Multi-degree-of-freedom motion control and attitude adaptive adjustment method for underwater booster

Abstract

The invention discloses a multi-degree-of-freedom motion control and attitude adaptive adjustment method for an underwater booster, relates to the technical field of underwater multi-degree-of-freedom attitude control and thrust distribution, and is used for solving the problems of attitude instability and control failure of the underwater booster in a narrow channel, thrust saturation and disturbance environment. Firstly, based on spatial isolation arrangement of a power supply and a sensing assembly, a stable posture in which the gravity center and the buoyancy center are coaxial is constructed; secondly, unlocking a gear shifting passage after water entry judgment, and analyzing a user gear and a posture state to generate a six-degree-of-freedom target; furthermore, a high-precision attitude is estimated by adopting extended Kalman filtering, and thrust offset in a saturation region is corrected through a potential energy gradient driven by a residual error, so that attitude asymptotic elimination is realized. The method has the capabilities of multi-dimensional disturbance identification, energy-phase linkage judgment and dynamic boundary obstacle avoidance, and is suitable for high-precision boosting control in a complex underwater environment.

Description

Technical Field

[0001] This invention relates to the field of underwater multi-degree-of-freedom attitude control and thrust distribution technology, and more specifically, to a method for underwater booster multi-degree-of-freedom motion control and attitude adaptive adjustment. Background Technology

[0002] Existing underwater boosters are mostly used in underwater gliders, autonomous underwater vehicles, and underwater observation platforms. Their control strategies are mostly based on six-degree-of-freedom dynamic modeling and classical drive-by-wire algorithms, such as proportional-integral-derivative (PID) control, sliding mode control (SMC), and robust optimal control (H∞). These methods usually rely on accurate modeling and environmental stability assumptions, making them difficult to adapt to non-ideal operating scenarios such as thrust system constraints, frequent water flow disturbances, or narrow and complex structures.

[0003] In typical confined spaces (such as inside pipes or at the bottom of a ship), the system's attitude response may be sluggish or even unstable due to potential thruster confinement, uneven boundary fluid pressure, or intensified wake vortex interference. In traditional control methods, when thrust output enters the saturation region, the closed-loop control logic often fails to detect compensation failures, resulting in slow attitude drift without triggering a fault response, thus leading to pseudo-steady state and hysteretic runaway control.

[0004] Furthermore, existing methods for identifying disturbance sources mostly remain at the level of statistical residuals or filtering enhancement, making it difficult to elicit structural responses to periodic disturbances such as wake vortices, local vortices, and lateral surges. Especially when the thrusters are saturated, the system has no redundant thrust for standard path tracking, necessitating the introduction of non-target-oriented potential field correction mechanisms to construct adaptive "non-optimal but safe and reachable" paths.

[0005] Furthermore, current solutions generally lack coupled control over the thrust-attitude-medium identification process, often failing to respond robustly to user gear shift requests in complex environments, and are prone to accidental touches, malfunctions, or low attitude stability issues; in sudden imbalance scenarios, they also lack risk warning mechanisms based on joint analysis of inertia and residuals.

[0006] To address the above problems, this invention proposes a solution. Summary of the Invention

[0007] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a method for multi-degree-of-freedom motion control and attitude adaptive adjustment of underwater boosters, so as to solve the problems mentioned in the background art.

[0008] To achieve the above objectives, the present invention provides the following technical solution: In a preferred embodiment, it includes: A symmetrical mass distribution is constructed and attitude convergence verification is completed to establish a natural and stable initial configuration; the thrust gear table is written, the water entry state is identified through dual-channel water-sensitive determination, the shift control is unlocked, and the power output preparation is initiated. The attitude quaternion is obtained and combined with the user gear input. The six-degree-of-freedom motion target is analyzed, decomposed into a three-dimensional thrust vector, and the initial duty cycle is generated by inverse interpolation. The thrust direction and amplitude are corrected through the proportional-integral channel. After detecting thrust saturation and attitude drift, a multi-dimensional state sequence is collected to identify disturbance segments, a residual vector is constructed and a pseudo-force field is generated to guide the remaining thrust to adaptively correct the deviation, and the boundary potential energy and attitude monitoring are integrated.

[0009] In a preferred embodiment, the power supply and sensors are arranged in a non-overlapping, isolated layout in three-dimensional space to record the angular deviation curve of the underwater booster in a still water pool in real time. The reading of the bubble level is used as a verification signal to compare the convergence rate of the angular deviation over time and adjust the attitude of the underwater booster to achieve natural stability.

[0010] In a preferred embodiment, after attitude stabilization is achieved, a thrust level table is established; shifting and resetting are controlled by buttons; thrust output is unlocked only when the underwater booster is powered on and the water entry determination is successful, otherwise it is locked and cleared to zero.

[0011] In a preferred embodiment, the current attitude state of the underwater booster is obtained based on sensor data, and the six-degree-of-freedom moving target is analyzed in combination with user input to generate a three-dimensional thrust distribution; attitude deviations are continuously corrected during the control process.

[0012] In a preferred embodiment, the target thrust is mapped to an initial duty cycle command, the actual thrust is compared with the target thrust to generate an attitude error, and the duty cycle increment is obtained by performing a proportional-integral calculation on the error. The drive command is updated in real time to track the thrust vector in a closed loop.

[0013] In a preferred embodiment, thrust vector, angular velocity information, external lateral velocity estimation results, and attitude drift trajectory are used to construct multidimensional state sequence data, and residual estimates of the desired attitude trajectory are reconstructed and output in real time.

[0014] In a preferred embodiment, a time-frequency domain multi-scale circular coordinate system is constructed based on a multidimensional state sequence to obtain the rotational decoupling gradient tensor. The difference between the current rotational decoupling gradient tensor and the reference tensor established in the still water stable region is calculated and input into the cyclic autocorrelation operator to extract the resonance period difference between the first two main peaks and determine the phase offset index. Energy calculations are performed on the time-series data of the rotational decoupled gradient tensor to obtain the instantaneous energy curve. A higher-order perturbation discrimination pointer is constructed by multiplying the phase offset index with the instantaneous energy. Then, a first-order high-sensitivity mask M1 is generated based on the dynamic changes of the higher-order perturbation discrimination pointer. The mask is then input into a perturbation anomaly discrimination function constructed based on the variance of the multidimensional state sequence slice and the principal direction offset ratio index to determine the perturbation mask vector M. The phase offset index is regarded as a continuous quantity that characterizes the amplitude of the wake vortex-phase jump. The perturbation mask vector M is regarded as a time-domain weight describing the time span of energy anomalies. The product of the two is calculated in a unified time grid. When the product continuously crosses the dynamic threshold band formed by the historical static steady-state mean plus double the deviation, the consistency test of the ratio of the variance to the mean of the phase offset within the segment is performed to determine the isolated anomalous subsequence and generate a unique anomaly mask on the global timeline.

[0015] In a preferred embodiment, the isolated anomalous subsequence is controlled by a mask vector to participate only in the potential function offset correction operation during the residual construction process, thereby suppressing the directional perturbation of the residual estimation and determining the new residual R.

[0016] In a preferred embodiment, a pseudo-force field is established using the new residual gradient. The remaining thrust margin is distributed to each propulsion channel in decreasing order along the equipotential surface. The main gradient of the pseudo-force field is weighted using a unique anomaly mask, and the instantaneous amplitude near the disturbance source is weakened. Then, the minimum safe distance obtained from real-time ranging and position estimation is converted into barrier potential energy and superimposed on the pseudo-force field, so that the synthetic gradient automatically deviates from the restricted area when approaching the wall or narrow angle.

[0017] In a preferred embodiment, attitude and depth safety monitoring vectors are constructed and mapped to a multi-level envelope consisting of a safety zone, a critical zone, and a limit zone. When entering the critical zone, the thrust level is reduced, and when exceeding the limit zone, an alarm is triggered and a forced downshift is implemented. A dynamic buffer zone is also established to replace reactive fault handling with preventative attitude convergence.

[0018] The technical effects and advantages of the multi-degree-of-freedom motion control and attitude adaptive adjustment method for underwater boosters of this invention are as follows: This invention achieves natural attitude stabilization through mass reconstruction, enabling the system to have a self-stabilizing foundation even before active electrical control. The introduction of water-sensitive triggering and dynamic latching logic effectively avoids accidental activation in mid-air, ensuring operational safety before and after water entry. The six-degree-of-freedom target is constructed jointly through quaternion calculation and user gear settings, and EKF filtering is used to maintain high-precision attitude estimation, providing a stable control benchmark under any transition state.

[0019] In the saturation region where traditional control fails, this invention no longer forcibly tracks the original target attitude. Instead, it extracts the residual structure through an autoencoder and establishes a self-guided correction mechanism by combining it with pseudo-force field guidance. This method can effectively suppress slow attitude drift caused by thrust saturation and provide a separation response to microscopic deviations caused by wake vortex disturbances.

[0020] Meanwhile, the energy-phase dual-index disturbance identification and anomaly mask construction method is adopted to enhance the system's accuracy in identifying periodic disturbances, and the navigation boundary constraint gradient potential energy is integrated to achieve stable control under dynamic obstacle avoidance conditions. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the structure of the underwater booster multi-degree-of-freedom motion control and attitude adaptive adjustment method of the present invention.

[0022] Figure 2 This is a timing diagram of the underwater booster multi-degree-of-freedom motion control and attitude adaptive adjustment method of the present invention. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some 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.

[0024] Example This invention discloses a method for multi-degree-of-freedom motion control and attitude adaptive adjustment of underwater boosters, such as... Figure 1 As shown, it includes: A symmetrical mass distribution is constructed and attitude convergence verification is completed to establish a natural and stable initial configuration; the thrust gear table is written, the water entry state is identified through dual-channel water-sensitive determination, the shift control is unlocked, and the power output preparation is initiated. The attitude quaternion is obtained and combined with the user gear input. The six-degree-of-freedom motion target is analyzed, decomposed into a three-dimensional thrust vector, and the initial duty cycle is generated by inverse interpolation. The thrust direction and amplitude are corrected through the proportional-integral channel. After detecting thrust saturation and attitude drift, a multi-dimensional state sequence is collected to identify disturbance segments, a residual vector is constructed and a pseudo-force field is generated to guide the remaining thrust to adaptively correct the deviation, and the boundary potential energy and attitude monitoring are integrated.

[0025] First, such as Figure 2 As shown, with the goal of naturally stabilizing the initial attitude of the underwater booster, a mass reconfiguration operation of the internal components is performed.

[0026] Specifically, the power supply is slightly adjusted by Δr1 outward from the center of the body along the longitudinal axis, and at the same time, the inertial measurement component is adjusted by Δr2 towards the symmetric orientation, so that the power supply and the sensor are arranged in isolation without overlapping in three-dimensional space. Among them, this arrangement constrains the non-diagonal elements of the overall mass matrix to be minimized through the three-axis symmetric mass distribution criterion, so that the buoyancy center and the center of gravity tend to be coaxial in the vertical section. Among them, the three-axis symmetric mass distribution criterion refers to using the constrained least squares method to iteratively solve the mass matrix.

[0027] Subsequently, in the static water tank, a small perturbation is applied through a pitch-roll two-way rotatable bracket, and the high-precision pitch angle-roll angle fiber optic gyroscope is used to record the angle offset curve in real time. At the same time, the reading of the bubble level is used as an artificial calibration signal to compare the convergence rate of the angle offset over time. When the two curves converge to ±0.2° within 30 s and the bubble is centered and stationary, it is determined that the layout is qualified; if the two curves do not converge to ±0.2° within 30 s, it is necessary to immediately return to the layout stage and adjust the radial distance of the components until the convergence standard is met.

[0028] After the attitude is naturally stabilized, the PWM duty cycle of the brushless motor is mapped to a discrete thrust gear set {P1, P2,..., Pn}, and a propulsion gear table is written into the firmware for real-time call.

[0029] Next, the user operation is input through a waterproof button: short-term closing triggers the step-by-step increasing shift function, and sequentially calls the next duty cycle after the previous gear; long-term closing triggers the zero reset function, forcibly sets the duty cycle to zero and records a safe shutdown event.

[0030] To prevent the propeller from rotating at high speed due to accidental touch in the air, the underwater booster automatically enters the thrust lock state after being powered on. When the embedded controller starts, it first samples the medium impedance in real time through the dual-redundancy contact type water-sensitive switch and the external micro-conductivity probe of the housing. When the determination function g(zprobe, t) outputs that the water entry is established and remains stable, the PWM counter is unlocked to allow the user to perform a shift operation; if the water entry signal is interrupted, the duty cycle is immediately cleared and the drive is locked again.

[0031] Furthermore, on the basis of the initial static balance, the three-axis angular velocity, the three-axis linear acceleration, and the three-axis geomagnetic vector are continuously collected. And in each sampling period, the current angular velocity is inserted into the quaternion differential equation and the first-order explicit integration is performed to obtain the attitude increment, which is then left-multiplied by the quaternion of the previous period to generate the predicted quaternion; Subsequently, the predicted quaternion, the gyro zero bias, the acceleration measurement deviation, and the magnetic deviation are jointly used to form the extended Kalman filter state vector, and the state prior propagation is completed in the "prediction stage"; At the same time, the filter then introduces the latest acceleration and geomagnetic observation values, performs the "observation correction stage" to calculate the Kalman gain and correct the prior quaternion, and represents the current attitude state in the form of a quaternion.

[0032] After obtaining high-precision attitude, the user's current gear input and attitude information are jointly analyzed into a six-degree-of-freedom motion target. Specifically, the controller reads the user's gear and finds the corresponding thrust amplitude in the power-thrust characteristic curve; then, it constructs a volume coordinate-inertial coordinate rotation matrix using instantaneous quaternions, mapping the preset thrust direction vector in the volume coordinate system to the linear velocity target in the inertial coordinate system; if the inertial measurement unit reports a non-zero angular velocity, the angular velocity direction component is analyzed in the same matrix and multiplied by the thrust amplitude to generate the angular velocity target vector.

[0033] Furthermore, based on the calculated linear velocity and angular velocity target vectors, the six-degree-of-freedom target vector is split along the axis, and the three-dimensional thrust distribution vector is calculated. Specifically, the target linear velocity vector is decomposed into longitudinal, lateral, and normal components. Combined with the target angular velocity vector, pitch, roll, and yaw requirements are superimposed onto the corresponding component weights through small-angle approximations. Then, the thrust direction in volume coordinates is calculated using the current quaternion to obtain the three-dimensional thrust distribution vector, which is then written into the drive duty cycle interpolation function to output the single-channel thrust.

[0034] Meanwhile, to avoid thrust offset or target resolution degradation when directions intersect, the thrust distribution is corrected in real time using attitude feedback error. Within each refresh cycle, the controller represents the attitude deviation as the axis-angle error between the target quaternion and the filtered output quaternion. If an axis deviates beyond its limit, the thrust component of that axis is reduced accordingly, and the three-dimensional thrust vector is replanned in the next cycle. This achieves spatial multi-degree-of-freedom linkage response under arbitrary attitude and thrust combinations, ensuring overall control decoupling and consistency.

[0035] Based on the obtained target thrust and target direction vectors, the controller first calls the motor thrust-PWM inverse interpolation function to directly map the target thrust amplitude Fd to the initial duty cycle command PWM0. The inverse interpolation function calculates the thrust in a piecewise, linear manner based on the pre-calibrated static thrust curve of the motor. The duty cycle resolution matches the thrust calibration resolution, thus converting the physical thrust requirement into an executable digital control quantity. Then, quantization errors are eliminated through numerical inverse calculation rather than empirical estimation, achieving a one-to-one correspondence between the target thrust and the drive command. After the initial duty cycle command is output, the motor generates actual thrust. Within the same refresh cycle, the controller reads the real-time attitude quaternion and converts the actual thrust to the inertial coordinate system. It then calculates the vector difference between the actual thrust and the target thrust amplitude Fd to generate a three-dimensional attitude error vector. Subsequently, the three-dimensional attitude error vector is input into the proportional-integral correction channel. The proportional loop performs instantaneous scaling on the three-dimensional attitude error vector with gain kp to quickly reduce directional deviation. The integral loop performs time integration on the three-dimensional attitude error vector with gain ki to accumulate and offset residual amplitude error. Then, the duty cycle increments output by the two loops are superimposed to obtain the update command. This achieves dual closed-loop tracking of the thrust vector to the moving target, enabling the underwater booster to maintain consistent and robust power response even when attitude changes rapidly, user gear shifts frequently, or environmental disturbances intensify.

[0036] It should be noted that when the underwater booster enters a narrow, enclosed underwater space and encounters both wake vortex interference and a decrease in thrust, the intensity of the external crossflow increases significantly. Meanwhile, the thruster in the corresponding direction, due to limitations or malfunctions, experiences insufficient maximum output. The resulting attitude error vector persists, but the duty cycle adjusted by the PI correction channel, even after reaching its physical upper limit, cannot meet the compensation requirements, leading to saturation of the thrust direction adjustment. The original closed-loop control logic no longer possesses actual correction capabilities and continues to accumulate errors, ultimately causing slow attitude drift or even control instability. Furthermore, because this type of thrust saturation behavior manifests as a stable, small-amplitude attitude error in the inertial navigation feedback, it is difficult to distinguish from ordinary perturbations and is easily misjudged as residual disturbances, failing to trigger effective handling. This results in a hidden failure where the control closed loop appears to operate normally on the surface but the actual deviation continues to expand. Therefore, in this embodiment… After acquiring the target attitude and the actual attitude state, when it is detected that the thrust direction adjustment of the thruster has reached the physical limit of the duty cycle and the attitude error continues to increase, the thrust vector, angular velocity information, external lateral flow velocity estimation results and attitude drift trajectory in the last sixty seconds are first called to form multi-dimensional state sequence data.

[0037] The thrust vector is calculated by linear interpolation using a pre-calibrated thrust curve after real-time current reading from a Hall current sensor. Then, a rotation matrix is ​​constructed based on the current quaternion attitude to map the thrust from the body coordinate system to the inertial coordinate system. Angular velocity information is obtained by outputting from a three-axis gyroscope and correcting for temperature drift using a second-order low-pass filter. The external lateral flow velocity estimation results are synchronously acquired by four miniature Doppler ultrasonic velocity profile sensors deployed around the shell and solved using the weighted least squares method. The attitude drift trajectory is tracked by recording the quaternion difference between the target attitude and the actual attitude in each cycle and mapping it to an axis-angular offset scalar, forming a complete time series.

[0038] Subsequently, the multidimensional state sequence data is compressed and mapped to the latent space representation z=fenc(x) by embedding the structure encoder network fenc(), and the residual estimate of the desired attitude trajectory is reconstructed in real time: r=fdec(z)–xref.

[0039] Due to factors such as limited local thrusters, inertial navigation frequency oscillations, or unsteady changes in flow characteristics during abrupt attitude changes, isolated anomalous subsequences appear in the multidimensional state sequence as short-term, high-amplitude drifts in the time-series structure but do not significantly affect the global trend. Therefore, before entering the self-encoding compression stage, the multidimensional state sequence is first subjected to Fast Fourier Transform to extract the principal energy period, and a folding operation is performed on the time axis based on this period to construct a time-series-frequency domain multi-scale annular coordinate system. Furthermore, singular value decomposition is used to extract the component with the most significant rotational variation characteristics in the multidimensional state sequence, constructing a rotational decoupled gradient tensor to significantly improve the sensitivity to periodic disturbances such as the high-frequency component wake pulsations of external lateral flow velocity.

[0040] Next, the difference between the current spin-direction decoupled gradient tensor and the reference tensor established in the still water stable region is calculated and input into the cyclic autocorrelation operator. The resonance period difference between the first two main peaks is extracted and defined as the phase offset index πc of the isolated perturbation, which is used to quantify the phase jump caused by the wake perturbation.

[0041] Simultaneously, energy calculations are performed on the temporal data of the rotationally decoupled gradient tensor to obtain the instantaneous energy curve. A higher-order perturbation discrimination pointer is constructed by multiplying the phase offset index by the instantaneous energy. Then, a sliding time sampling window is set, and the average and median of the higher-order perturbation discrimination pointer are continuously tracked within the sliding time window. A discrimination benchmark curve is plotted and updated in real time with the sliding time sampling window. When the higher-order perturbation pointer at the current moment is higher than the discrimination benchmark curve in two consecutive sampling windows, it is determined that the current state has exhibited a dual abrupt jump characteristic of period and phase. A first-order high-sensitivity mask M1 is immediately generated to mark anomalous mutation segments. Subsequently, the segments marked by M1 are input into a perturbation anomaly discrimination function constructed based on the variance and principal direction offset ratio index of multidimensional state sequence slices to determine the perturbation mask vector M.

[0042] Furthermore, the phase offset index πc obtained from the previous refresh cycle is regarded as a continuous quantity characterizing the amplitude of the wake vortex-phase jump. The perturbation mask vector M generated by the higher-order perturbation pointer is regarded as a time-domain weight describing the time span of the energy anomaly. The product of the two is calculated in a unified time grid, so that the amplitude information of the phase jump and the continuous information of the energy anomaly are naturally coupled on the same continuous curve. The deviation of this continuous curve from the still water steady-state reference band is monitored in real time and suppressed by exponential decay kernel smoothing.

[0043] When the product continuously crosses the dynamic threshold band formed by the historical static steady-state mean plus double deviation, it is determined that the segment has simultaneously exhibited the dual jump characteristics of phase and energy. Then, a consistency test is performed on the ratio of the variance to the mean of the phase offset within the segment. Only when the ratio is less than the preset upper limit and the segment length does not exceed half of the main energy cycle is it finally confirmed as an isolated anomalous subsequence, and a unique anomalous mask is generated on the global timeline.

[0044] Subsequently, by controlling the isolated anomalous subsequence to only participate in the potential function offset correction operation in the residual construction process without entering the main path network encoding, the directional perturbation of the residual estimation is suppressed, and a new residual R is determined. The new residual obtained in this way accurately reflects the macroscopic position trend caused by thrust saturation on the one hand, and completely retains the microscopic nonlinear offset information brought about by the wake perturbation on the other hand.

[0045] Subsequently, the reverse gradient of the new residual is used as the spatial direction of the vector potential energy depression, and the exponential decay form of the square of the new residual modulus is defined as the new residual potential energy function φ(R)=exp(–||R||² / σ²), where σ represents the steady-state residual variance calibration constant, determined by the wake vortex free decay experiment. Then, the new residual potential energy function is used to perform a continuously differentiable potential energy mapping on the residual field, and the descent direction of the potential energy depression gradient is solved in real time. Thus, a pseudo-potential field model that is not physical but mathematically differentiable is constructed in a mathematical sense. Then, each gradient vector of the pseudo-potential field is projected onto the thruster coordinate system. First, the new residual modulus is used as the scale of the thrust amplitude increment, and then the proportional coefficient is allocated according to the gradient direction. The remaining available thrust margin is mapped to the increment of each channel to generate a self-guided correction thrust distribution field. It should be noted that the thrust distribution always decreases along the new residual equipotential surface rather than continuing to track the original attitude target. At the same time, when the wake vortex continues to act or the thrust is saturated, it can automatically adjust the direction and amplitude so that the corrected trajectory gradually approaches the new residual minimum point along the path of lowest potential energy, thereby making a built-in response to the real disturbance trend and realizing the asymptotic elimination of attitude deviation.

[0046] Furthermore, to suppress abrupt jumps in the pseudo-force field near the perturbation source, the mask vector of the isolated anomalous subsequence is expanded into a perturbation weight mask with the same dimension as the new residual gradient, and a vector offset modulation with a hyperbolic tangent function as the kernel is superimposed on the principal gradient direction: as the gradient magnitude gradually increases, the weight of the perturbation direction component is automatically reduced, so that the thrust guidance trajectory exhibits monotonically continuous changes after entering the buffer transition zone; when the new residual further decreases and moves away from the perturbation source, the weight is restored to the full-amplitude state consistent with the principal gradient, ensuring that the overall convergence speed is not weakened, and that the thrust distribution remains coherent and smooth throughout the entire field, so that no matter how the perturbation intensity fluctuates, it will not trigger abrupt changes in direction or generate secondary oscillations.

[0047] Simultaneously, a navigation boundary constraint function is embedded within the pseudo-force field framework. Real-time ranging sonar and inertial position estimation are used to jointly determine the minimum safe distances from the outer surface of the hull to the walls, cabin corners, and unidirectional flow boundaries. The distance error is then incorporated into the gradient barrier potential energy, which is in the form of the inverse square. This gradient barrier potential energy rapidly increases the potential energy threshold near the boundary, causing the synthesized gradient to spontaneously deviate from the impassable region. Far from the boundary, the gradient barrier decays to zero, without affecting the convergence of the main gradient. This ensures that the thrust guidance direction can avoid walls and dead angles in real time under dynamic spatial deformation and abrupt channel turns, completely eliminating the deadlock risk caused by the single-axis thrust limit.

[0048] Subsequently, to ensure that the underwater booster does not become unstable, roll over, or rapidly descend due to attitude deviations during multi-degree-of-freedom maneuvers, the pitch and roll angles calculated in real time by the inertial measurement unit and the diving depth output by the pressure sensor are called up in the same refresh cycle. The three together form a three-dimensional safety monitoring vector. Then, using the multi-level limit range in the factory calibration table as a reference, the real-time attitude-depth points are mapped to a three-segment dynamic envelope of "safe zone - critical zone - limit zone".

[0049] When the monitoring vector enters the critical region, the thrust amplitude is immediately reduced by weighting the current residual potential field gradient, and the duty cycle is reduced to the next higher level within one refresh cycle, thus reducing power output in advance and slowing down the attitude evolution rate. A safety limit is set, and if the monitoring vector exceeds the safety limit, a bright red warning light at the front of the shell is activated simultaneously, and an acoustic alarm is issued via a bone conduction buzzer. At the same time, the existing thrust command is frozen, a dynamic downshifting routine is executed, and the duty cycle of all motors is forcibly reduced to the lowest safe level. The system then switches to a low-speed attitude maintenance strategy to achieve emergency stop instability suppression.

[0050] To further avoid mid-course loss of control due to power decay, dynamic buffers with hysteresis are set in the three channels of attitude, depth and remaining power: each channel predicts the future trend with exponential smoothing and compares the predicted value with the current safety envelope; once the predicted curve touches the critical zone boundary within the reserved time window, the thrust is reduced or the potential field convergence path is extended in advance across cycles, so as to replace the ex-post fault handling with preventive attitude convergence.

[0051] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0052] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0053] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and inventive constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0054] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0055] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0056] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for multi-degree-of-freedom motion control and attitude self-adaptive adjustment of an underwater booster, The application is characterized in that it comprises: The symmetric mass distribution is constructed and the attitude convergence verification is completed, and the natural stable initial configuration is established; The thrust gear table is written, the water entry state is identified through the double-channel water sensitivity judgment, the gear shifting control is unlocked and the power output preparation is entered; The attitude quaternion is obtained and combined with the user gear input, the six-degree-of-freedom motion target is analyzed, the three-dimensional thrust vector is decomposed, the initial duty ratio is generated by calling the reverse interpolation, and the thrust direction and amplitude are corrected through the proportional-integral channel; After detecting the thrust saturation and attitude drift, the multi-dimensional state sequence is collected to identify the disturbance segment, the residual vector is constructed and the pseudo potential field is generated, the remaining thrust is guided to adaptively correct the deviation, and the boundary potential energy and attitude monitoring are fused.

2. The method of claim 1, wherein: The power supply and the sensor are arranged in isolation in the three-dimensional space, the angle deviation curve of the underwater booster in the still water pool is recorded in real time, the bubble level reading is taken as the checking signal, the convergence rate of the angle deviation with time is compared, and the attitude of the underwater booster is adjusted to be naturally stable.

3. The method of claim 2, wherein: After the attitude is stabilized, the thrust gear table is established; the gear shifting and resetting are controlled by the button; the underwater booster is powered on, the water entry judgment is established, and the thrust output is unlocked; if the underwater booster is not in water, the thrust output is locked and reset.

4. The method of claim 3, wherein the method further comprises: The current attitude state of the underwater booster is obtained based on the sensor data, and the six-degree-of-freedom motion target is analyzed combined with the user input to generate a three-dimensional thrust distribution; the attitude deviation is continuously corrected during the control process.

5. The method of claim 4, wherein: The target thrust is mapped to the initial duty ratio command, the actual thrust and the target thrust are compared to generate the attitude error, and the proportional-integral calculation is performed on the error to obtain the duty ratio increment, and the driving command is updated in real time to track the thrust vector in a closed loop.

6. The method of claim 5, wherein: The multi-dimensional state sequence data is composed of the thrust vector, angular velocity information, external transverse flow velocity estimation result and attitude drift trajectory, and the residual estimation of the expected attitude trajectory is reconstructed in real time.

7. The method of claim 6, wherein: A time-frequency multi-scale ring coordinate system is constructed based on the multi-dimensional state sequence, a rotational decoupling gradient tensor is obtained, the current rotational decoupling gradient tensor is subtracted from the reference tensor established in the still water stable interval, and the difference is input into a circular autocorrelation operator to extract the resonance period difference between the first two main peaks to determine the phase offset index; The time series data of the rotational decoupling gradient tensor is subjected to energy operation to obtain an instantaneous energy curve, and a high-order disturbance discrimination pointer is constructed by multiplying the phase offset index and the instantaneous energy, and then a first-order high-sensitivity mask is generated based on the dynamic change of the high-order disturbance discrimination pointer, and is input into a disturbance anomaly discrimination function constructed based on the variance and main direction offset ratio index of the multi-dimensional state sequence slice to determine the disturbance mask vector; The phase offset index is regarded as a continuous quantity describing the amplitude of the tail vortex-phase jump, and the disturbance mask vector is regarded as a time domain weight describing the time span of the energy anomaly, and the product of the two in the unified time grid is calculated, when the product continuously crosses the dynamic threshold band composed of the historical still water stable mean value plus double deviation, the phase offset variance and mean value ratio in the internal segment are subjected to consistency test to determine the isolated anomaly subsequence, and a unique anomaly mask is generated on the global timeline.

8. The method of claim 7, wherein the method further comprises: The isolated abnormal sub-sequence is only involved in the potential function offset correction operation in the residual error construction process through the mask vector control, the directional disturbance of the residual error estimation is inhibited, and a new residual error R is determined.

9. The method of claim 8, wherein: A pseudo-potential field is established with the new residual error gradient, the remaining thrust margin is distributed to each propulsion channel along the equipotential surface in descending order, the pseudo-potential field main gradient is weighted by the unique abnormal mask, and the instantaneous amplitude close to the disturbance source direction is weakened; then the minimum safety distance obtained by real-time ranging and position estimation is converted into a barrier potential, which is superimposed into the pseudo-potential field, so that the synthesized gradient automatically deviates from the forbidden area when approaching the wall surface or narrow angle.

10. The method of claim 9, wherein the method further comprises: The attitude and depth safety monitoring vectors are constructed and mapped to the multi-level envelope of the safety zone-critical zone-limit zone, the thrust gear is reduced when entering the critical zone, the alarm is triggered and forced to downshift when breaking through the limit zone; and a dynamic buffer zone is set up to replace the post-fault disposal with preventive attitude convergence.

Images