps at the corresponding location to operate in a high-frequency "spray-stop" pulse mode. By applying a series of discrete and precisely controllable thrust pulses, the inertial effects in the fluid dynamics system can be overcome more effectively, thereby enabling fine-tuning of the robot's posture and avoiding overcorrection.

[0108] Regarding dynamic pulse width modulation, the key to this refined leveling process lies in the central control unit dynamically adjusting the core parameter of pulse injection—the pulse width—based on the real-time tilt degree ΔD. The core logic is: the larger ΔD is, the longer the pulse width.

[0109] In practical implementation, the pulse width is represented by the duty cycle of the pulse width modulation (PWM) signal. Within a fixed duty cycle T (e.g., 50 milliseconds), the actual spraying time T of the water pump... on The proportion occupied is called the duty cycle. To achieve this correlation adjustment, the central control unit has a preset mapping function or look-up table that establishes a deterministic correspondence between the tilt difference ΔD and the pulse width (duty cycle). For example, when the initial ΔD is large, the controller outputs a large duty cycle (e.g., 80%, i.e., 40 milliseconds of injection followed by 10 milliseconds of pause) to provide a strong initial correction torque for rapid coarse adjustment; as the attitude recovers and ΔD decreases, the controller smoothly reduces the duty cycle (e.g., to 30%, i.e., 15 milliseconds of injection followed by 35 milliseconds of pause) to weaken the correction force and achieve high-precision fine adjustment.

[0110] The mapping function or lookup table mentioned above can be established through offline experimental calibration methods, specifically including:

[0111] 1. In the experimental environment, a series of tilt states with different ΔD are artificially created by placing counterweights of known weight at different positions of the robot.

[0112] 2. For each tilt state, adjust the pulse width using an optimization algorithm (e.g., grid search or gradient descent) to find a value that makes the evaluation function J = w1*T. settle +w2*M p The optimal pulse width value to minimize, where T settle The time required for M to restore equilibrium p w1 and w2 are preset weighting coefficients representing the maximum overshoot during the adjustment process. By collecting data and performing curve fitting on multiple sets of correspondences between "tilt difference - optimal pulse width", the mapping function can be generated or it can be stored as a lookup table and pre-stored in the memory of the central control unit.

[0113] 3. By collecting data and fitting curves to the correspondence between multiple sets of "tilt difference - optimal pulse width", the mapping function can be generated or fixed into a lookup table.

[0114] The entire leveling process is a high-precision closed-loop feedback system. The controller calculates the initial pulse width based on the initial ΔD and sends the corresponding PWM signal to the driver of the target water pump. During pulse jetting, the tilt detection module continuously feeds back the latest attitude data at a sufficiently high frequency (e.g., a sensor sampling rate of 100Hz, much higher than the pulse frequency of 20Hz). In each control loop, the controller updates the pulse width parameter using the latest ΔD value through a mapping function and issues a new PWM signal.

[0115] The termination condition for adjustment is set so that when the controller detects that the latest ΔD is stably maintained within the preset threshold range for several consecutive times (e.g., 3 times), the system determines that the robot has returned to level, and at this time the duty cycle of all target water pumps is set to 0, and all pulse spraying is stopped.

[0116] This embodiment introduces a refined leveling control method based on pulse-jet injection. Compared with continuous linear control, it has the following technical advantages: 1. By dynamically adjusting the pulse width, stepless adjustment from strong to weak force is achieved, effectively suppressing over-adjustment and oscillation, making the leveling process smoother and more stable; 2. Pulse control can provide instantaneous high thrust and rapidly decrease it when the attitude approaches equilibrium, making the adjustment process responsive and convergent quickly; 3. Since the water pump does not work continuously, the average energy consumption may be lower under certain operating conditions, which helps to extend the driving time.

[0117] Example 4

[0118] To efficiently handle complex tilt attitudes (such as diagonal sinking or other asymmetric tilt patterns) that are difficult to process using the method described in Embodiment 3, this embodiment further optimizes the pulse-based leveling control by employing a more precise vector pulse combination control method based on tilt axis analysis. This method elevates leveling control from a point or line level to a vector level, achieving more accurate and faster attitude correction. The system hardware configuration of this embodiment is the same as that of the aforementioned implementation methods; its core innovation lies in the attitude analysis and control allocation algorithm running within the central control unit.

[0119] The core innovation of this method lies in the algorithm of the balance adjustment step (S3), which is implemented as follows:

[0120] 1. Coordinate system establishment and data mapping

[0121] When the leveling procedure is triggered, the central control unit first acquires real-time distance data (D1-D4) from the four liquid level sensors (S1-S4). The controller internally establishes a two-dimensional Cartesian coordinate system with the robot's geometric center as the origin, where the position coordinates of the four sensors (x, y, y) are... i y iGiven that, the controller has information about four three-dimensional spatial points at any given time: P1(x1,y1,D1), P2(x2,y2,D2), P3(x3,y3,D3), and P4(x4,y4,D4).

[0122] 2. Solution of Inclined Axis

[0123] The central control unit executes an algorithm to calculate the tilt axis, which characterizes the overall tilt direction and degree of the robot's tilt. First, using data from four three-dimensional spatial points, a vector calculation (e.g., calculating the cross product of two vectors consisting of three points, N = (P2-P1) × (P3-P1)) is used to determine the normal vector N = (A, B, C) representing the spatial plane of the robot's current top posture. Then, this normal vector N is projected onto the robot's own horizontal coordinate system (XY plane), and the two-dimensional vector in its opposite direction is defined as the tilt axis. The direction of this axis precisely characterizes the direction of the robot's most severe overall sinking, while its magnitude reflects the overall degree of tilt.

[0124] 3. Differentiated pulse width allocation

[0125] The tilting axis divides the robot's bottom into a "sinking half" and a "lifting half." The controller identifies all the water pumps located in the sinking half and treats them as a coordinated pulse array. For each water pump P in this array... i The controller calculates the vertical distance d from its position to the tilt axis. i The distance d i This intuitively reflects the contribution of the corner point to the current complex tilt attitude. The principle for allocating pulse width is: the larger the distance, the longer the allocated pulse width. Its linear mapping relationship can be expressed as:

[0126] DutyCycle i =C base *|V tilt |*d i

[0127] Where C base It is a basic proportionality coefficient, |V tilt | represents the overall degree of tilt, while d i This is then used as a weighting factor for the water pump. For example, suppose the robot tilts complexly along a left-front-right-rear diagonal, with the tilt axis roughly along this diagonal direction, and the left-front corner (P1) sinks the most. In this case, the other water pumps located within the sinking half-zone (such as P2 and P3) have non-zero vertical distances d2 and d3 from this axis, while point P1 itself contributes the most to the tilt. The controller will use the distance d from each point to the axis as a weighting factor. iThe differentiated pulse widths were calculated so that the corner point farthest from the axis and with the greatest contribution to the sinking received the longest pulse width, while other corner points located in the sinking half-zone received correspondingly shorter pulse widths. The water pumps located in the lifting half-zone were not activated (duty cycle was 0).

[0128] When multiple water pumps in the cooperative array simultaneously pulse-jet according to the aforementioned differentiated pulse parameters, the multiple upward thrusts they generate will combine at the bottom of the robot into a vector lifting torque that is precisely opposite in direction to the tilting torque and dynamically matched in magnitude. The entire process is also under high-precision closed-loop control. The controller will continuously and frequently re-execute the above steps, updating the tilting axis and the pulse parameters of each pump in real time until the robot's posture returns to horizontal.

[0129] This embodiment elevates attitude correction to the vector level, enabling precise analysis and handling of complex tilting conditions, including diagonal tilt, thereby effectively improving the efficiency, stability, and accuracy of leveling.

[0130] Example 5

[0131] To enhance the robot's autonomous balancing capability from passive response to active prevention, this embodiment, based on the pulse-based balancing method described in Embodiments 3 and 4, further adds a control scheme with self-learning and optimization capabilities. The core of this scheme lies in introducing a predictive pulse compensation mechanism based on historical data correlation. Specifically:

[0132] This embodiment requires no additional hardware components, but it does require the central control unit to have a certain capacity of non-volatile storage space (e.g., Flash memory or eMMC). Based on the original dynamic balance adjustment method, this method further includes the following steps:

[0133] 1. Record and store tilt events

[0134] The central control unit embeds an event log module, which automatically records key information about each tilt event in structured data entries to local storage after each successful attitude correction. A structured tilt event record may include:

[0135] Event identifier and timestamp: Used to uniquely identify and record the time when an event occurred;

[0136] Trigger context: Records the work task information before the leveling program is triggered, such as task mode (e.g., "cleaning along the wall"), task parameters (e.g., direction is clockwise) and work position information;

[0137] Tilt characteristics: A detailed description of the tilt, which may include the tilt axis vector calculated in Example 4, or the maximum sensor distance difference before leveling.

[0138] Correction scheme: Record the key pulse parameters used to complete this leveling, such as the average or peak pulse width of each water pump.

[0139] 2. Establish association rules

[0140] The central control unit activates a correlation analysis engine during idle periods (e.g., while charging) or periodically. This engine processes stored historical tilt event data to uncover high-probability causal relationships between specific work behaviors and specific tilt patterns. Specifically:

[0141] Data preprocessing: Before analysis, the continuous tilt feature information is discretized. For example, the tilt axis vector is quantized into eight discrete direction terms (such as 'left front tilt') according to its direction angle, and the geographical location is divided into different grid area terms (such as 'shallow water area of ​​the pool').

[0142] Association rule mining: This analysis engine uses association rule mining algorithms (such as Apriori or FP-Growth algorithm) to discover frequently occurring itemsets in the dataset and calculate the support and confidence of these rules;

[0143] Policy library generation and updates: When the confidence level of a rule exceeds a preset threshold (e.g., >75%), the system learns a reliable association rule and generates a predictive compensation policy based on it, which is stored in a dedicated policy library. This policy library is dynamically updated. The system can introduce weighting or aging mechanisms to increase the weight of frequently validated effective policies and decrease or even remove the weight of ineffective control policies. At the same time, new policies are generated periodically to achieve continuous optimization of the system.

[0144] 3. Implement anticipatory compensation

[0145] During the robot's subsequent work, the central control unit monitors its current working status in real time. Once the current status matches the trigger condition of a policy in the "policy library," the system activates the corresponding compensation action in advance. This compensation policy can be constructed as an "IF-THEN" rule, specifically:

[0146] IF (Trigger Condition): Precisely defines high-risk work behaviors that lead to a tilt in predictability, such as: IFTask ID =='CounterClockwise_Wall_Following'AND Task_Duration>

[0147] 5 minutes;

[0148] THEN (Compensation Action): Defines the compensation action that needs to be performed in advance to counteract this anticipated tilt. This action includes:

[0149] (1) Target water pump: Based on the tilt direction obtained from the correlation analysis, determine one or more water pumps that need to be compensated for (usually the side that is predicted to sink).

[0150] (2) Compensation pulse parameters: Define a set of relatively weak pulse parameters, such as low-frequency pulses with a duty cycle of 2%-5%.

[0151] The counter-thrust generated by these weak compensating pulses is not intended to correct the tilt that has already occurred, but rather to act as a continuous, minute counter-balancing torque, providing early and synchronous dynamic cancellation throughout the process in which the center of gravity begins to shift unevenly due to a specific behavioral pattern.

[0152] This embodiment, through a closed loop of "memory-analysis-prediction-prevention," endows the robot with the ability to self-optimize based on its own work experience. This learning mechanism enables the robot to proactively anticipate and mitigate potential imbalance risks, significantly reducing the frequency of severe tilting events and making the robot's autonomous operation more stable and reliable, thereby elevating its intelligence and environmental adaptability to a new level.

[0153] The embodiments described above are for illustrative purposes only and are not intended to limit the scope of the invention. Various modifications and variations can be made to the above embodiments by those skilled in the art without departing from the spirit and scope of the invention, and all such modifications and variations should fall within the protection scope of the invention.

Claims

1. A dynamic balance adjustment system for an underwater cleaning robot, characterized in that, include: The tilt detection module includes multiple liquid level sensors all located on the same horizontal plane. Each liquid level sensor is positioned on the top of the cleaning robot and is distributed at least at the four diagonal positions of the cleaning robot. Each liquid level sensor is used to measure the vertical distance from its respective position to the water surface in real time to obtain distance data characterizing the robot's posture. The dynamic leveling module includes multiple water pumps all located on the same horizontal plane. Each water pump is installed at the bottom of the cleaning robot and corresponds vertically to each of the liquid level sensors. The central control unit is electrically connected to both the tilt detection module and the dynamic leveling module. The central control unit determines whether the body of the cleaning robot is tilted and the location of the tilt based on the distance data measured by each of the liquid level sensors. When the cleaning robot is detected to be tilted, the central control unit controls the water pump located at the tilted position to start spraying water downwards, so as to correct the balance posture of the cleaning robot body by generating an upward water flow thrust.

2. A control method executed by the dynamic balance adjustment system for an underwater cleaning robot according to claim 1, characterized in that, Includes the following steps: S1. Distance data acquisition steps: When the cleaning robot floats to the water surface, the vertical distance data from each sensor to the water surface, which represents the posture of the cleaning robot, is acquired in real time through the various liquid level sensors set on the top of the cleaning robot. S2. Tilt determination step: Based on the difference between each distance value in the vertical distance data group, determine whether the cleaning robot has tilted and determine the position of the tilt. S3. Balance adjustment step: If it is determined that the body of the cleaning robot is tilted, control one or more water pumps located at the bottom of the cleaning robot corresponding to the tilt position of the body to spray water downwards, so as to generate an upward water flow thrust and correct the balance posture of the cleaning robot. During the correction process, steps S1, S2 and S3 are continuously executed in a loop until the difference between the distance values ​​in the vertical distance data set is less than a preset tilt threshold.

3. The control method according to claim 2, characterized in that, In step S1, the number of liquid level sensors is set to four and distributed at the four diagonal positions on the top of the cleaning robot; in step S2, the number of water pumps is set to four and corresponding vertically to the four liquid level sensors.

4. The control method according to claim 2, characterized in that, Step S2 further includes: calculating a distance difference by measuring the maximum and minimum distance values ​​in the vertical distance data set; and comparing the distance difference with the preset tilt threshold to determine whether the cleaning robot has tilted its posture.

5. The control method according to claim 4, characterized in that, In step S3, the water spraying power of the water pump is proportional to the portion of the distance difference that exceeds the tilt threshold.

6. The control method according to claim 2, characterized in that, The control methods for the water pump in step S3 include: When it is determined that the cleaning robot is tilted at a single corner, the water pump corresponding to that corner is activated individually; and when it is determined that the cleaning robot is tilted at two corners on the same side, the water pumps corresponding to both corners on that side are activated simultaneously.

7. The control method according to claim 2, characterized in that, Prior to step S1, an underwater attitude pre-calibration step is also included, which includes: Control the cleaning robot to perform a hovering maneuver at a preset depth underwater; Control each water pump to simultaneously execute a preset pulse water spray action to apply an upward pulse thrust to the machine body; During the application of pulse thrust, the attitude angle change data of the fuselage are acquired through the inertial measurement unit; If the change in the robot's attitude angle exceeds the preset stable threshold, a preventative calibration intervention will be performed underwater on the cleaning robot until the robot's attitude angle returns to the stable threshold range before it continues to ascend.

8. The control method according to claim 2, characterized in that, In step S3, the method of controlling one or more water pumps to spray water downwards is as follows: Pulse injection is employed; and the pulse width of the pulse injection is related to the difference between the distance values ​​in the vertical distance data set. The larger the difference, the longer the pulse width of the pulse injection.

9. The control method according to claim 8, characterized in that, Step S3 further includes: Based on the aforementioned vertical distance data set, the tilt axis, characterizing the overall tilt direction and degree of the cleaning robot's body, is calculated; and... For multiple water pumps located on the downward side of the inclined axis, differentiated pulse widths are assigned to each water pump based on its distance from the inclined axis; the greater the distance, the longer the assigned pulse width.

10. The control method according to claim 8, characterized in that, The method also includes: The tilt events that occur during the balance adjustment step are recorded and stored, and the tilt events include robot task information and tilt feature information when the adjustment is triggered. Correlation analysis of historical tilt events is conducted to establish correlation rules between specific work tasks and specific tilt patterns; and, When the robot performs a specific task that matches the association rule, it predicts the position where the robot body will tilt according to the association rule, and controls one or more water pumps corresponding to the predicted position in advance to perform preset compensating pulse spray.

Images