A method and control system for generating a cleaning path for a pool robot

Abstract

The application relates to the technical field of swimming pool cleaning equipment, and discloses a swimming pool robot cleaning path generation method and a control system, which comprises the following steps: reading environment parameters, extracting instantaneous depth and depth change rate, constructing a depth sequence, calculating local depth variance, activating the isobath tracking logic or the yaw angle locking traversal mode according to the comparison result, executing global traversal work, extracting the net surface resistance index by using the difference between the real-time total load current and the theoretical reference current, stopping the global traversal when the resistance index is greater than the resistance threshold, generating an orthogonal covering path instruction, improving the speed of a duct motor and increasing the current limiting threshold of a track and a roller brush motor, and when the resistance index is not greater than the resistance threshold, reducing the speed of the duct motor and resetting the current limiting threshold, and then completing the global work. Through coupling of the path instruction and the bottom load state, the application dynamically configures the driving system and the down pressure power parameter, and reduces the probability of robot motion jam in the variable resistance area.

Description

Technical Field

[0001] This invention relates to the field of swimming pool cleaning equipment technology, specifically to a method and control system for generating cleaning paths for a swimming pool robot. Background Technology

[0002] When performing underwater cleaning operations, pool robots still have some defects in their path generation and power control mechanisms.

[0003] In terms of path planning, existing control schemes typically use fixed trajectory generation logic, which cannot effectively switch the corresponding path coverage instructions dynamically based on the tilt state and flatness characteristics of the underlying working surface. This leads to the robot being prone to missing its path when facing complex terrain boundaries such as flat bottoms, slopes, and inclined pool walls.

[0004] In terms of low-level state perception, existing equipment mostly directly collects the total operating current of the drive motor to assess the resistance of the contact surface. However, because this electrical load data is mixed with the gravitational load caused by changes in the spatial orientation of the fuselage, as well as the fluid damping load caused by the aquatic environment, these interference factors cannot be eliminated. Consequently, the extracted load parameters cannot objectively and quantitatively characterize the true physical frictional resistance state of the working surface.

[0005] Due to the lack of objective resistance determination parameters and corresponding power scheduling mechanisms, robots often cannot obtain sufficient downforce and torque output limits in time when passing through high resistance areas, which can easily cause mechanical jamming of the drive system or track slippage. Furthermore, when leaving such areas, the instantaneous drop and sudden decrease in the power parameters of the existing equipment motors can easily disrupt the force balance of the chassis movement, thereby causing transient directional deviation of the body and reducing the continuity and stability of cleaning operations. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a method and control system for generating cleaning paths for a swimming pool robot. This solves the problems of missed paths due to poor terrain adaptability when existing swimming pool robots perform cleaning operations on complex terrains and surfaces with varying resistance, as well as cleaning termination and operational stagnation due to the lack of a power compensation strategy when facing high-resistance soiled areas.

[0007] To achieve the above objectives, the present invention provides the following technical solution: The first aspect of this invention provides a method for generating a cleaning path for a swimming pool robot. This method constructs a depth data sequence by fusing environmental parameters, assesses terrain smoothness by combining local depth variance, and switches between contour tracking control logic and yaw angle locking traversal mode accordingly to generate motion trajectory commands for executing a global traversal cleaning operation. During the global traversal cleaning operation, the difference between the real-time total load current and the theoretical reference current is extracted to calculate the net surface resistance index. When the net surface resistance index is greater than the resistance threshold, a local intervention strategy is triggered to stop the global traversal cleaning operation, generate an orthogonal coverage path instruction, and synchronously increase the output speed of the duct motor and adjust the current limit threshold of the track motor and the brush motor according to the net surface resistance index. When the net surface resistance index is not greater than the resistance threshold, the output speed of the duct motor is reduced, the current limit threshold of the track motor and the roller brush motor is reset, the path planning mode is called again according to the local depth variance, and the global traversal cleaning operation is completed.

[0008] This method establishes control steps that couple path commands with the underlying electrical load state, enabling dynamic configuration of the drive system and downward pressure for different friction states, thereby reducing the probability of the robot experiencing motion stagnation and path deviation in high-resistance areas.

[0009] Preferably, the process of reading multidimensional basic environmental parameters, extracting instantaneous depth and depth change rate, and constructing a depth data sequence includes: simultaneously acquiring hydrostatic pressure and spatial triaxial linear acceleration to obtain the vertical acceleration after removing the gravity component; calculating the original roughness depth based on the pressure difference between hydrostatic pressure and reference atmospheric pressure; dynamically allocating filtering confidence coefficients based on the vertical relative velocity obtained by integrating the vertical acceleration and the recent hydrostatic pressure fluctuation variance to perform data fusion calculations and output the instantaneous depth; obtaining the depth change rate by dividing the depth difference between adjacent sampling times by the global sampling time step, and binding the instantaneous depth with a timestamp for caching. This step uses spatial kinematic parameters to numerically correct the hydrostatic depth, reducing the depth measurement error caused by water ripples and aircraft sway.

[0010] Preferably, the process of extracting local depth variance from the depth data sequence includes: establishing a sliding sampling window in the internal memory space, verifying the timestamp difference, filling in the instantaneous depth to form a depth evaluation sequence; calculating the arithmetic mean of the sequence, extracting the squared deviation of the height of discrete sampling points relative to the arithmetic mean, performing a cumulative summation operation, and multiplying by the discrete sample averaging coefficient to output the local depth variance. This step outputs a numerical feature quantity used to characterize the local bottom surface flatness.

[0011] Preferably, the process of activating the isobath tracking control logic that incorporates the depth change rate includes: introducing the depth change rate into the motion control closed loop; when the error of the depth change rate deviating from zero exceeds the set error dead zone range, generating a differential compensation command and superimposing it on the basic rotational speed parameters to produce a differential torque; controlling the motion heading to lock on the gravity equipotential surface to perform a spiral approximation maneuver; and simultaneously establishing a meshed isobath pseudo-3D topology map based on the distance traveled. This control logic is used to output lateral pose compensation commands in the inclined surface region.

[0012] Preferably, the process of activating the yaw angle locking traversal mode includes: triggering a state machine switch when the local depth variance remains no greater than the flatness characteristic threshold within the anti-jitter time window, suspending the differential speed compensation command update, and instead using the yaw angle obtained by the angular velocity integral as the absolute heading feedback reference, combined with preset plane coverage step size parameters to generate a motion trajectory command in a straight line interlacing pattern or a planar spiral pattern. This mode is used to output trajectory planning commands with a set step size in flat areas.

[0013] Preferably, the process of obtaining the reference current and extracting the net surface drag index includes: calculating the equivalent electrical load obtained by mapping the gravity component load, and the electrical load generated by overcoming the fluid resistance of the water body; summing the two with the no-load foundation current constant to obtain the reference current under theoretical pose conditions; and performing normalization processing using the difference between the real-time total load current and the reference current to obtain the net surface drag index. This calculation process isolates the electrical load components generated by fluid damping and gravity attitude, and extracts quantitative parameters characterizing the friction state of the chassis.

[0014] Preferably, the process of dynamically increasing the output speed of the ducted motor and raising the current limit threshold includes: calculating the speed compensation increment using the difference between the net surface resistance index and the resistance threshold, combined with the feedforward gain mapping proportional coefficient, and superimposing it on the base cruise speed to obtain the target control speed, driving the ducted motor to accelerate and generate the hydrodynamic downpressure increment; calculating the additional load torque based on the hydrodynamic downpressure increment and the static friction coefficient, converting it into a current compensation boundary using transmission constants such as the physical reduction ratio, and superimposing it on the initial configuration limits of the track motor and the brush motor. When the resistance characteristic exceeds the limit, this step outputs a control command that synchronously increases the downpressure of the ducted motor and the current limit value of the drive system, providing a torque range to overcome high frictional resistance conditions.

[0015] Preferably, the process of reducing the output speed of the ducted motor and resetting the current limiting threshold includes: calculating the discrete average value of the net surface resistance index and the resistance fluctuation variance within the verification time window; determining that the local area has been effectively cleaned when the discrete average value is less than the target cleaning confirmation benchmark and the resistance fluctuation variance is within the discrete distribution boundary value; subsequently, controlling the target control speed of the ducted motor to decrease according to the step-decreasing time constant, and simultaneously gradually reducing the current compensation boundary of the track motor and the brush motor according to the downpressure attenuation corresponding to the actual feedback speed, until it is restored to the initial configuration parameters. This step reduces the fuselage attitude deviation caused by transient changes in power parameters by extracting the resistance characteristic parameters of multiple sampling periods to trigger control switching and outputting a speed reduction command with a stepped time constant.

[0016] Preferably, the process of continuing the global traversal cleaning operation includes: reading the original breakpoint spatial coordinate data at the time of the interruption and performing matrix alignment verification; calculating the yaw angle difference between the relative target heading angle and the real-time yaw angle; inputting the yaw angle difference into the proportional-integral controller to generate a differential speed compensation adjustment, superimposing it on the basic straight-line speed, and outputting a left-side linear speed command or a right-side linear speed command to drive the track motor. This process is used to calculate and output the linear speed compensation command that drives the fuselage to converge towards the breakpoint coordinates after the intervention control command ends.

[0017] A second aspect of this invention provides a pool robot cleaning path generation and control system. This control system applies the aforementioned pool robot cleaning path generation method and includes a sensing module, a main control module, and a drive module. The sensing module includes a pressure sensor for collecting hydrostatic pressure, an inertial measurement unit for collecting spatial triaxial linear acceleration and spatial triaxial angular velocity, and a current sampling circuit for collecting steady-state operating current, used to transmit environmental measurement signals and electrical signals to the main control module; The main control module is equipped with an analog-to-digital converter and a hardware floating-point arithmetic unit, which are used to perform physical state calculation and path generation logic, and output low-level pulse width modulation signals and control commands to the driver module. The drive module includes track motors configured on both sides of the chassis to control linear movement and differential steering, roller brush motors to drive rotating brushing of the drive surface, and duct motors to generate negative pressure and hydrodynamic downforce. The duct motors are equipped with speed feedback units that monitor the operating status.

[0018] This system constitutes the physical hardware connection structure for performing environmental perception sampling, electrical load calculation, and motor power scheduling.

[0019] This invention provides a method and control system for generating cleaning paths for a swimming pool robot. It has the following beneficial effects:

[0020] 1. This invention extracts instantaneous depth and calculates local depth variance. Based on the comparison result of this variance with a threshold, it switches between the contour tracking control logic and the yaw angle locking traversal mode. This enables the robot to output corresponding path coverage instructions based on the tilt state and flatness characteristics of the underlying working surface, thereby reducing the trajectory omission rate in complex terrain boundary areas.

[0021] 2. This invention generates a theoretical reference current by calculating the gravity component load and fluid damping load of the robot, and extracts the net surface resistance index by using the difference between the real-time total load current and the reference current. This eliminates the interference caused by the water environment and the spatial posture of the robot body on the electrical load data of the motor, and outputs a quantitative judgment parameter for objectively characterizing the real frictional resistance state of the working surface.

[0022] 3. This invention ensures that the drive system has the upper limit of downforce and torque output without mechanical jamming when passing through high resistance areas by simultaneously increasing the duct motor speed and the drive motor current limit threshold when the net surface resistance index is greater than the set resistance threshold, and by adjusting the target speed and current compensation boundary by step-decreasing time constant after confirming that the area is clean. At the same time, it reduces the fuselage heading deviation caused by sudden drop in motor power parameters. Attached Figure Description

[0023] Figure 1 This is a block diagram of the pool robot cleaning path generation and control system of the present invention; Figure 2 This is the main flowchart of the swimming pool robot cleaning path generation method of the present invention; Figure 3 This is a logic diagram of the navigation mode switching and trajectory planning method of the present invention; Figure 4 This is a flowchart of the multi-parameter electromechanical-hydraulic coordinated intervention and state recovery closed-loop control of the present invention; Figure 5 This is a multi-source sensor deep fusion and terrain feature variance determination diagram of the present invention; Figure 6 This is a monitoring graph of the net surface resistance index after decoupling the dirt and friction characteristics of the present invention; Figure 7 This is a diagram illustrating the dynamic compensation of resistance feedforward speed regulation and underlying current limiting in this invention. Detailed Implementation

[0024] 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.

[0025] See attached document Figure 1 The present invention provides a pool robot cleaning path generation and control system, comprising: a sensing module, a main control module and a drive module.

[0026] The sensing module is configured inside the pool robot body. It includes a pressure sensor for collecting the hydrostatic pressure at the robot's location, an inertial measurement unit for collecting the robot's three-axis linear acceleration and three-axis angular velocity in space, and a current sampling circuit configured in the motor drive link for collecting the real-time steady-state operating current of the actuator motor.

[0027] The main control module is internally equipped with an analog-to-digital converter and a hardware floating-point arithmetic unit, and communicates with the sensing module to receive underlying environmental measurement signals and electrical signals, thereby performing signal decoding, filtering, and model calculation processing. Simultaneously, the main control module is electrically connected to the drive module, outputting underlying pulse-width modulation signals and control commands to it.

[0028] The drive module includes track motors configured on both sides of the chassis for controlling the robot's linear movement and differential steering, a roller brush motor configured at the bottom for driving the mechanical cleaning mechanism to perform surface rotation and brushing, and a duct motor configured in the top water outlet channel for drawing internal water flow and spraying it to the outside to generate negative pressure and hydrodynamic downforce. The duct motor or its drive control link is equipped with a speed feedback unit for monitoring the rotor's operating status.

[0029] See attached document Figure 2 This invention provides a method for generating a cleaning path for a swimming pool robot, comprising the following steps: S10. Read multi-dimensional basic environmental parameters and perform complementary fusion calculations using sensors. After the robot enters the water and lands on the bottom, the main control module reads the hydrostatic pressure output by the pressure sensor in real time, and simultaneously reads the vertical acceleration output by the inertial measurement unit. The main control module uses a filter model to perform complementary fusion calculations in the frequency domain of the hydrostatic pressure and vertical acceleration, extracting the instantaneous depth and depth change rate after filtering out high-frequency water flow disturbance noise, and continuously caching the instantaneous depth and its corresponding timestamp to construct a depth data sequence.

[0030] S20. Calculate the local depth variance of the terrain to trigger the navigation dimensionality reduction matrix. During the sliding sampling period of the movement trajectory, the main control module calculates the local depth variance based on the depth data sequence and compares it with a set flatness feature threshold. When the local depth variance is greater than the flatness feature threshold, it is determined that the current environment is an irregular curved surface, thus activating the contour tracking control logic; when the local depth variance is less than or equal to the flatness feature threshold, it is determined that the current environment is a flat terrain, thus activating the yaw angle locking traversal mode. Subsequently, the main control module generates the corresponding motion trajectory command based on the activated mode, and the driver module receives the motion trajectory command and executes the global traversal cleaning operation.

[0031] S30. Extracting Net Surface Resistance Characteristic Parameters through Joint Spatial Pose Calculation. During the global traversal cleaning operation, the current sampling circuit collects the real-time total load current of the track motor and the roller brush motor. The main control module combines the pitch angle and running linear velocity calculated by the inertial measurement unit and substitutes them into the underlying dynamic model to calculate the reference current under theoretical pose conditions. By calculating the difference between the real-time total load current and the reference current, the main control module performs normalization processing to extract the net surface resistance index, which reflects the physical surface friction resistance and the degree of dirt load.

[0032] S40. A multi-parameter electromechanical-hydraulic collaborative intervention strategy is introduced based on resistance characteristics and feedforward. When the main control module determines that the net surface resistance index continuously calculated and extracted at the current coordinate position exceeds the set resistance threshold, a local intervention strategy is triggered, the current global traversal expansion is stopped, and a new orthogonal coverage path instruction around the current coordinate point is re-planned and generated. At the same time, the main control module extracts the difference parameter between the net surface resistance index and the resistance threshold as the feedforward gain input of the control loop, and dynamically increases the output speed of the duct motor according to the feedforward gain input to increase the hydrodynamic pressure of the robot on the working surface of the pool bottom. In addition, the main control module simultaneously calculates the current limiting compensation amount of the transmission system according to the increment of the hydrodynamic pressure, and then dynamically adjusts the current limiting threshold of the track motor and the roller brush motor to maintain the dynamic balance between the electrical system and the hydrodynamic system.

[0033] S50. Perform local area cleaning verification and global state machine control loop recovery. During the execution of orthogonal coverage path instructions, the main control module cyclically executes the resampling calculation program for the net surface resistance index. After confirming that the net surface resistance index of the local grid area has stably fallen back to within the set resistance threshold, the main control module executes the reverse recovery timing sequence. First, it gradually reduces the output speed of the ducted motor to the base cruise speed, and then resets the current limiting thresholds of the track motor and the brush motor according to the attenuation of the hydrodynamic pressure. Finally, the main control module calls back the contour tracking control logic or yaw angle locking traversal mode according to the local depth variance parameter status of the current area, and controls the robot to return to the breakpoint coordinates to continue the global traversal cleaning operation.

[0034] In conjunction with the main control module's mechanism for reading the underlying measurement signals of the sensing module, to address the issue of high-frequency disturbance noise caused by hydrostatic pressure aliasing in the pressure sensor output due to dynamic pressure interference from water pump flow and surface waves during underwater operation of the pool robot, and to integrate the frequency domain response characteristics of heterogeneous sensors to extract reliable depth parameters, the main control module executes step S10, which involves reading multi-dimensional basic environmental parameters and performing sensor complementary fusion calculations. Step S10 executes the following specific sub-steps: S11. Synchronous Acquisition of Multi-Source Basic Environmental Parameters and Calculation of Raw Coarse Depth. After the robot touches the bottom of the water, the main control module reads the data from the sensing module based on the global sampling time step set by the internal hardware timer. For physical conditions where the sampling frequencies of heterogeneous sensors are inconsistent, the main control module introduces hardware timestamp alignment logic. Using the sampling clock of the high-frequency inertial measurement unit as the time cross-section reference, zero-order hold interpolation sampling is performed on the data from the low-frequency pressure sensor, thereby achieving temporal alignment of multi-source measurement data in the time dimension. For the timestamp synchronization processing of zero-order hold interpolation sampling, those skilled in the art can use standard digital signal processing interpolation functions. Its underlying time-domain synchronization mechanism is a well-known technology in the field and will not be elaborated upon here.

[0035] After completing data time alignment, the main control module synchronously extracts the hydrostatic pressure including dynamic pressure disturbances, and the spatial triaxial linear acceleration output by the inertial measurement unit. For the coordinate system rotation and gravitational acceleration component stripping of the output spatial triaxial linear acceleration to extract vertical acceleration, those skilled in the art can use a general spatial Euler angle matrix transformation algorithm; its spatial attitude calculation is a well-known technique in the field and will not be elaborated upon here. The main control module obtains the vertical acceleration after stripping the gravity component.

[0036] The main control module has preset reference atmospheric pressure, pool fluid density, and gravitational acceleration. The reference atmospheric pressure is obtained by the robot through initial zero-point calibration in the air environment before entering the water using a pressure sensor; the pool fluid density and gravitational acceleration are directly read as physical empirical parameters pre-configured in non-volatile memory; the global sampling time step is set by the system's main frequency division configuration.

[0037] The pressure difference between the hydrostatic pressure and the reference atmospheric pressure is calculated, and then divided by the product of the pool fluid density and the acceleration due to gravity. Before executing the division instruction in the underlying software algorithm, a parameter validity check program is called to confirm that the pool fluid density and acceleration due to gravity of the divisors are both greater than zero, thus avoiding the error of a zero denominator, thereby obtaining the original roughness depth. The original roughness depth reflects the depth reference based on absolute pressure conversion, which has the physical characteristics of an accurate low-frequency reference, but is mixed with high-frequency dynamic pressure noise signals.

[0038] S12. Vertical Relative Velocity Calculation and Complementary Fusion Filter Sequence Construction. The vertical acceleration acquired by the inertial measurement unit responds quickly to instantaneous high-frequency motion changes and possesses high-frequency response characteristics, but low-frequency zero-bias drift occurs when performing time integration on it. The main control module multiplies the vertical acceleration of the current cycle by the global sampling time step and accumulates it with the vertical relative velocity of the previous sampling cycle to calculate the vertical relative velocity of the current cycle.

[0039] Based on the physical principle of complementary high- and low-frequency characteristics of the sensor, the main control module constructs a complementary filtering model in the discrete-time domain. The complementary filtering model utilizes high-pass filtering characteristics to handle the vertical relative velocity variable and low-pass filtering characteristics to handle the original roughness depth variable. The main control module calculates the instantaneous depth based on the complementary filtering model. The calculation formula corresponding to the complementary filtering model is as follows: ; in, Indicates the current sampling time The instantaneous depth extracted; This represents the filter confidence coefficient, and its value range is set to the value within the interval (0,1); Indicates the previous sampling time Output instantaneous depth; This represents the vertical relative velocity calculated based on the inertial measurement unit. Indicates the current sampling time Compared with the previous sampling time The time difference between them, i.e., the global sampling time step; This represents the original roughness depth calculated based on the pressure sensor. This represents the high-frequency integral term derived from high-frequency acceleration integrals, used to reflect the high-frequency dynamic displacement response; This represents a low-frequency pressure calculation term based on low-frequency physical hydrostatic pressure conversion, used to anchor the absolute depth zero-bias constraint. It represents the increment of physical displacement change within a single global sampling time step.

[0040] Regarding the filter confidence coefficient The system employs a dynamic allocation mechanism where the filter confidence coefficient is dynamically allocated based on a comparison between the system's internally assessed recent hydrostatic pressure fluctuation variance and a preset benchmark hydrostatic pressure variance threshold. Specifically, the benchmark hydrostatic pressure variance threshold is obtained through long-term sampling statistics of the robot in a stable, static water environment during the factory calibration phase. When the assessed recent hydrostatic pressure fluctuation variance exceeds the benchmark hydrostatic pressure variance threshold, the system determines that it is currently in a high dynamic pressure disturbance condition and increases the value of the filter confidence coefficient to increase the weight of the velocity integral. When the recent hydrostatic pressure fluctuation variance is less than or equal to the benchmark hydrostatic pressure variance threshold, the filter confidence coefficient remains at a preset basic constant. Based on this dynamic allocation mechanism, when calculating instantaneous depth, the main control module tends to rely on the velocity integral of the inertial measurement unit in the high-frequency dynamic change range, while relying on the static benchmark of the pressure sensor in the low-frequency steady-state range, thereby filtering out high-frequency dynamic pressure interference and suppressing integral zero-bias drift.

[0041] The main control module binds the instantaneous depth output to the current system clock timestamp and writes it into an internal cache matrix. As the sampling period progresses, a continuous depth data sequence is built into the cache matrix. To prevent memory exhaustion caused by prolonged operation, the main control module employs a circular buffer data structure management mechanism in the cache matrix. When data writes to a node and reaches the physical boundary pointer of the matrix, it automatically loops back to the head node and overwrites stale data, ensuring stable operation of the computation flow. For the boundary pointer overflow judgment and node overwriting logic of the circular buffer, those skilled in the art can use standard circular arrays or doubly linked lists; the underlying memory addressing mechanism is well-known in the field and will not be elaborated here. Simultaneously, the main control module directly inputs the vertical relative velocity of the current period... The depth change rate for the current cycle is output. The processed instantaneous depth and depth change rate together constitute the basic reference input for subsequent navigation and control loops.

[0042] See attached document Figure 3 In this embodiment, combining the depth data sequence and depth change rate output in step S10, to overcome the accumulated odometry error caused by the lateral component of gravity during underwater dead reckoning of the pool robot, and to solve the adaptability problem of a single navigation mode in flat terrain, the main control module executes step S20, which calculates the local depth variance of the environmental terrain to trigger the navigation dimensionality reduction matrix. Step S20 executes the following specific sub-steps:

[0043] S21. Sliding Estimation of Local Depth Variance and Terrain Status Assessment. The main control module allocates a sliding sampling window with a fixed data capacity in its internal memory space. The fixed data capacity of the sliding sampling window is set based on the robot's front-to-back axis distance and the global sampling frequency to ensure that the sliding sampling window exactly covers the physical movement cycle of the robot crossing a single typical slope. Within the update cycle of the robot's movement trajectory, the main control module extracts instantaneous depths from the cache matrix according to the timestamp order. Before writing the data, the timestamp difference between adjacent extracted instantaneous depths is checked. To address the risk of data loss in the underlying data transmission, it is determined whether the timestamp difference is within the preset effective sampling time span. The effective sampling time span is calculated and determined based on the maximum packet loss tolerance of the communication bus and the basic sampling clock cycle. If the timestamp difference meets the effective sampling time span, the extracted instantaneous depth is filled into the sliding sampling window, thereby forming a depth evaluation sequence for the current local terrain.

[0044] The main control module calculates the local depth variance based on the depth evaluation sequence. The calculation process combines the data capacity of the sliding sampling window with the instantaneous depths in the depth evaluation sequence through mathematical statistical operations. The calculation logic for the variance is as follows: ; in, This represents the local depth variance calculated based on the sliding sampling window; Represents the average coefficient for discrete samples; Indicates the first The instantaneous depth corresponding to each discrete sampling point; This represents the arithmetic mean of all instantaneous depths in the depth evaluation sequence; This represents the relative height deviation between the instantaneous depth corresponding to a single discrete sampling point and its arithmetic mean. This represents the squared deviation of the instantaneous depth from the arithmetic mean corresponding to a single discrete sampling point; Indicates the first to the second The formula sums the squared deviations of each discrete sampling point. The physical meaning of the local depth variance output by this formula is to quantify the absolute elevation dispersion within the current sampling trajectory segment.

[0045] The smoothness feature threshold is pre-programmed into the non-volatile memory of the main control module. This smoothness feature threshold represents the amount of terrain undulation that the robot can tolerate without experiencing gravity-induced sideslip. As a preferred method, the smoothness feature threshold is obtained by the system through prior underwater traction experiments, combining the static friction coefficient model of the chassis tracks with the fuselage counterweight parameters. When the sliding sampling window is at full data load, the main control module compares the real-time calculated local depth variance with the preset smoothness feature threshold.

[0046] S22. Irregular surface contour tracing logic triggering and topology map construction. When the local depth variance exceeds the flatness feature threshold, the main control module determines that the robot is currently in an irregular surface terrain environment with undulating slopes. Under the physical conditions of an irregular surface terrain environment, the underwater gravity component will cause asymmetrical slip friction on both sides of the robot's tracks.

[0047] The main control module activates the contour tracking control logic, introducing the depth change rate output in step S10 into the motion control closed loop, and dynamically constraining the target feedback value of the depth change rate to approach zero through a proportional-integral-derivative (PID) control algorithm. Specifically, the error amount of the depth change rate deviating from zero is calculated. To address the high-frequency oscillation phenomenon of the actuator caused by small error inputs, a dead-zone nonlinear element is introduced for filtering. When the calculated error amount exceeds the set error dead-zone range, the main control module outputs a differential compensation command based on the error amount. The error dead-zone range is obtained by comprehensively calibrating the minimum starting dead-zone duty cycle of the track motor in the drive module and the mechanical backlash parameter of the reduction transmission mechanism. Simultaneously, to address the accumulation problem of the integral term in the PID control algorithm, the main control module applies anti-saturation boundary limiting processing, which is determined based on the upper limit of the maximum allowable pulse width duty cycle of the motor drive circuit.

[0048] The main control module superimposes the differential compensation command, after boundary truncation, onto the basic speed parameters of the drive module, causing the track motors to generate differential torque on both sides. Under the action of this differential torque, the robot's lateral yaw angle is continuously corrected, forcing the robot's heading to be locked on the gravitational equipotential surface. The robot thus performs a helical approximation motion along the isobath, counteracting the pose divergence caused by gravitational lateral slippage at the physical mechanism level.

[0049] During contour tracking, the main control module combines the robot's movement distance and instantaneous depth to build a gridded pseudo-3D topological map in the background system memory. For spatial gridding and coordinate node updates, those skilled in the art can use general grid map construction algorithms; the underlying grid mapping mechanism is a well-known technology in the field and will not be elaborated upon here.

[0050] S23. Navigation dimensionality reduction matrix triggering and yaw angle locking based on flat terrain. When the local depth variance is less than or equal to the flatness feature threshold, the main control module determines that the robot is currently in a flat terrain environment. In a flat terrain environment, the physical bottom surface lacks height gradient reference. If heading constraints are continuously applied based on the depth change rate, the water pressure measurement noise will be amplified by the control loop, causing yaw control oscillations.

[0051] The main control module executes a state machine transition to trigger the navigation dimensionality reduction matrix. To address control state oscillations at terrain boundaries, an internal timer is activated for state delay confirmation. The state machine transition is only executed when the continuously calculated local depth variance is less than or equal to the flatness feature threshold within a set anti-jitter time window. This anti-jitter time window is calculated by dividing the robot's body length by the current base cruising speed. After confirming the transition conditions are met, the main control module suspends the differential compensation instruction update in the contour tracking control logic, cuts off the depth change rate control feedback path to the drive module, and then activates the yaw angle lock traversal mode.

[0052] The main control module uses the spatial three-axis angular velocities collected by the inertial measurement unit to obtain the robot's yaw angle through integration. For the process of integrating the spatial three-axis angular velocities to calculate the Euler yaw angle, those skilled in the art can use a common quaternion update algorithm or direction cosine matrix algorithm; the spatial attitude integration mechanism is a well-known technology in this field and will not be elaborated further here. Based on the physical premise that the lateral component of gravity disappears under flat terrain, the physical resistance of the left and right tracks of the chassis tends to be symmetrically distributed. Under the state of disappearing lateral component of gravity, the inherent drift error of angular velocity integration is constrained within an acceptable engineering margin. The main control module uses the yaw angle as the absolute heading feedback reference for the control closed loop, and combines it with preset planar coverage step size parameters to generate motion trajectory commands in a straight-line interlacing pattern or a planar spiral pattern, and sends them to the drive module. The drive module controls the track motors to perform constant speed drive on both sides of the chassis or differential steering in place, achieving stable global traversal cleaning operations on the zero-gradient plane.

[0053] In conjunction with the global traversal cleaning operation performed in step S20, to accurately assess the physical friction characteristics and dirt adhesion level of the pool bottom surface, and to eliminate the interference of robot posture changes and hydrodynamic fluctuations on the load of the underlying execution motors, the main control module performs step S30, which involves joint spatial pose calculation to extract net surface resistance characteristic parameters. Step S30 executes the following specific sub-steps:

[0054] S31. Low-level load current synchronous sampling and steady-state feature extraction. In this embodiment, during the movement and cleaning operations performed by the drive module, the current sampling circuit acquires the main circuit current signals of the track motor and the brush motor in real time through the built-in shunt resistor and operational amplifier. The main control module calls the internal analog-to-digital converter to convert the main circuit current signal into a digital real-time total load current. Since the track motor generates high-frequency spike interference at the moment of commutation and load change, the main control module uses a set cutoff frequency to perform first-order hysteresis low-pass filtering on the digital real-time total load current to extract a smooth steady-state total load current.

[0055] As a preferred approach, the cutoff frequency is calculated based on the commutation fundamental frequency corresponding to the number of pole pairs of the motor in the drive module and the rated operating speed. The cutoff frequency is twice this commutation fundamental frequency, which can filter out high-frequency electromagnetic noise while preserving the complete mechanical load envelope signal. For the signal discretization of the analog-to-digital conversion unit and the construction of the software digital filter, those skilled in the art can use general digital signal processing algorithms. The underlying data conversion and filtering mechanisms are well-known technologies in the field and will not be elaborated upon here. The steady-state total load current reflects the robot's total energy consumption level under the current terrain and fluid environment.

[0056] S32. Multi-dimensional pose observation and theoretical reference current mathematical modeling. During the actual movement of the robot below the water surface, the electrical load borne by the drive module is affected by the combined effects of foundation friction, slope gravity components, and water flow resistance. To extract the foundation physical contact resistance reflecting the dirt on the bottom surface, the electrical load caused by the robot's posture and dynamic operation must be decoupled and eliminated. The main control module extracts the pitch angle calculated in real time by the inertial measurement unit, and simultaneously reads the robot's linear velocity through the encoder built into the drive module.

[0057] Because of frequency deviations in the data sampling clocks of different hardware components within the sensing module, the main control module uses the sampling timestamp of the current sampling circuit as a global reference. It then calls a linear interpolation function to perform time-domain alignment and resampling of the pitch angle and operating linear velocity, ensuring that the multi-dimensional environmental state variables involved in subsequent decoupling calculations remain consistent in time phase. The specific calculations for aligning the timestamps using the linear interpolation function can be implemented using a common first-order differential resampling algorithm, whose time-domain alignment mechanism is well-known in the field and will not be elaborated upon here.

[0058] After completing time series alignment, and combining the underlying dynamic response characteristics, the main control module establishes a reference current estimation model under theoretical pose conditions. The calculation formula corresponding to the reference current estimation model is as follows: ; in, Represents the reference current under theoretical pose conditions; The unloaded basic current constant represents the state when the robot is suspended underwater after being detached from the physical contact surface. It is measured and burned into the read-only memory under the suspension test conditions before the system leaves the factory. It represents the linear mapping coefficient from gravitational load to electrical current, reflecting the comprehensive conversion relationship between motor torque constant and mechanical reduction ratio; The underwater equivalent physical mass parameter of the robot body is obtained by subtracting the buoyancy of the water displaced by the robot from the robot body's physical gravity in the air, and then dividing this force difference by the gravitational acceleration. Represents gravitational acceleration; Indicates pitch angle The corresponding sine trigonometric function values; This represents the gravitational component load that the robot experiences along the direction of the running slope. This represents mapping the mechanical component of gravity along the slope to the equivalent electrical load on the drive link; The hydrodynamic damping mapping coefficient depends on the projected area of ​​the robot's outer shell facing the water and the surface streamline drag coefficient. This indicates the linear velocity of operation fed back by the drive module; This refers to the electrical load generated by the robot overcoming the resistance of the water fluid.

[0059] When the robot is determined to be in a flat terrain environment in step S20, the pitch angle... Approaching zero degrees, the value of the sine trigonometric function The equivalent zeroing process automatically eliminates the gravity component load term in the reference current estimation model, ensuring that the reference current calculation conforms to the actual physical state of operation on flat ground. The main control module calculates and outputs the reference current based on the current state variables.

[0060] S33. Resistance Characteristic Stripping and Dimensionless Index Normalization Calculation. The main control module subtracts the calculated reference current from the extracted steady-state total load current to obtain the net friction load current characterizing the additional friction load current applied to the external physical contact surface. To reduce individual hardware differences and establish a unified judgment benchmark, the main control module divides the net friction load current by the preset calibration current, performs normalization mapping processing, and finally extracts the net surface resistance index.

[0061] Before executing the division instruction, the main control module calls the boundary detection logic to verify the calibration current. If the calibration current is detected to be close to zero, the default exception handling mechanism is triggered, and a preset non-zero constant is assigned to the calibration current to avoid the microprocessor throwing a calculation overflow error. The specific value of the preset non-zero constant is set based on the minimum sustaining current value of the track motor during no-load operation. Regarding the acquisition of the calibration current under normal operating conditions, the calibration current is obtained by the robot automatically finding a flat test area with a continuous local depth variance of zero during the initial water landing stage, and by performing uniform linear movement and periodically averaging and filtering the net friction load current at this time. The formula for calculating the net surface resistance index is as follows: ; in, This represents the dimensionless net surface resistance index after normalization. This represents the steady-state total load current acquired by the current sampling circuit and extracted after filtering. This represents the reference current calculated by combining pose and hydrodynamic characteristics; This represents the net triboelectric load current after stripping the pose and fluid interference; This indicates the calibration current obtained during system calibration; This represents the comprehensive calculation process of performing division normalization on the obtained net friction load current.

[0062] After filtering the raw current data and decoupling the calculation from the theoretical benchmark, the net surface resistance index reduces the environmental physical interference from terrain undulations and water flow resistance. The net surface resistance index can independently reflect the frictional characteristics and the degree of micro-dirt adhesion of the current track motor and roller brush motor contact surface, thus providing a basis for subsequent judgment on the initiation of local intervention strategies.

[0063] In this embodiment, based on the net surface resistance index extracted in step S30, to address the issues of chassis slippage and reduced cleaning efficiency caused by insufficient adhesion when the robot encounters stubborn stains or high-resistance areas, the main control module executes step S40, which introduces a multi-parameter electromechanical-hydraulic collaborative intervention strategy based on resistance characteristics feedforward. Step S40 executes the following specific sub-steps:

[0064] S41. Local Area Resistance Anomaly Detection and Path Replanning. The main control module reads the preset resistance threshold from the internal non-volatile memory. As a preferred method, the resistance threshold represents the upper limit reference value of physical friction resistance of the robot chassis when running on a standard clean and flat surface. It is calculated by adding three times the standard deviation to the statistical average of the net surface resistance index extracted during the global traversal of the robot in the standard test water tank during the factory calibration phase.

[0065] Considering that data fluctuations at a single moment may lead to false triggers, the main control module constructs a resistance assessment window over multiple consecutive sampling periods. The main control module not only determines whether the real-time extracted net surface resistance index is greater than the resistance threshold, but also calculates the variance of the net surface resistance index fluctuation within the resistance assessment window. Only when the net surface resistance index is greater than the resistance threshold for multiple consecutive periods, and the variance of the fluctuation is less than the set steady-state variance threshold, does the main control module determine that the local area currently contacted by the robot has a stable presence of stubborn attachments or is in a high-friction coefficient state, thereby triggering a local intervention strategy. The steady-state variance threshold is determined by the average variance calculated during the robot's uniform steady-state operation in a level test pool during the factory calibration phase.

[0066] To ensure the recoverability of breakpoints in the global traversal path planning, the main control module pushes the current spatial coordinate data, absolute yaw angle, and global path expansion progress pointer into the internal stack memory for breakpoint protection the instant the local intervention strategy is triggered. For the data pushing and pointer management operations in the memory stack, those skilled in the art can use standard data structure push instructions; the underlying memory addressing and saving mechanism is a well-known technology in the field and will not be elaborated upon here.

[0067] After completing breakpoint protection, the main control module suspends the current global traversal expansion, resets the path generator parameters, and plans to generate orthogonal coverage path instructions around the current coordinate point. Specifically, the orthogonal coverage path instructions control the robot to perform reciprocating linear movements that interweave lateral and longitudinal directions within the local grid boundary, thereby increasing the frequency of physical cleaning of the current dirty area. The main control module sends the orthogonal coverage path instructions to the drive module, which receives the instructions and controls the track motors to perform gridded motion.

[0068] S42. Resistance Feature Feedforward and Dynamic Allocation of Culvert Hydrodynamics. While generating orthogonal coverage path commands, the main control module calculates the difference parameter between the net surface resistance index and the resistance threshold. This difference parameter reflects the overshoot of localized fouling resistance exceeding the standard level. The main control module extracts this difference parameter as the feedforward gain input for the control loop. Based on the feedforward compensation mechanism in automatic control principles, traditional control closed loops based on feedback errors suffer from physical defects such as response lag. Introducing resistance feature feedforward can reduce physical inertia to adjust the actuator parameters.

[0069] To overcome the high resistance caused by dirty areas and prevent track slippage, the main control module dynamically increases the output speed of the ducted motor using the feedforward gain input. The main control module calculates the target control speed of the ducted motor based on the feedforward compensation mathematical model. The calculation formula corresponding to the feedforward compensation mathematical model is as follows: ; in, This indicates the target control speed of the ducted motor after the application of feedforward intervention; This indicates the base cruise speed of the ducted motor when the robot is cruising on a standard clean surface; This represents the pre-calibrated feedforward gain mapping ratio coefficient, which reflects the required ducted motor speed increment corresponding to the unit resistance exponential overshoot. It is obtained by comprehensive calibration based on the difference between the maximum allowable speed of the ducted motor and the basic cruise speed. This represents the dimensionless net surface resistance index after normalization. This indicates the resistance threshold configured in the system. The parameter representing the difference between the net surface drag index and the drag threshold is the feedforward gain input. This represents the speed compensation increment calculated based on the feedforward gain input and the feedforward gain mapping ratio.

[0070] The main control module converts the calculated target rotational speed into a corresponding underlying pulse width modulation signal and sends it to the drive module. The drive module controls the ducted motor to accelerate its rotation. The increased output speed of the ducted motor causes the water inside the robot to be sprayed outward at an accelerated speed. Based on the principle of fluid reaction force, the hydrodynamic pressure of the robot on the working surface at the bottom of the pool increases accordingly, thereby increasing the physical normal pressure between the chassis and the contact surface and the track traction.

[0071] S43. Mechanical Limiting Dynamic Compensation and Electrical System Balance Control. As the output speed of the ducted motor increases and the hydrodynamic pressure increases, the robot is forced to adhere to the bottom surface of the pool. Under high hydrodynamic pressure, the mechanical friction load on the track motor driving the body and the roller brush motor performing rotational brushing increases accordingly. The increase in mechanical load forces the motor rotor speed to be physically suppressed, resulting in a decrease in the back electromotive force generated by the stator winding cutting the magnetic field lines inside the motor. According to the basic electrical principles of motors, the decrease in back electromotive force will cause an increase in the physical current of the main circuit. If the system maintains the current limiting threshold configured under normal conditions, the increased main circuit current will touch the limiting boundary, causing the main control module to misjudge a short circuit or stall anomaly, thereby triggering hardware overcurrent protection or even cutting off the drive circuit.

[0072] To maintain the dynamic balance between the electrical and hydrodynamic systems, the main control module performs dynamic limiting compensation calculations based on the increase in hydrodynamic pressure resulting from the increased speed of the ducted motor. As a preferred approach, the main control module constructs electromechanical-hydraulic conversion equations based on the chassis traction physical model and motor characteristics, and calculates the current compensation boundary accordingly. The calculation formulas corresponding to the electromechanical-hydraulic conversion equations are as follows: ; in, This represents the calculated current compensation boundary. This represents the preset safety compensation coefficient, which is set to a constant greater than one. It is obtained through analysis of high-load bench anti-lockdown tests and is used to prevent false triggering due to transient current surges during high-load startup. This indicates the increase in hydrodynamic pressure caused by the increase in the speed of the ducted motor; This represents the static friction coefficient between the chassis tracks and the working surface; Indicates the physical radius of the driving track wheel; This indicates the additional load torque generated at the chassis wheel ends; Indicates the physical reduction ratio of the drive transmission mechanism; This represents the fixed motor torque constant of the bottom track motor; It represents the comprehensive transmission constant that converts the electrical current of the motor into the mechanical torque at the wheel end; This represents the theoretical current compensation reference value equivalent to the electrical circuit. By dividing the added load torque by the comprehensive transmission constant and introducing a safety compensation coefficient, the main control module achieves compensation calculations across physical domains.

[0073] The main control module superimposes the current compensation boundary onto the initial current limiting thresholds configured for the track motor and the roller brush motor, performing a dynamic upward adjustment. By dynamically increasing the current limiting thresholds for the track motor and the roller brush motor, the main control module widens the allowable operating range of the underlying drive circuit, preventing the drive link from shutting down due to hardware overcurrent protection when subjected to heavy-load cleaning tasks, thereby ensuring the stable operation of high-load electromechanical-hydraulic collaborative operations in local areas. The writing and activation mechanism of the underlying chip current limiting register can be implemented using general driver chip configuration instructions by those skilled in the art; its register rewriting mechanism is a well-known technology in the field and will not be elaborated upon here.

[0074] See attached document Figure 4 In this embodiment, in conjunction with the resistance-feedforward electromechanical-hydraulic multi-parameter collaborative intervention strategy executed in step S40, to ensure that stubborn local dirt is effectively removed and to guide the robot smoothly back to the original global traversal task after the local intervention, preventing cleaning blind spots or repeated coverage, the main control module executes step S50 of cleaning status verification and global state machine recovery sequence. Step S50 executes the following specific sub-steps:

[0075] S51. Real-time monitoring and hysteresis verification logic for cleaning efficiency. During the execution of the orthogonal coverage path command for localized meshing by the drive module controlling the robot, the main control module continuously and synchronously calls the decoupling algorithm in step S30 to extract the real-time updated net surface resistance index. To accurately determine whether stubborn deposits in a local area have been removed by the physical brush and suctioned by the fluid system, the main control module opens a verification time window in the system memory. The length of the verification time window is set according to the physical time required for the robot to complete a single complete mesh cell traversal.

[0076] The main control module collects the net surface resistance index for multiple consecutive sampling periods within the verification time window and introduces a hysteresis verification model for state assessment. Based on the hysteresis comparison principle in control system engineering, the hysteresis verification model aims to prevent frequent erroneous state switching near the critical resistance point. The calculation formula for the hysteresis verification model is as follows: ; in, This represents the calculated cleanliness verification deviation. This represents the total number of sampling periods contained within the verification time window. The value is a positive integer, and its value is obtained by dividing the length and width of the local grid by the linear velocity of the chassis foundation. Indicates the first Net surface resistance index extracted from each sampling period; This indicates that all net surface resistance indices within the verification time window are summed. Represents the average coefficient for discrete samples; This represents the mathematical discrete average of the net surface resistance index within the verification time window; This indicates the resistance threshold read by the system in step S41; This represents the pre-configured hysteresis convergence coefficient, whose value range is set to a constant within the interval (0.8, 0.95). It is used to establish a tolerance lower boundary that allows for minor residual friction on the original resistance determination boundary. Its specific value is determined by technicians through repeated testing and comparison by changing the roughness of the friction surface in the water tank calibration experiment. This represents the target cleanliness verification benchmark after hysteresis convergence processing.

[0077] Before performing discrete sample averaging, the main control module detects the total number of sampling periods. The value. If the extracted effective net surface resistance index is missing due to packet loss in sensor bus communication, resulting in a decrease in the total number of sampling periods. If the result is zero, the main control module triggers the division-by-zero anomaly protection mechanism, suspends the current delayed verification logic and maintains the electromechanical-hydraulic collaborative intervention state until sufficient valid sample data is acquired again.

[0078] To avoid misjudgments based on a single average value, the main control module simultaneously calculates the variance of the fluctuation of all net surface resistance indices within the verification time window while calculating the cleanliness verification deviation. The main control module then determines whether the cleanliness verification deviation is less than or equal to zero. If the cleanliness verification deviation is greater than zero or the calculated variance of fluctuation is greater than the preset discrete distribution boundary value, the main control module determines that there is still high frictional resistance or dirt adhesion in the current local area, maintains the current electromechanical-hydraulic multi-parameter collaborative intervention strategy, and continues to execute the orthogonal coverage path command.

[0079] If the cleanliness verification deviation is less than or equal to zero within a single verification time window and the variance fluctuation is within the discrete distribution boundary value, but the condition for two consecutive windows is not met, the main control module maintains the current monitoring state and continues to execute the orthogonal coverage path instruction. If the cleanliness verification deviation is less than or equal to zero for two consecutive verification time windows and the variance fluctuation is within the discrete distribution boundary value, the main control module determines that the local area has been effectively cleaned, thereby triggering the system to enter the parameter degradation and state recovery process. The discrete distribution boundary value is determined by adding the average variance calculated during the robot's uniform and steady-state operation in the flat test area during the factory calibration phase to the preset data fluctuation tolerance.

[0080] If the cleanliness verification deviation is less than or equal to zero for two consecutive verification time windows and the fluctuation variance is within the discrete distribution boundary value, the main control module determines that the local area has been effectively cleaned, and then triggers the system to enter the parameter degradation and state recovery process.

[0081] S52. Smooth degradation of electromechanical and hydraulic parameters and reset of the underlying drive. After confirming the completion of local cleaning and triggering the recovery process, because the chassis and working surface are still under high hydrodynamic pressure and the underlying motor is operating at a wide limit boundary, based on fluid inertia and pipeline hydrodynamics principles, if all compensation interventions are instantly withdrawn, it can easily cause a sudden drop in negative pressure and water hammer effect within the fluid system. Sudden changes in water pressure will not only cause the recently sucked suspended debris to overflow again, but will also generate a large reverse electromotive force impact on the drive module.

[0082] The main control module suspends real-time calculation of the feedforward gain input and activates the ramp degradation function. As a preferred method, the main control module gradually reduces the target control speed of the ducted motor according to a set step-decreasing time constant, causing it to return to the base cruise speed with a preset slope. The step-decreasing time constant is calculated jointly from the rotor inertia of the ducted motor and the damping coefficient of the water fluid to ensure a smooth transition during the deceleration process.

[0083] During the deceleration of the ducted motor, the main control module synchronously reads the current actual feedback speed of the ducted motor and, based on the corresponding decrease in hydrodynamic pressure attenuation at the actual feedback speed, gradually reduces the current compensation boundaries of the track motor and the brush motor. Once the ducted motor returns to its base cruising speed, the main control module restores the current limiting thresholds of the track motor and the brush motor to their initial configuration parameters under normal conditions. For the linear decreasing control of the pulse width modulation signal duty cycle and the smooth transition of the drive parameters, those skilled in the art can use a general ramp generator algorithm or a first-order inertial filter algorithm. The underlying software smoothing control mechanism is well-known in the field and will not be elaborated upon here.

[0084] S53. Global Breakpoint Pop and Navigation State Machine Trajectory Recovery. After completing the steady-state reset of the electromechanical and hydraulic underlying physical parameters, the main control module accesses the internal stack memory and performs a data pop operation. The main control module extracts the spatial coordinate data, absolute yaw angle, and global path extension progress pointer saved at the moment the local intervention strategy was triggered in step S41 from the stack memory.

[0085] To ensure spatial consistency in tracking calculations, the main control module performs matrix alignment verification between the extracted original breakpoint spatial coordinate data and the current global odometry reference frame. After successful verification, the main control module acquires the robot's current spatial coordinate data and calculates the relative distance and relative target heading angle between the current spatial coordinate data and the extracted original breakpoint spatial coordinate data. Subsequently, based on the chassis kinematic differential model, the difference in yaw angle between the relative target heading angle and the real-time yaw angle fed back by the current inertial measurement unit is used as the error input for the proportional-integral feedback loop.

[0086] Combining the error input term corresponding to the yaw angle difference, the main control module calculates the target control linear velocity of the left and right tracks at the bottom layer based on the discretized proportional-integral control law and the chassis differential speed distribution model. The calculation formula corresponding to the discretized proportional-integral differential control model is as follows: ; ; ; in, This represents the differential compensation adjustment amount calculated by the proportional-integral feedback loop; This represents the proportional gain coefficient for yaw closed-loop control. This represents the calculated relative heading angle to the target. This represents the real-time yaw angle fed back by the inertial measurement unit at the current moment; This represents the yaw angle difference at the current moment, i.e., the error input of the proportional control term; Indicates the proportional adjustment speed component; This represents the integral gain coefficient of the yaw closed-loop control; This indicates the current sampling period number after entering the breakpoint recovery process; Indicates the first Real-time yaw angle feedback per cycle; This represents the discrete integral sum of all historical yaw angle differences since entering the regression discontinuity state; Indicates the integral speed component; This represents the calculated target control linear velocity of the left track motor; This represents the calculated target control linear velocity of the right track motor; This represents the basic straight-line velocity set by the robot during its approach to the breakpoint; This indicates the left-side linear speed command obtained by subtracting the differential compensation adjustment from the base straight-line speed; This indicates the right-side linear speed command obtained by superimposing differential compensation adjustment on the base straight-line speed.

[0087] The main control module converts the target control linear velocity into a pulse distribution sequence for the motors on both sides of the chassis, thereby controlling the track motors to perform differential steering until the robot's front end aligns with the original breakpoint. Subsequently, it controls the robot to travel in a straight line to approach the breakpoint. The specific calculations of the proportional-integral feedback loop and the differential command conversion can be implemented using a general PID motion control algorithm by those skilled in the art. Its underlying closed-loop servo control mechanism is a well-known technology in the field and will not be elaborated upon here.

[0088] When the positioning error between the robot's current spatial coordinate data and the original breakpoint spatial coordinate data is less than the set spatial convergence tolerance, the main control module determines that the robot has physically returned to the breakpoint. The spatial convergence tolerance is set based on a combination of the robot's positioning module's static drift variance and the chassis's maximum external dimensions, and is typically one-tenth of the chassis's physical width.

[0089] After reverting to the breakpoint, the main control module adjusts the robot's yaw angle to the absolute yaw angle extracted from the stack and restores the interrupted internal state of the path generator based on the global path expansion progress pointer. Based on the reset parameter state, the main control module reactivates the corresponding irregular surface contour tracking control logic or flat terrain yaw angle locking traversal mode according to the current terrain environment information determined in step S20, and continues to issue the unfinished global motion trajectory instructions, thereby ensuring that the robot can complete the cleaning traversal task of the entire underwater environment with an effective planned coverage area.

[0090] See attached document Figure 5 To be continued Figure 7To aid in understanding the technical solution of this invention, a specific swimming pool cleaning application embodiment is provided. This embodiment is tested in a test pool containing a flat bottom and sloping sidewalls. Combining the aforementioned steps and logic, the main control module performs the following operations to complete environmental perception, resistance decoupling, and electromechanical coordinated intervention.

[0091] The robot's system clock frequency is set to 100Hz, and the global sampling time step within the system is set to... The configuration is 0.01 seconds. After the robot enters the water and lands on the bottom, it initially positions itself on the flat bottom of the pool. The main control module triggers the synchronous reading of multi-source basic environmental parameters via an internal hardware timer interrupt. The physical settings are: the sampling frequency of the pressure sensor is 10Hz, and the sampling frequency of the inertial measurement unit is 100Hz. The main control module uses the high-frequency sampling clock of the inertial measurement unit as the time cross-section reference and performs zero-order hold interpolation sampling on the data from the low-frequency pressure sensor, mapping the discrete pressure data to the 100Hz time domain, thus achieving time-domain alignment of the data.

[0092] The main control module synchronously extracts the hydrostatic pressure including dynamic pressure disturbance, and uses the spatial Euler angle matrix transformation algorithm to rotate the coordinate system of the spatial triaxial linear acceleration to remove the gravitational component and extract the vertical acceleration.

[0093] The preset non-volatile memory contains a pool fluid density of 1000 kg / m³. 3 gravitational acceleration It is 9.81 m / s 2 The main control module calls the underlying divider instruction to divide the pressure difference between the hydrostatic pressure and the reference atmospheric pressure by the product of the fluid density and the acceleration due to gravity, thus obtaining the original roughness depth. The main control module multiplies the vertical acceleration of the current cycle by the global sampling time step of 0.01 seconds, and adds it to the result of the previous cycle to calculate the vertical relative velocity of the current cycle. .

[0094] The main control module calculates instantaneous depth based on a discrete-time domain complementary filtering model. During operation, the main control module evaluates the recent hydrostatic pressure fluctuation variance and dynamically allocates filter confidence coefficients based on the variance results. The value of is used to calculate the instantaneous depth of the output. The data is appended with a clock timestamp and written into a circular buffer configured in system memory to construct a continuous sequence of depth data.

[0095] The main control module allocates a sliding sampling window in its internal memory space. The fixed data capacity of the sliding sampling window is configured to 200 sampling points, covering a 2-second physical movement cycle of the robot. During the runtime of 0 to 30 seconds, the robot maintains its position on a flat bottom surface. The main control module extracts the depth data sequence and calculates the local depth variance. At this stage, the calculated local depth variance stabilizes at 0.001m. 2 Fluctuations within a small range. The system's pre-calibrated flatness feature threshold is 0.02m. 2 If the local depth variance remains below the flatness characteristic threshold, the main control module activates its internal sliding timer counter to confirm the anti-jitter state delay. Once the state is confirmed to meet the requirements, the robot is determined to be in a flat terrain environment. The main control module executes a state machine switch to trigger the navigation dimensionality reduction matrix and activates the yaw angle locking traversal mode. The main control module performs integral calculations on the spatial three-axis angular velocities to obtain the yaw angle, which is used as the heading feedback reference to generate staggered motion trajectory commands, which are then sent to the drive module to control the track motors to perform constant-speed chassis movement.

[0096] At time 30 seconds, the robot crossed the boundary and entered a region with a sloping sidewall. The local depth variance calculated by the main control module increased and exceeded the set flatness characteristic threshold of 0.02m. 2 The main control module determines that it has entered an irregular curved terrain environment and activates the contour line tracking control logic. During contour line tracking, the main control module builds a pseudo-3D topological map in the background system memory, which is constructed by combining two-dimensional occupancy grids and depth and elevation information.

[0097] The current sampling circuit reads the main circuit current signals of the track motor and the brush motor in real time. The main control module discretizes the signals using an analog-to-digital converter and uses a first-order hysteresis low-pass filter with a cutoff frequency of 10Hz to filter out commutation spike noise and extract the steady-state total load current. The main control module introduces a first-order polynomial interpolation function based on the slope of the preceding and following sampling points to perform pitch angle interpolation. With running linear velocity Implement time-domain aligned resampling.

[0098] After alignment, the main control module uses the reference current estimation model under theoretical pose conditions for calculation. (Foundation no-load constant) Based on the factory-set suspension calibration of 1.5A, the robot's equivalent physical mass underwater is... The load is set to 2.0 kg. The main control module subtracts the calculated reference current from the steady-state total load current to obtain the net friction load current. It then calls the underlying division boundary detection logic and verifies the preset calibration current. After configuring it to 0.5A, a division operation is performed to obtain the dimensionless net surface resistance index. During the 30 to 60-second run, the extracted net surface resistance index remained stable below the system resistance threshold of 2.0 because the baseline current model completely decoupled the slope gravity load from the water flow damping.

[0099] At 60 seconds into the runtime, the robot chassis encountered a dirty working surface with stubborn deposits. Within a resistance evaluation window consisting of multiple consecutive sampling periods, the main control module detected that the net surface resistance index rapidly increased to 3.5, significantly exceeding the preset resistance threshold of 2.0, and the arithmetic variance of the fluctuation was less than the preset steady-state variance threshold. The main control module determined that the physical resistance was abnormal and triggered a local intervention strategy. The main control module invoked a data structure push instruction to store the current spatial coordinates, absolute yaw angle, and global path expansion progress pointer into the internal stack memory, creating a breakpoint protection. The main control module aborted the global traversal operation, reset the parameters, and generated orthogonal coverage path instructions, which were then executed by the driver module.

[0100] Simultaneously, the main control module calculates the difference between the net surface drag index and the drag threshold (the calculation result is 1.5), and uses this as the input to the feedforward gain. This is based on the feedforward compensation mathematical model. System base cruise speed Set to 2500 RPM, feedforward gain mapping ratio. Set at 800 RPM, the main control module calculates that the target control speed of the ducted motor is increased to 3700 RPM. The drive module drives the ducted motor to accelerate, resulting in a significant increase in physical hydrodynamic pressure.

[0101] To prevent false overcurrent alarms caused by increased adhesion, the main control module calculates the theoretical current compensation boundary based on the electromechanical-hydraulic conversion equation and superimposes the compensation boundary onto the original set threshold. The current limit threshold of the track motor and the roller brush motor is dynamically increased from the conventional 3.0A and locked to 4.5A to ensure sufficient output torque of the actuator motor.

[0102] During the localized orthogonal coverage operation, the main control module allocates a 5-second verification time window. The main control module introduces a hysteresis verification model to determine the physical cleanliness status and calculates the cleanliness verification deviation. The resistance threshold is configured to 2.0, and the hysteresis convergence coefficient is... Setting the threshold to 0.9, the target cleanliness verification benchmark was determined to be 1.8. After 80 seconds of operation, localized contaminants were completely physically removed. The main control module monitored that within two consecutive verification time windows, the discrete average of the extracted net surface resistance index fell back to 1.5, the cleanliness verification deviation was calculated as negative, and the resistance fluctuation variance fell back within the discrete distribution boundary. The main control module determined that the contaminants were completely removed.

[0103] The system activates a first-order linear decay iterative equation with a time constant, suspending the feedforward gain input. The duct motor speed smoothly decreases to the base cruise speed of 2500 RPM according to the step-decreasing time constant. With the elimination of physical hydrodynamic pressure, the main control module uses the drive chip configuration instructions to inversely reduce the current limiting threshold of the track motor, smoothly restoring it to the initial parameter of 3.0A.

[0104] After the underlying electromechanical parameters are reset, the main control module accesses the stack memory to perform a pop operation, extracting the coordinates of the protected breakpoint and the heading parameters. The main control module substitutes the current spatial coordinates and the breakpoint coordinates into the matrix for verification, and calculates the differential compensation command based on the discrete proportional-integral control law, sending it to the chassis motor. The drive module controls the chassis to steer and move towards the breakpoint. When the relative spatial positioning error converges to within a tolerance of 0.05 meters, the physical regression breakpoint is determined. Subsequently, the main control module calls the global path progress pointer to restore the trajectory generator's operating environment, guiding the robot to continue completing the global cleaning along the original grid.

[0105] Figure 5 The light gray solid line represents the original roughness depth, the black solid line represents the instantaneous depth after complementary filtering, the pure black dotted line represents the local depth variance, and the dark gray dashed line represents the smoothness feature threshold. Combined with... Figure 5 It can be seen that in the flat terrain range of 0 to 30 seconds, the local depth variance value remains below the flatness feature threshold, and the system maintains a locked yaw angle state. In the terrain climbing range after 30 seconds, the local depth variance increases sharply and exceeds the flatness feature threshold boundary, triggering the system's irregular surface determination. At the same time, the instantaneous depth curve shows that the complementary fusion filter has a strong smoothing ability for high-frequency hydrodynamic disturbances.

[0106] Figure 6 The solid black line represents the real-time extracted net surface resistance index, the dark gray dashed line represents the resistance threshold, and the pure black dotted line represents the target cleanliness confirmation benchmark. (See attached diagram.) Figure 6 It can be seen that at 60s, the net surface resistance index rises sharply and exceeds the resistance threshold, indicating contact with a high-resistance dirty area; at 80s, the resistance drops and continuously crosses the target cleaning confirmation benchmark, indicating that the area has been cleaned.

[0107] Figure 7 The solid black line represents the target control intervention speed, the gray dashed line represents the basic cruise setting speed, and the solid black dotted line represents the limiting threshold of the track and brush motors. As can be seen from the curves, upon receiving the feedforward input, the main control module forces the ducted motor to exceed the basic cruise setting speed, rapidly increasing and maintaining it to enhance the hydrodynamic downforce. Simultaneously, to maintain electromechanical balance across the physical domain, the system dynamically pushes up the limiting thresholds of the track and brush motors during intervention, thus preventing hardware overcurrent malfunctions under heavy-load, high-pressure scrubbing conditions, until the 80-second cleaning is completed and the system enters the ramp degrading buffer phase.

[0108] In the verification of the embodiments of the technical solution of the present invention, combined with Figures 5 to 7 The recorded multi-dimensional operational data provided a comprehensive evaluation of the underlying algorithm's innovative performance. The complementary fusion filtering model completely resolved the amplification defect of dynamic pressure disturbances caused by the traditional architecture's absolute pressure calibration. Combined with... Figure 5 The dual-axis observation data maintained convergence and tracking of instantaneous depth in the climbing data segment. The local depth variance parameter constructed in this way successfully eliminated misjudgments caused by fluid surge and ensured the rigor of state machine switching.

[0109] in accordance with Figure 6 and Figure 7 The response relationship observed during the 60-80 second intervention period confirmed that the underlying decoupling model independently extracted the physical friction loads on the environmental surface, and the cross-physical domain feedforward intervention strategy bridged the boundaries of independent operation of the mechanical, electrical, and hydraulic systems. After the net resistance deviation was injected as feedforward gain, the ducted motor speed directly generated a significant hydrodynamic downforce, while the main control hardware simultaneously expanded the safety limit window of the actuator motor. Ultimately, the entire intervention and regression process maintained high steady-state characteristics in the multi-system coupling balance, comprehensively preventing negative pressure drop impacts and achieving optimal dynamic scheduling and configuration of mechanical, electrical, and hydraulic resources under complex terrain and unsteady load conditions.

Claims

1. A method for generating a cleaning path for a swimming pool robot, characterized in that, Includes the following steps: Read multidimensional basic environmental parameters, extract instantaneous depth and depth change rate, and construct depth data sequence; The local depth variance is extracted from the depth data sequence. Based on the comparison between the local depth variance and the flatness feature threshold, the contour tracking control logic or yaw angle locking traversal mode that introduces the depth change rate is activated, and a motion trajectory command is generated to execute the global traversal cleaning operation. During the global traversal cleaning operation, a reference current is acquired, and the net surface resistance index is extracted using the difference between the real-time total load current and the reference current. When the net surface resistance index is greater than the resistance threshold, a local intervention strategy is triggered to stop the global traversal cleaning operation and generate an orthogonal coverage path instruction. The output speed of the duct motor is dynamically increased according to the net surface resistance index, and the current limit threshold of the track motor and the roller brush motor is increased. When the net surface resistance index is not greater than the resistance threshold, the output speed of the duct motor is reduced and the current limiting threshold of the track motor and the roller brush motor is reset. Based on the local depth variance, the contour tracking control logic or the yaw angle locking traversal mode is called again to continue the global traversal cleaning operation.

2. The method for generating a cleaning path for a swimming pool robot according to claim 1, characterized in that, The process of reading the multidimensional basic environmental parameters, extracting the instantaneous depth and the rate of change of depth, and constructing the depth data sequence includes: Simultaneously extract hydrostatic pressure and spatial triaxial linear acceleration to obtain the vertical acceleration after removing the gravitational component; Calculate the pressure difference between the hydrostatic pressure and the reference atmospheric pressure, and divide the pressure difference by the product of the pool fluid density and the gravitational acceleration to obtain the original roughness depth; The vertical acceleration of the current period is multiplied by the global sampling time step and accumulated with the vertical relative velocity of the previous sampling period to calculate the vertical relative velocity of the current period. Obtain the recent hydrostatic pressure fluctuation variance for evaluation, and dynamically allocate filter confidence coefficients based on the comparison results between the recent hydrostatic pressure fluctuation variance and the benchmark hydrostatic pressure variance threshold. The instantaneous depth extracted at the current sampling time is calculated using the filtered confidence coefficient, the instantaneous depth output at the previous sampling time, the product of the vertical relative velocity and the global sampling time step, and the original roughness depth. The depth difference result is obtained by subtracting the instantaneous depth extracted at the current sampling time from the instantaneous depth extracted at the previous sampling time, and the depth difference result is obtained by dividing the global sampling time step. The instantaneous depth and the corresponding timestamp are synchronously cached to construct the depth data sequence.

3. The method for generating a cleaning path for a swimming pool robot according to claim 2, characterized in that, The process of extracting the local depth variance from the depth data sequence includes: A sliding sampling window with a fixed data capacity is allocated in the internal memory space. The timestamp difference of the timestamps of adjacent extracted instantaneous depths is checked and filled into the sliding sampling window to form a depth evaluation sequence. The arithmetic mean of all the instantaneous depths in the depth evaluation sequence is calculated. The squared deviation of the relative height between the instantaneous depth and the arithmetic mean from the first discrete sampling point to the last discrete sampling point is calculated and summed. The result of the summation is multiplied by the discrete sample averaging coefficient to obtain the local depth variance.

4. The method for generating a cleaning path for a swimming pool robot according to claim 1, characterized in that, The process of activating the contour tracking control logic that introduces the depth change rate includes: The depth change rate is introduced into the motion control closed loop, the error amount of the depth change rate deviating from zero is calculated, and a differential compensation command is output when the error amount exceeds the set error dead zone range. The differential compensation command after boundary truncation is superimposed on the basic speed parameters to generate differential torque, and the motion heading is locked on the gravity equipotential surface to perform a spiral approximation action. During the execution of the spiral approximation action, a meshed pseudo-3D topological map of equal depth is established by combining the movement distance and the instantaneous depth.

5. The method for generating a cleaning path for a swimming pool robot according to claim 1, characterized in that, The process of activating the yaw angle locking traversal mode includes: A state machine switching action is triggered when the continuously calculated local depth variance is not greater than the smoothness feature threshold within the anti-jitter time window. The differential compensation command update in the contour tracking control logic is suspended, and the yaw angle is obtained by integral calculation using the collected spatial three-axis angular velocities. The yaw angle is used as the absolute heading feedback reference of the motion control closed loop, combined with the preset planar coverage step size parameter to generate the motion trajectory command in the form of straight line interlacing or planar spiral.

6. The method for generating a cleaning path for a swimming pool robot according to claim 2, characterized in that, The process of obtaining the reference current and extracting the net surface resistance index using the difference between the real-time total load current and the reference current includes: The product of the underwater equivalent physical mass parameters and the sinusoidal trigonometric function values ​​of the gravitational acceleration and pitch angle is used to obtain the gravitational component load along the running slope direction. The gravitational component load is then multiplied by the linear mapping coefficient from the gravitational load to the electrical current to obtain the equivalent electrical load. The electrical load generated to overcome the fluid resistance of the water body is obtained by multiplying the hydrodynamic damping mapping coefficient by the square of the running linear velocity. The reference current under theoretical pose conditions is obtained by summing the unloaded foundation current constant when suspended underwater, the equivalent electrical load, and the electrical load generated to overcome the resistance of the water fluid. The net surface resistance index is extracted by performing normalization processing on the difference between the real-time total load current and the reference current.

7. The method for generating a cleaning path for a swimming pool robot according to claim 1, characterized in that, The process of dynamically increasing the output speed of the ducted motor and raising the current limiting threshold of the track motor and the brush motor based on the net surface resistance index includes: The difference between the net surface resistance index and the resistance threshold is extracted as the feedforward gain input. The target control speed is obtained by adding the base cruise speed to the speed compensation increment, and the duct motor is controlled to run at the target control speed. The speed compensation increment is obtained by multiplying the feedforward gain input and the feedforward gain mapping ratio coefficient. The additional load torque is calculated by multiplying the increase in hydrodynamic pressure caused by the increase in the output speed of the ducted motor by the static friction coefficient between the chassis track and the working surface and the physical radius of the drive track wheel. Calculate the comprehensive transmission constant, which is obtained by multiplying the physical reduction ratio by the fixed motor torque constant; The current compensation boundary is obtained by dividing the new load torque by the comprehensive transmission constant and multiplying it by the safety compensation coefficient. The current compensation boundary is then superimposed on the current limiting threshold initially configured for the track motor and the roller brush motor.

8. The method for generating a cleaning path for a swimming pool robot according to claim 1, characterized in that, The process of reducing the output speed of the ducted motor and resetting the current limiting threshold of the track motor and the brush motor includes: Calculate the discrete average value of the net surface resistance index within the verification time window, calculate the product of the resistance threshold and the hysteresis convergence coefficient to obtain the target cleanliness verification benchmark, and subtract the target cleanliness verification benchmark from the discrete average value to obtain the cleanliness verification deviation. While calculating the cleanliness verification deviation, the resistance fluctuation variance of all net surface resistance indices within the verification time window is also calculated. When the cleanliness verification deviation is not greater than zero for two consecutive verification time windows and the resistance fluctuation variance is within the discrete distribution boundary value, the local area is determined to have been effectively cleaned. After determining that the local area has been effectively cleaned, the target control speed of the duct motor is reduced back to the basic cruise speed according to the set step-decreasing time constant; The actual feedback speed of the duct motor is read synchronously, and the current compensation boundary of the track motor and the roller brush motor is gradually reduced according to the hydrodynamic pressure attenuation corresponding to the actual feedback speed. When the ducted motor returns to the base cruise speed, the current limiting threshold is restored to the initial configuration parameters.

9. A method for generating a cleaning path for a swimming pool robot according to claim 1, characterized in that, The process of completing the global traversal cleaning operation includes: Read the original breakpoint spatial coordinate data recorded when the global traversal cleaning operation is stopped, and perform matrix alignment verification with the current global odometer reference frame; Calculate the yaw angle difference between the relative target heading angle and the real-time yaw angle, and multiply the yaw angle difference by the proportional gain coefficient to obtain the proportional speed component; Calculate the discrete integral sum of all historically generated yaw angle differences, and multiply the discrete integral sum by the integral gain coefficient to obtain the integral controlled speed component. The differential compensation adjustment amount is obtained by adding the proportional speed adjustment component and the integral speed adjustment component. The left or right linear speed command is obtained by subtracting or adding the differential speed compensation adjustment to the basic straight-line speed. The track motor of the pool robot is driven to run according to the left or right linear speed command to continue to complete the global traversal cleaning operation.

10. A pool robot cleaning path generation and control system, characterized in that, The method for generating a cleaning path for a swimming pool robot as described in any one of claims 1-9 includes: The sensing module includes a pressure sensor for collecting hydrostatic pressure, an inertial measurement unit for collecting spatial triaxial linear acceleration and spatial triaxial angular velocity, and a current sampling circuit for collecting steady-state operating current, and transmits environmental measurement signals and electrical signals to the main control module. The main control module is internally configured with an analog-to-digital conversion unit and a hardware floating-point arithmetic unit, and outputs low-level pulse width modulation signals and control commands to the drive module; The drive module includes a track motor configured on both sides of the chassis to control linear movement and differential steering, a roller brush motor to drive the rotating brushing surface, and a duct motor to generate negative pressure and hydrodynamic downforce. The duct motor is equipped with a speed feedback unit to monitor the operating status.

Images