A lumped disturbance observation and adaptive super-twisting sliding mode control method and system for a micro autonomous underwater robot

Abstract

The application discloses a kind of collective disturbance observation and adaptive super-twisting sliding mode control method and system for micro autonomous underwater robot, belong to underwater robot motion control technical field, the method establishes six degrees of freedom dynamics model and model uncertainty and external disturbance are uniformly expressed as acceleration layer collective disturbance, using super-twisting nonlinear disturbance observer is continuously estimated on line to collective disturbance and outputs disturbance estimate, according to pose error and velocity error, construct nonsingular terminal sliding mode surface, design includes nominal equivalent control term, disturbance feedforward compensation term based on disturbance estimate and adaptive super-twisting robust switching term Compound control law, wherein the control gain of super-twisting robust switching term is on-line self-adapting adjustment according to sliding mode surface envelope and disturbance envelope, the synthesized six degrees of freedom generalized control force is converted into eight propeller thrust command output by thrust distribution matrix to drive robot motion.

Description

Technical Field

[0001] This invention relates to the field of underwater robots, specifically to a lumped disturbance observation and adaptive super-twisted sliding mode composite control method for micro-sized autonomous underwater robots or eight-propeller omnidirectional underwater robots. Background Technology

[0002] Miniature autonomous underwater vehicles (AUVs) have high application value in tasks such as port inspection, underwater pipeline inspection, ship observation, precision operations in narrow waters, and fixed-point hovering. These robots are typically characterized by small size, low inertia, high maneuverability, and ease of deployment, but they are also more susceptible to factors such as uncertainties in hydrodynamic parameters, external ocean current disturbances, thruster coupling, and sensor noise.

[0003] For omnidirectional underwater robots with eight thrusters, they are typically capable of generating control forces and torques in six degrees of freedom: pitch, sway, heave, roll, pitch, and yaw. While these platforms possess good omnidirectional maneuverability, their control systems often face the following challenges:

[0004] 1. Parameters such as added mass, damping, and restoring torque are difficult to obtain accurately in the dynamic model of underwater robots;

[0005] 2. External disturbances such as ocean currents, local eddies, and near-wall effects are time-varying and cannot be directly measured;

[0006] 3. Micro-platforms have smaller mass and inertia, making them more sensitive to model errors and external disturbances;

[0007] 4. The six-degree-of-freedom omnidirectional motion system has the characteristics of strong coupling, multiple inputs and multiple outputs, and thruster redundancy;

[0008] 5. Traditional sliding mode control is prone to chattering, while fixed gain overtwitch control is difficult to balance robustness and smoothness.

[0009] Existing control methods mainly include PID control, backstepping control, model predictive control, traditional sliding mode control, fixed-gain super-twisted sliding mode control, and compensation control methods based on extended state observers or disturbance observers. These methods can achieve motion control of underwater robots under certain operating conditions, but they still have shortcomings in the scenarios of strong disturbances, strong coupling, and low-speed precision operations of micro-sized omnidirectional AUVs.

[0010] Specifically, existing technologies typically have the following problems:

[0011] Strong model dependency leads to decreased control accuracy under parameter mismatch.

[0012] Existing controllers often rely on nominal dynamic models. When the actual added mass, damping parameters, load state, or center of buoyancy and center of gravity relationship change, deviations will occur between the nominal model and the real system, leading to inaccurate control compensation, which manifests as increased trajectory error, reduced convergence speed, or attitude oscillation.

[0013] Linear disturbance observers are insufficient for responding to fast time-varying disturbances.

[0014] Methods such as linear extended state observers are typically based on fixed bandwidth designs. When the bandwidth is small, the response is lag-prone; when the bandwidth is large, noise is easily amplified. Therefore, it is difficult to simultaneously achieve both the speed and smoothness of disturbance estimation.

[0015] Traditional sliding mode control suffers from high-frequency chattering.

[0016] Traditional sliding mode control relies on sign function switching terms, which can easily generate high-frequency chattering in the control input. For thruster-driven underwater robots, this chattering can cause frequent fluctuations in thruster commands, increased energy consumption, wear on actuators, and unmodeled high-frequency dynamic excitation.

[0017] Fixed-gain overtwist control lacks online adjustment capability.

[0018] While fixed-gain overtorsion control can reduce traditional sliding mode chattering, its gain is typically selected based on the worst-case disturbance. When the disturbance is small, a fixed high gain can lead to an excessively large control input; when the disturbance increases, a fixed low gain may result in insufficient robustness. Therefore, the fixed-gain method struggles to simultaneously meet the requirements of speed, robustness, and smoothness under different operating conditions.

[0019] Insufficient coordination between observer and controller

[0020] Existing disturbance observers are often designed independently as external compensation modules, and the controller still mainly relies on the switching term to suppress errors. The lack of a coordination mechanism between the observation information and the robust control term results in a heavy burden on the robust switching term, making it difficult to fully utilize the combined advantages of "disturbance feedforward compensation + robust feedback control".

[0021] Therefore, a composite control method for micro-sized omnidirectional underwater robots is needed, which can uniformly observe model uncertainties and external disturbances, and combine non-singular terminal sliding mode and adaptive super-twist control to achieve high-precision, low-bounce, and robust trajectory tracking with six degrees of freedom. Summary of the Invention

[0022] To address the challenges faced by miniature omnidirectional autonomous underwater vehicles (AUVs) in scenarios involving strong disturbances, strong coupling, and low-speed precision operations, such as strong model dependence, delayed disturbance response, sliding mode chattering, and insufficient observer-controller coordination, this invention provides a lumped disturbance observation and adaptive super-twisted sliding mode control method and system for miniature AUVs. The specific technical solution includes the following steps:

[0023] S1. State Acquisition and Construction of Lumped Perturbation Model

[0024] The current state (including current pose and velocity) and desired trajectory of the underwater robot are obtained. A six-degree-of-freedom dynamic model of the underwater robot is established, and model uncertainties and external disturbances are uniformly represented as lumped disturbances in the acceleration layer. Specifically, the underwater robot's additional mass error, damping error, restoring force error, unmodeled dynamics, and external ocean current disturbances are uniformly converted into variables to be estimated in the acceleration layer, so that the controller can compensate for a single unified disturbance object, thereby reducing the complexity of the controller design.

[0025] S2. Online estimation of lumped perturbation based on super-twisted nonlinear perturbation observer (STNDO)

[0026] A super-twisted nonlinear perturbation observer is used to continuously estimate the lumped perturbation of the acceleration layer online, outputting the lumped perturbation estimate. The specific implementation mechanism is as follows: an extended state containing both velocity and acceleration layer lumped perturbations is constructed. Under the assumption that the perturbation derivative is bounded, a second-order sliding mode perturbation reconstruction is achieved by placing discontinuous terms in the integration channel through a super-twisted structure, thereby outputting a continuous acceleration layer lumped perturbation estimate and avoiding chattering caused by directly outputting the sign switching term.

[0027] S3, Non-singular terminal sliding surface structure

[0028] A non-singular terminal sliding surface is constructed based on the acquired pose and velocity errors. The exponential parameters of the non-singular terminal sliding surface satisfy preset conditions, ensuring that the power term in its derivative does not exhibit a negative exponent when the sliding surface approaches the origin. This avoids the singularity problem of the control input tending to infinity and guarantees the rapid convergence characteristics of the six-degree-of-freedom trajectory error.

[0029] S4, Three-layer Composite Control Law Design

[0030] Design a composite control law consisting of a nominal equivalent control term, a disturbance feedforward compensation term, and an adaptive super-twisted robust switching term:

[0031] (1) Calculate the nominal equivalent control term to offset the known dynamic term and transform the nonlinear coupled system into an error system that is easy to control;

[0032] (2) Calculate the disturbance feedforward compensation term based on the lumped disturbance estimate output by S2, which is used to actively offset the main components of the lumped disturbance and reduce the burden of the subsequent robust switching term;

[0033] (3) Design an adaptive super-twisted robust switching term to suppress observation residuals and unmodeled errors.

[0034] S5, Online adaptive adjustment of super-twist control gain

[0035] In the adaptive super-twist robust switching term of S4, the super-twist control gain is adaptively adjusted online, specifically including:

[0036] The sliding mode error and the normalized disturbance estimate are filtered to generate sliding mode envelope variables and disturbance envelope variables;

[0037] Based on the sliding mode envelope variable, the disturbance envelope variable, and the preset threshold, a dynamic lower bound for the control gain is generated and first-order smoothing is performed.

[0038] The first-level gain is adjusted in segments according to the magnitude of the sliding mode error, and the upper and lower bounds of the first-level gain are limited by the projection operator, while the second-level gain is linked with the first-level gain.

[0039] The super-twist control gain is increased when the sliding mode error is large or the disturbance is enhanced, and the super-twist control gain is decreased when the sliding mode error is reduced and the disturbance is weakened, so as to balance robustness and smoothness.

[0040] S6, Control Law Synthesis and Thrust Distribution Output

[0041] The nominal equivalent control term, disturbance feedforward compensation term, and adaptive super-twisted robust switching term are combined into a six-degree-of-freedom generalized control force. The six-degree-of-freedom generalized control force is converted into an eight-thruster thrust command through a thrust allocation matrix and sent to the thruster execution unit to drive the underwater robot to perform omnidirectional motion control.

[0042] Corresponding to the above control method, the present invention provides a control system for micro-sized autonomous underwater robots, the system comprising:

[0043] The status acquisition module is used to acquire the current status and desired trajectory of the underwater robot;

[0044] The disturbance modeling module is used to establish a six-degree-of-freedom dynamic model and to represent model uncertainties such as additional mass error and damping error, along with external disturbances, as a unified lumped disturbance of the acceleration layer.

[0045] The disturbance observation module is used to continuously estimate the lumped disturbance of the acceleration layer online using a super-torsional nonlinear disturbance observer and output the lumped disturbance estimate.

[0046] The sliding surface construction module is used to construct a non-singular terminal sliding surface based on the pose error and velocity error, and its exponential parameters satisfy the preset conditions for avoiding singular control.

[0047] The composite control calculation module is used to calculate the nominal equivalent control term, the disturbance feedforward compensation term based on the lumped disturbance estimate, and the adaptive super-twisted robust switching term; wherein, the super-twisted control gain is adaptively adjusted online by combining the projection operator according to the sliding mode envelope variable, the disturbance envelope variable, and the dynamic lower bound.

[0048] The thrust allocation module is used to convert the synthesized six-degree-of-freedom generalized control forces into eight-thrust commands through the thrust allocation matrix.

[0049] The present invention also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the various steps in the above-described control method.

[0050] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the various steps of the above-described control method.

[0051] The advantages of this invention are:

[0052] 1. More unified disturbance handling: Model errors and external disturbances are unified into lumped disturbances, reducing the complexity of controller design;

[0053] 2. Faster perturbation estimation: STNDO can achieve fast perturbation estimation under the condition that the perturbation derivative is bounded;

[0054] 3. Smoother perturbation estimation: The super-twisted structure avoids directly outputting the traditional sign switching term, reducing observation chattering;

[0055] 4. Faster trajectory error convergence: The non-singular terminal sliding surface enables the error to have finite-time convergence characteristics;

[0056] 5. Smoother control input: Disturbance feedforward compensation reduces the burden of robust switching terms, and adaptive gain mechanism avoids high gain throughout the entire time period;

[0057] 6. Enhanced robustness: It can maintain high tracking accuracy even under model parameter mismatch and strong time-varying disturbances;

[0058] 7. Clearer engineering implementation: The controller output is a six-degree-of-freedom generalized control force, which can be directly converted into eight-thrust commands through the thrust distribution matrix;

[0059] 8. Wide range of applications: Suitable for fixed-point hovering, low-speed trajectory tracking, attitude control, inspection in narrow waters and operation in highly disturbed environments of micro-sized omnidirectional AUVs. Attached Figure Description

[0060] Figure 1 Structure diagram of a lumped disturbance observation and adaptive super-torsional sliding mode control system for a miniature omnidirectional AUV;

[0061] Figure 2 This is a structural diagram of the STNDO super-twisted nonlinear perturbation observer;

[0062] Figure 3 Flowchart for adaptive super-twisted gain adjustment. Detailed Implementation

[0063] 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.

[0064] This invention provides a lumped disturbance observation and adaptive super-torsional sliding mode composite control method for micro-sized autonomous underwater robots. The method includes:

[0065] 1. Six-degree-of-freedom dynamic modeling and lumped disturbance reconstruction;

[0066] 2. Design of the STNDO (Super-Twisted Nonlinear Perturbation Observer);

[0067] 3. Non-singular terminal sliding surface construction;

[0068] 4. Design of nominal equivalent control terms;

[0069] 5. Design of disturbance feedforward compensation term based on STNDO;

[0070] 6. Adaptive super-twisted robust switching design;

[0071] 7. Control law synthesis and thruster output distribution.

[0072] The inputs to the control system of this invention are the desired trajectory, current position and attitude, current velocity, and sensor measurement information; the outputs are six-degree-of-freedom generalized control force and eight-jet thrust commands.

[0073] Actual robot platform and control system input / output

[0074] This invention takes the BlueROV Heavy eight-propeller underwater robot as the implementation object.

[0075] (1) System input

[0076] The inputs to the control system of this invention include:

[0077] A. Task Input

[0078] • Target location;

[0079] • Target posture;

[0080] • Or target trajectory.

[0081] It can be uniformly written as:

[0082] in: For the target location; For the target posture; Indicates the desired pose.

[0083] If necessary, you can also input the target velocity and target acceleration:

[0084] in: Indicates the rate of change of the target pose; This represents the acceleration due to the change in the target's pose.

[0085] B. Sensor Input

[0086] Provided by BlueROV Heavy airborne sensors:

[0087] • Attitude and angular velocity measured by the IMU;

[0088] • Depth measured by a depth sensor;

[0089] • Linear velocity provided by the DVL or velocity estimation module;

[0090] • The current position and attitude output by other state estimation modules.

[0091] This information collectively constitutes the current state:

[0092] (2) System output

[0093] The output of the control system of this invention includes two stages:

[0094] A. First-level output: Six-degree-of-freedom control input

[0095] The controller first outputs:

[0096] in: For forward and backward control force; For left and right control force; For vertical control force; For roll control torque; For pitch control torque; This is the control torque in the yaw direction.

[0097] B. Secondary Output: Eight-Thrust Command

[0098] After thrust distribution, eight thruster commands are output:

[0099] in: These represent the thrust commands for the eight thrusters of the BlueROV Heavy.

[0100] Therefore, the control system of the present invention can be simply understood as:

[0101] Input: Target trajectory + Current state

[0102] Output: Six-DOF control input + eight-jet thrust command

[0103] Control object and state definition

[0104] Define the pose vector of the underwater robot in the inertial coordinate system as:

[0105] in, These represent the position coordinates, These represent the roll angle, pitch angle, and yaw angle, respectively.

[0106] Define the velocity vector in the carrier coordinate system as:

[0107] in, For three-axis velocities, It represents the angular velocity of the three axes.

[0108] The six-degree-of-freedom kinematic equations are:

[0109] in, This is the coordinate transformation matrix.

[0110] The six-degree-of-freedom dynamic equations of the underwater robot are:

[0111] in, The inertia matrix, The matrix represents the Coriolis force and the centripetal force. Here is the damping matrix. For restoring force, For six degrees of freedom generalized control, It is a lumped disturbance consisting of model uncertainty and external disturbance.

[0112] For an eight-thrust omnidirectional configuration, the thrust vector of the thrusters is defined as follows:

[0113] The generalized control force and the thruster thrust satisfy the following:

[0114] in, Assign a matrix to the thrust.

[0115] Unified modeling of lumped disturbances

[0116] To avoid dealing with additional mass error, damping error, restoring force error and external ocean current disturbance separately, this invention converts the above uncertainties into a unified acceleration layer lumped disturbance.

[0117] Define a known dynamic term:

[0118] Define the input matrix:

[0119] Define the lumped perturbation of the acceleration layer:

[0120] The underwater robot's dynamic equations can then be reconstructed as follows:

[0121] This form unifies complex uncertainties into a single variable to be estimated. This facilitates observer design and controller feedforward compensation.

[0122] Design of a super-twisted nonlinear perturbation observer

[0123] To estimate lumped perturbations online, this invention designs a super-twisted nonlinear perturbation observer (STNDO).

[0124] Define extended state:

[0125] The extended system is then:

[0126] Suppose that the derivative of the perturbation is bounded, that is, that there exists a positive constant. , so that:

[0127] Define velocity observation error:

[0128] Constructing STNDO:

[0129] in, This is a speed estimate. This is the estimated value of the lumped disturbance in the acceleration layer. , This is the gain matrix of the positive definite observer.

[0130] The observer output is:

[0131] It can also be converted into force and moment layer disturbance estimates:

[0132] Compared to the linear extended state observer, this STNDO achieves second-order sliding mode disturbance reconstruction through a super-twisted structure. The discontinuity term is located in the integral channel, so the output disturbance estimate has continuity and can be directly used for controller feedforward compensation.

[0133] Trajectory error and non-singular terminal sliding surface design

[0134] Let the desired trajectory be:

[0135] And assume , , Smooth and available.

[0136] Define pose error:

[0137] Define speed error:

[0138] To achieve fast convergence and avoid the singularity problem of traditional terminal sliding mode, a non-singular terminal sliding surface is constructed:

[0139] in, For sliding mode variables, For a positive definite diagonal matrix, the exponent parameter satisfies:

[0140] because ,exist When the power term in the sliding surface derivative does not have a negative exponent singular term, the problem of the control input tending to infinity in the traditional terminal sliding mode control law is avoided.

[0141] Design of Nominal Equivalent Control Items

[0142] This invention first designs a nominal equivalent control term to compensate for the known dynamics of the underwater robot and makes the error system dynamically evolve according to a preset sliding mode.

[0143] according to:

[0144] We can obtain:

[0145] Combination:

[0146] Design equivalent control terms without considering disturbances. It is used to counteract known dynamics and achieve the desired error acceleration.

[0147] A preferred equivalent control term can be expressed as:

[0148] This item corresponds to model compensation and expected sliding mode dynamic generation.

[0149] The function of this equivalent control item is:

[0150] 1. Counteract the known inertia, hydrodynamic damping, and restoring force terms;

[0151] 2. Transform the six-degree-of-freedom nonlinear coupled system into an error system that is easy to control using sliding mode;

[0152] 3. Provides basic control inputs for subsequent disturbance compensation and robust switching control.

[0153] Design of perturbation feedforward compensation term based on STNDO

[0154] Perturbation estimates are obtained from STNDO:

[0155] Design disturbance feedforward compensation term:

[0156] Right now:

[0157] This term is used to actively offset the main components of the lumped disturbance. After compensation, the remaining disturbance in the system is:

[0158] Or acceleration layer residuals:

[0159] Therefore, the controller no longer relies entirely on robust switching terms to combat all disturbances, but only needs to handle observation residuals and unmodeled residuals, thereby reducing switching gain requirements and minimizing control chattering.

[0160] Adaptive Super-Twisted Robust Switching Design

[0161] To suppress observation residuals and unmodeled errors, this invention further designs an adaptive super-twisted switching control term.

[0162] Define and switch virtual control inputs:

[0163] in, , This is the adaptive gain matrix. This is the super-distorted internal integral state. It is a smooth saturation function with a boundary layer.

[0164] Preferably, the smoothing saturation function is taken as:

[0165] The internal integral state satisfies:

[0166] in, This is the leakage integral coefficient matrix. The leakage term is used to avoid long-term accumulation of integral states and enhance the boundedness in engineering implementation.

[0167] Mapping virtual control inputs to generalized control forces:

[0168] An equivalent mapping method can also be selected based on the actual control coordinates.

[0169] Adaptive gain adjustment mechanism

[0170] One of the key innovations of this invention is that the super-torsion control gain is not a fixed constant, but is adjusted online according to the sliding mode error and the disturbance estimation state.

[0171] Define sliding mode envelope variables:

[0172] Define the perturbation envelope variable:

[0173] in, , The filtering time constant is The normalized perturbation estimator can be taken as:

[0174] in, For the first Estimated values ​​of the perturbation for each degree of freedom. The perturbation normalization scale, To prevent positive numbers from being divided by zero.

[0175] Constructing a dynamic lower bound for first-order gain:

[0176] in, , This is the minimum gain of the first stage. The perturbation threshold, , The upper and lower thresholds of the sliding mode envelope, , This is the adjustment coefficient.

[0177] First-order smoothing of the dynamic lower bound:

[0178] Level 1 gain The update law is:

[0179] in, To implement a soft-start function, avoid over-activation of the controller during the initial phase: Determined based on sliding envelope and dynamic lower bound segmentation:

[0180] And by limiting it through the projection operator:

[0181] The secondary gain is linked to the primary gain:

[0182] The function of this adaptive mechanism is:

[0183] 1. When the sliding mode error is large, automatically increase the value. , This improves the ability to suppress disturbances.

[0184] 2. When the error enters a small neighborhood and the disturbance is weak, the gain is automatically reduced to decrease control chattering;

[0185] 3. When the disturbance estimate is still large, a dynamic lower bound is used to prevent the gain from decreasing prematurely;

[0186] 4. Projection constraints ensure that the control gain is always bounded, which facilitates engineering implementation.

[0187] Composite control law synthesis

[0188] The final controller output consists of three parts:

[0189] in:

[0190] This is the nominal equivalent control term, used to compensate for known dynamics;

[0191] This is a perturbation feedforward compensation term used to offset the lumped perturbation estimated by STNDO;

[0192] An adaptive super-twisted robust switching term is used to suppress observation residuals and unmodeled errors.

[0193] The six-degree-of-freedom generalized control force output by the controller is:

[0194] The thrust command for the eight thrusters is obtained through thrust distribution:

[0195] in, The pseudoinverse or weighted pseudoinverse of the thrust assignment matrix is ​​obtained.

[0196] It was then distributed to the underwater robot thruster actuator unit.

[0197] The present invention is preferably performed according to the following steps:

[0198] S100: Get the current state

[0199] Obtain the current pose through an IMU, depth sensor, DVL, or state estimation module. and speed .

[0200] S200: Input desired trajectory

[0201] Input desired pose Expected speed and expected acceleration .

[0202] S300: Calculation of lumped disturbance observation model

[0203] Calculated based on the nominal dynamics model:

[0204] And construct an acceleration layer system:

[0205] S400: Running STNDO disturbance observer

[0206] Update based on current velocity, control input, and nominal model terms:

[0207] And output the lumped disturbance estimate:

[0208] S500: Calculates tracking error and sliding mode variables

[0209] calculate:

[0210] And generate non-singular terminal sliding mode variables:

[0211] S600: Calculate the three-term control law

[0212] Calculate the nominal equivalent control item:

[0213] Calculate the disturbance feedforward compensation term:

[0214] Calculate the adaptive super-twist switching control term:

[0215] S700: Composite six-degree-of-freedom control quantity

[0216] S800: Thruster Distribution and Output

[0217] Based on the thrust distribution matrix:

[0218] Output the thrust command for the eight thrusters.

[0219] S900: Closed-loop feedback

[0220] After the robot executes the thruster command, the sensors re-collect the current state and enter the next control cycle.

[0221] Without departing from the core idea of ​​this invention, each core element in the above-described specific technical solutions can be replaced or extended with the following equivalent alternatives to achieve the purpose of this invention:

[0222] Alternatives to the perturbation observer: The super-twisted nonlinear perturbation observer (STNDO) is not limited to the super-twisted structure and can be replaced by other perturbation observers with finite-time convergence or high robustness, such as higher-order sliding mode perturbation observers, nonlinear extended state observers, or adaptive perturbation observers, as long as they are still used for online estimation of lumped perturbations and for feedforward compensation.

[0223] Alternatives for the sliding surface: The non-singular terminal sliding surface can be replaced by a fast non-singular terminal sliding surface, an integral non-singular terminal sliding surface, or other finite-time convergent sliding surfaces, as long as they can avoid singular control and achieve rapid convergence of trajectory errors.

[0224] Alternatives to the adaptive law: The adaptive update law for super-twisted gain is not limited to mechanisms based on envelope variables and dynamic lower bounds. Other online update mechanisms, such as piecewise, continuous, projective, or with leakage terms, can also be used, as long as the gain can be adjusted online according to the sliding mode error or disturbance estimation state.

[0225] Thrust Allocation Alternatives: The process of converting six-degree-of-freedom generalized control forces into thruster thrust is not limited to a specific thrust allocation matrix and can be implemented using ordinary pseudo-inverse, weighted pseudo-inverse, quadratic programming, or other control allocation algorithms.

[0226] Platform Alternatives: This invention is not limited to the eight-thrust BlueROV Heavy platform, but is also applicable to other micro-sized omnidirectional AUVs, ROVs, or multi-thrust underwater robot platforms for point hovering, low-speed trajectory tracking, and operation in narrow waters.

Claims

1. A lumped disturbance observation and adaptive super-twisted sliding mode control method for micro-sized autonomous underwater robots, characterized in that, Includes the following steps: The current state and desired trajectory of the underwater robot are obtained, a six-degree-of-freedom dynamic model is established, and the model uncertainty and external disturbances are uniformly represented as lumped disturbances in the acceleration layer. An ultra-twisted nonlinear perturbation observer is used to continuously estimate the lumped perturbation of the acceleration layer online, and the lumped perturbation estimate is output. Construct a non-singular terminal sliding surface based on pose error and velocity error; Design a composite control law, which includes a nominal equivalent control term, a disturbance feedforward compensation term, and an adaptive super-twisted robust switching term; wherein, the nominal equivalent control term is used to cancel the known dynamics term, the disturbance feedforward compensation term is actively canceled based on the lumped disturbance estimate, and the adaptive super-twisted robust switching term is used to suppress observation residuals and unmodeled errors; The super-torsion control gain of the adaptive super-torsion robust switching term is adaptively adjusted online based on the sliding mode envelope variable, the disturbance envelope variable, and the dynamic lower bound. The composite control law is synthesized into a six-degree-of-freedom generalized control force, and then converted into an eight-thruster thrust command through a thrust allocation matrix to drive the underwater robot to perform omnidirectional motion control.

2. The method according to claim 1, characterized in that, The specific steps for unifying model uncertainties and external disturbances into a unified expression of lumped disturbances in the acceleration layer are as follows: Additional mass error, damping error, restoring force error, unmodeled dynamics, and external ocean current disturbances are uniformly converted into variables to be estimated in the acceleration layer, so that the controller can compensate for a single, uniform disturbance object.

3. The method according to claim 1, characterized in that, The specific steps for online continuous estimation of the lumped perturbation of the acceleration layer using a super-torsional nonlinear perturbation observer are as follows: An extended state containing lumped perturbations of velocity and acceleration layers is constructed. Under the assumption that the perturbation derivatives are bounded, a second-order sliding mode perturbation reconstruction is achieved by placing discontinuous terms in the integral channel through a super-twisted structure, thereby outputting a continuous estimate of the lumped perturbation of the acceleration layer.

4. The method according to claim 1, characterized in that, In the step of constructing a non-singular terminal sliding surface based on pose error and velocity error: The exponential parameters of the constructed non-singular terminal sliding surface satisfy the preset conditions, so that when the sliding surface approaches the origin, the power term in its derivative does not have a negative exponent, thereby avoiding the singular problem of the control input tending to infinity.

5. The method according to claim 1, characterized in that, The step of online adaptively adjusting the super-torsion control gain of the adaptive super-torsion robust switching term based on the sliding mode envelope variable, the perturbation envelope variable, and the dynamic lower bound specifically includes: The sliding mode error and the normalized disturbance estimate are filtered to generate sliding mode envelope variables and disturbance envelope variables; Based on the sliding mode envelope variable, the disturbance envelope variable, and the preset threshold, a dynamic lower bound for the control gain is generated and first-order smoothing is performed. The first-level gain is adjusted in segments according to the magnitude of the sliding mode error, and the upper and lower bounds of the first-level gain are limited by the projection operator, while the second-level gain is linked with the first-level gain. The super-twist control gain is increased when the sliding mode error is large or the disturbance is enhanced, and the super-twist control gain is decreased when the sliding mode error is reduced and the disturbance is weakened.

6. A control system for micro-sized autonomous underwater robots, characterized in that, include: The status acquisition module is used to acquire the current status and desired trajectory of the underwater robot; The disturbance modeling module is used to establish a six-degree-of-freedom dynamic model and to represent model uncertainties and external disturbances as lumped disturbances in the acceleration layer. The disturbance observation module is used to continuously estimate the lumped disturbance of the acceleration layer online using a super-torsional nonlinear disturbance observer and output the lumped disturbance estimate. The sliding surface construction module is used to construct non-singular terminal sliding surfaces based on pose and velocity errors. The composite control calculation module is used to calculate the nominal equivalent control term, the disturbance feedforward compensation term based on the lumped disturbance estimate, and the adaptive super-twisted robust switching term based on online adaptive adjustment of the super-twisted control gain; The thrust allocation module is used to convert the synthesized six-degree-of-freedom generalized control forces into eight-thrust commands through the thrust allocation matrix.

7. An electronic device, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method of any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, It stores a computer program thereon, which, when executed by a processor, implements the method of any one of claims 1 to 5.

Images