An intelligent marine robot and an adaptive underwater motion control method thereof

Abstract

The application provides a marine intelligent robot and an adaptive underwater motion control method thereof, and belongs to the technical field of intelligent robots. A first convergence time constant is used to recursively estimate a quasi-static parameter perturbation and generate a feedforward compensation force. Meanwhile, an integral term corresponding to the action direction of the feedforward compensation force in a basic control law is controlled to attenuate to zero. In parallel with a slow channel estimator, motion state data, environment perception data and an expected control force are input into a fast channel observer to estimate a lumped time-varying disturbance and generate a disturbance compensation force with a second convergence time constant. A steady-state water flow feedforward force is calculated according to the environment perception data. The feedforward compensation force, the disturbance compensation force, the steady-state water flow feedforward force and a feedback output of the basic control law are superimposed to generate control instructions for each propeller. The scheme of the application can classify and process multi-source disturbances with significant time scale differences in underwater robot motion control.

Description

Technical Field

[0001] This application relates to the field of intelligent robot technology, specifically to a marine intelligent robot and its adaptive underwater motion control method. Background Technology

[0002] In deep-sea exploration and seabed facility maintenance, the motion control performance of underwater robots directly affects the accuracy and efficiency of operations. Disturbances in the underwater environment have significant multi-timescale characteristics: on the one hand, buoyancy materials undergo volume compression under high pressure in deep water, and the shift of the center of mass caused by load changes causes quasi-static drift in net buoyancy and restoring torque, with the timescale of change typically on the order of minutes; on the other hand, sudden changes in water flow caused by near-seafloor turbulence and internal waves, as well as the unmodeled hydrodynamic coupling effect generated by the robot's own high angle of attack maneuvers, constitute time-varying disturbances on the order of seconds or even sub-seconds.

[0003] Currently, active disturbance rejection control (ADRC) technology has been widely applied in the field of underwater robotics because it does not rely on precise dynamic models. This type of method uses an extended state observer to uniformly estimate model uncertainties, external disturbances, and unmodeled dynamics into a total disturbance, and then compensates for it centrally using a feedforward approach. During the development of deep-sea robots, the inventors discovered that when the robot is in steady-state hovering or uniform cruising, quasi-static parameter perturbations are the main source of motion deviation. However, due to its high-frequency response characteristics, the extended state observer's estimation results for slowly changing quasi-static quantities are easily contaminated by measurement noise, resulting in insufficient estimation accuracy. If the observer bandwidth is reduced to adapt to the slowly changing characteristics, the response to rapidly changing disturbances such as sudden changes in water flow will show a significant lag. Furthermore, when the integral term of the basic control law and the feedforward compensation force act simultaneously on the same motion channel, if both compensate for quasi-static deviations, redundant compensation may occur, leading to overshoot. In prolonged steady-state conditions, this can also cause integral saturation, affecting control quality. Therefore, how to classify and process multi-source disturbances with significant differences in time scale in the motion control of underwater robots, so that quasi-static parameter perturbations and lumped time-varying disturbances are estimated and compensated at the best matching time scale, while avoiding the repetitive effect between compensation forces, has become a challenge for the industry. Summary of the Invention

[0004] Based on this, this application provides a marine intelligent robot and its adaptive underwater motion control method, which is used to classify and process multi-source disturbances with significant differences in time scale in the motion control of underwater robots.

[0005] In a first aspect, this application provides an adaptive underwater motion control method for a marine intelligent robot, comprising the following steps: Acquire robot motion state data and environmental perception data; When the robot is determined to meet the steady-state condition, the motion state data and the expected control force currently output by the basic control law are synchronously input into the slow channel estimator. The perturbation of the quasi-static parameter is recursively estimated using the first convergence time constant, and a feedforward compensation force is generated. At the same time, the integral term in the basic control law corresponding to the direction of action of the feedforward compensation force is controlled to decay toward zero. In parallel with the slow-channel estimator, the motion state data, the environmental perception data, and the desired control force are input into the fast-channel observer to estimate the lumped time-varying disturbance with the second convergence time constant and generate the disturbance compensation force, and to calculate the steady-state flow feedforward force based on the environmental perception data. The feedforward compensation force, the disturbance compensation force, the steady-state water flow feedforward force, and the feedback output of the basic control law are superimposed to generate control commands for each thruster.

[0006] The motion state data includes the robot's posture angle, angular velocity, linear velocity, and depth, while the environmental perception data includes water flow velocity and direction.

[0007] In a preferred embodiment, when the robot is determined to meet the steady-state condition, the motion state data and the desired control force currently output by the basic control law are synchronously input into the slow-channel estimator. The quasi-static parameter perturbation is recursively estimated using a first convergence time constant, and a feedforward compensation force is generated, including: Based on the depth change rate and attitude angular rate in the motion state data, determine whether the robot currently meets the steady-state conditions; When it is determined that the robot meets the steady-state condition, the motion state data and the desired control force are used as identification input data pairs, and the quasi-static parameter perturbation is recursively estimated using the first convergence time constant. Based on the perturbation of the quasi-static parameters obtained by recursive estimation, a feedforward compensation force is generated.

[0008] The steady-state condition is that the depth change rate is lower than a preset depth change rate threshold and the attitude angular rate is lower than a preset attitude angular rate threshold, and this state is maintained for a preset number of consecutive judgment cycles; the quasi-static parameter perturbation refers to the restoring force deviation experienced by the robot body that exceeds a preset limit on a time scale.

[0009] Furthermore, simultaneously controlling the integral term in the basic control law corresponding to the direction of the feedforward compensation force to decay towards zero includes: Based on the motion channel where the non-zero component of the feedforward compensation force is located, determine the target channel for which integral term attenuation needs to be performed; On the target channel, the current accumulated value of the corresponding integral term in the basic control law is decayed towards zero by either decaying cycle by cycle or being directly cleared to zero.

[0010] In a preferred embodiment, in parallel with the slow-channel estimator, the motion state data, the environmental perception data, and the desired control force are input into the fast-channel observer to estimate the lumped time-varying disturbance with a second convergence time constant and generate the disturbance compensation force, including: In parallel with the slow channel estimator, the desired control force is used as the system input and the motion state data is used as the system output observation. The nominal response channel of the fast channel observer is constructed by combining the robot's nominal hydrodynamic model, and the residual between the nominal motion output and the actual motion output is calculated. The residual is input into the fast-channel observer, and the lumped time-varying disturbance is recursively estimated using the second convergence time constant. The environmental perception data is used to help distinguish between the component of the lumped time-varying disturbance caused by water flow and the component caused by unmodeled hydrodynamic coupling effects. Based on the lumped time-varying disturbance obtained by recursion estimation, a disturbance compensation force is generated.

[0011] Wherein, the second convergence time constant is at least one order of magnitude smaller than the first convergence time constant.

[0012] Furthermore, calculating the steady-state water flow feedforward force based on the environmental perception data includes: Extract water flow velocity and direction from the environmental perception data; Then, by combining the robot's frontal area and water flow resistance characteristics, the steady-state force acting on the robot under the current water flow conditions is determined, which serves as the steady-state water flow feedforward force.

[0013] In a preferred embodiment, the basic control law is a PID-type feedback control law that includes a proportional term, an integral term, and a derivative term.

[0014] Secondly, this application provides a marine intelligent robot, which includes an adaptive underwater motion control unit, the adaptive underwater motion control unit comprising: The acquisition module is used to acquire the robot's motion state data and environmental perception data; The processing module is used to synchronously input the motion state data and the desired control force currently output by the basic control law into the slow channel estimator when it is determined that the robot meets the steady-state conditions. The quasi-static parameter perturbation is estimated recursively with the first convergence time constant and a feedforward compensation force is generated. At the same time, the integral term in the basic control law corresponding to the direction of action of the feedforward compensation force is controlled to decay toward zero. The processing module is also used in parallel with the slow channel estimator to input the motion state data, the environmental perception data and the desired control force into the fast channel observer, estimate the lumped time-varying disturbance amount with the second convergence time constant and generate the disturbance compensation force, and calculate the steady-state flow feedforward force based on the environmental perception data. The execution module is used to superimpose the feedforward compensation force, the disturbance compensation force, the steady-state water flow feedforward force, and the feedback output of the basic control law to generate control commands for each thruster.

[0015] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects: In this embodiment, a fast and slow dual-channel parallel architecture is used to distribute disturbances of different time scales to different channels for separate processing. The slow channel estimator recursively estimates the quasi-static parameter perturbation under steady-state conditions with a larger first convergence time constant, avoiding the contamination of slowly varying parameter estimation by high-frequency measurement noise. The fast channel observer quickly tracks the lumped time-varying disturbances with a smaller second convergence time constant, ensuring timely response to sudden changes in water flow and unmodeled dynamics. At the same time, the integral term is controlled to decay toward zero in the direction in which the feedforward compensation force is effective, so that the feedforward compensation force takes over the steady-state compensation function of the integral term, avoiding overshoot caused by repeated action of the two in the same channel, and also eliminating the risk of integral saturation under long-term steady state. In addition, the steady-state water flow feedforward force is directly calculated based on environmental perception data, so that the fast channel observer only needs to focus on the estimation of sudden changes in water flow and unmodeled hydrodynamic coupling effects, further improving the pertinence estimation targeting. Attached Figure Description

[0016] Figure 1 This is an exemplary flowchart of an adaptive underwater motion control method for a marine intelligent robot according to some embodiments of this application; Figure 2 This is a schematic diagram illustrating an application scenario of an underwater motion control data processing system according to some embodiments of this application; Figure 3 This is a schematic diagram of the process for calculating steady-state flow feedforward force according to some embodiments of this application; Figure 4 This is a schematic diagram of the structure of an adaptive underwater motion control unit according to some embodiments of this application; Figure 5 This is a schematic diagram of the structure of a computer device for implementing an adaptive underwater motion control method for a marine intelligent robot, according to some embodiments of this application. Detailed Implementation

[0017] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0018] refer to Figure 1 The figure is an exemplary flowchart of an adaptive underwater motion control method for a marine intelligent robot according to some embodiments of this application. The adaptive underwater motion control method for the marine intelligent robot mainly includes the following steps: In step 101, the robot's motion state data and environmental perception data are acquired.

[0019] In a preferred embodiment, acquiring the robot's motion state data and environmental perception data specifically involves: acquiring the three-axis attitude angles and angular velocities, current depth value, and three-axis linear velocities using the inertial measurement unit, depth gauge, and Doppler velocimeter mounted on the robot body, respectively. All of these sensors are time-stamped using the same clock source to achieve data time synchronization. Simultaneously, a flow velocity measurement device mounted on the robot body, such as an acoustic Doppler current profiler or an electromagnetic current meter, directly measures the relative water flow velocity vector at the robot's depth layer. This relative flow velocity vector is then converted into water flow velocity and direction in a geodetic coordinate system by combining the robot's current heading angle. Alternatively, if the robot is not equipped with the flow velocity measurement device, the water flow velocity and direction can be obtained by inverting the robot's position drift over time while hovering. Finally, the attitude angles, angular velocities, linear velocities, and depth are integrated into the motion state data, and the water flow velocity and direction are integrated into the environmental perception data.

[0020] In some embodiments, reference Figure 2 As shown in the figure, this figure is a schematic diagram of the application scenario of the underwater motion control data processing system shown in some embodiments of this application. The figure includes three main components: acquisition device, server and data storage device. The acquisition device is responsible for collecting the robot's motion state data and environmental perception data, and sending the collected motion state data and environmental perception data to the server through the communication network. The underwater motion control data processing system runs in the server, and the server stores the processing results in the data storage device and visualizes them.

[0021] In step 102, when it is determined that the robot meets the steady-state condition, the motion state data and the expected control force currently output by the basic control law are synchronously input into the slow channel estimator. The quasi-static parameter perturbation is recursively estimated using the first convergence time constant, and a feedforward compensation force is generated. At the same time, the integral term in the basic control law corresponding to the direction of action of the feedforward compensation force is controlled to decay toward zero.

[0022] The basic control law is a PID-type feedback control law that includes a proportional term, an integral term, and a derivative term; the steady-state condition is that the depth change rate is lower than a preset depth change rate threshold and the attitude angular rate is lower than a preset attitude angular rate threshold, and this state is maintained for a preset number of consecutive judgment cycles.

[0023] Optionally, when it is determined that the robot meets the steady-state condition, the motion state data and the desired control force currently output by the basic control law are synchronously input into the slow-channel estimator. The quasi-static parameter perturbation is recursively estimated using the first convergence time constant, and a feedforward compensation force is generated. Specifically, this includes: Based on the depth change rate and attitude angular rate in the motion state data, determine whether the robot currently meets the steady-state conditions; When it is determined that the robot meets the steady-state condition, the motion state data and the desired control force are used as identification input data pairs, and the quasi-static parameter perturbation is recursively estimated using the first convergence time constant. Based on the perturbation of the quasi-static parameters obtained by recursive estimation, a feedforward compensation force is generated.

[0024] Specifically, in this implementation, the robot's current steady-state condition is determined based on the depth change rate and attitude angular rate in the motion state data. Specifically, this involves extracting a depth sequence and an attitude angular rate sequence from the motion state data for several consecutive control cycles; performing first-order difference processing on the depth sequence to obtain the current depth change rate; and simultaneously performing first-order difference processing on the attitude angular rate sequence to obtain the current attitude angular rate. The depth change rate is then compared with a preset depth change rate threshold, and the attitude angular rate is compared with a preset attitude angular rate threshold. The depth change rate threshold and the attitude angular rate threshold are set based on the minimum disturbance that the robot's thrusters can overcome in hovering mode. The corresponding motion response level, for example, for a deep-sea robot operating at a depth of 3000 meters and requiring a hovering accuracy of ±0.5 meters, the depth change rate threshold can be set to 0.05 m / s, and the attitude angular rate threshold can be set to 0.5 degrees / s; when the depth change rate is lower than the depth change rate threshold and the attitude angular rate is lower than the attitude angular rate threshold, and this state is maintained for a preset number of consecutive judgment cycles, such as 10 consecutive control cycles, then it is determined that the robot currently meets the steady-state conditions; otherwise, it is determined that the steady-state conditions are not met. The purpose of introducing the number of consecutive judgment cycles is to avoid frequent jumps between meeting and not meeting the steady-state judgment results due to instantaneous sensor noise or brief water flow fluctuations.

[0025] When the robot is determined to meet the steady-state condition, the motion state data and the desired control force are used as an identification input data pair. The quasi-static parameter perturbation is recursively estimated using a first convergence time constant. That is, the identification input data pair is constructed, where the motion state data provides the robot's current actual motion response, and the desired control force provides the active control input currently applied to the robot. Together, they constitute an identification data pair reflecting the robot's dynamic input-output relationship. Using the desired control force as the system input and the motion state data as the system output observation, a recursive least squares identifier is constructed in conjunction with the robot's nominal rigid body dynamics model. In each control cycle, the recursive least squares identifier calculates the residual between the predicted output and the actual output based on the current identification data pair, and updates the estimated value of the quasi-static parameter perturbation along the direction of residual descent. The first... The convergence time constant is set accordingly using the forgetting factor of the recursive least squares identifier. The smaller the forgetting factor, the faster the identifier converges and the smaller the corresponding first convergence time constant. For example, when the first convergence time constant is 10 seconds, the corresponding forgetting factor can be set to about 0.9995 (calculated with a control period of 5 milliseconds), so that the identifier can complete the convergence estimation of the quasi-static parameter perturbation within a time window of about 10 seconds. The quasi-static parameter perturbation is specifically manifested as the net restoring force deviation of the robot's vertical channel and the static restoring torque deviation of the tilt / roll channel. During the identification process, the net restoring force deviation and the static restoring torque deviation are used as augmented state variables to be estimated and are jointly recursively solved with the robot's nominal dynamic parameters, thereby obtaining the estimated value of the quasi-static parameter perturbation with a unique convergence solution under the steady-state conditions.

[0026] Based on the quasi-static parameter perturbation obtained by recursion estimation, a feedforward compensation force is generated. Specifically, the vertical net restoring force deviation estimate and the static restoring torque deviation estimate are extracted from the quasi-static parameter perturbation obtained by recursion estimation. The vertical net restoring force deviation estimate reflects the deviation of the robot's current actual net buoyancy from the nominal net buoyancy, and the static restoring torque deviation estimate reflects the static torque imbalance caused by the offset of the robot's current actual center of mass position from the nominal center of mass position. The vertical net restoring force deviation estimate is directly used as the feedforward compensation force component of the vertical channel, and the static restoring torque deviation estimate is directly used as the feedforward compensation torque component of the tilt and pitch channels. At the same time, the feedforward compensation force components of the other motion channels are set to zero, thereby forming a complete six-degree-of-freedom feedforward compensation force vector.

[0027] Simultaneously controlling the decay of the integral term corresponding to the direction of the feedforward compensation force in the basic control law towards zero specifically includes: determining the target channel for which the integral term decay needs to be performed based on the motion channel where the non-zero component of the feedforward compensation force is located; on the target channel, decaying the current accumulated value of the corresponding integral term in the basic control law towards zero by decaying it periodically or by directly clearing it to zero.

[0028] More specifically, firstly, the six degrees of freedom components of the feedforward compensation force are traversed, and channels whose absolute values ​​are greater than a preset small threshold are marked as non-zero component channels. For example, the force component threshold is on the order of 10⁻³ Newtons, and the torque component threshold is on the order of 10⁻³ N·m. These non-zero component channels indicate that the slow channel estimator has already output an effective quasi-static compensation force to that channel, and the feedforward compensation force needs to take over the steady-state deviation compensation task for that channel. Then, the non-zero component channels are identified as target channels for which integral term attenuation needs to be performed. For each target channel, the current accumulated value of the integral term corresponding to that target channel in the basic control law is obtained, and a zero-value attenuation operation is performed on the current accumulated value. This zero-value attenuation operation can be implemented using a cycle-by-cycle exponential attenuation method, that is, in each control cycle, the current accumulated value is multiplied by a preset attenuation factor, such as 0.95, so that the accumulated value of the integral term is reduced as the feedforward compensation force continues to output... The value gradually approaches zero during the output period, or it can be achieved by direct zeroing. That is, during the control cycle when the slow channel estimator first outputs the effective feedforward compensation force, the cumulative value of the integral term corresponding to the target channel is directly set to zero. Among them, the decay transition is smoother when using the period-by-period exponential decay method, but the handover speed is slower. The handover speed is the fastest when using the direct zeroing method, but it may cause a brief step jump in the control quantity of the channel. The specific decay method can be selected according to the duration of the robot's current steady-state condition and the thrust change rate tolerance of the thruster. Through the above operation, the feedforward compensation force on the target channel gradually takes over or takes over the quasi-static deviation compensation function originally undertaken by the integral term, while the integral term gradually withdraws or withdraws immediately. Thus, the quasi-static compensation on the target channel is undertaken solely by the feedforward compensation force, and the integral term only retains the ability to supplement and adjust the remaining high-frequency residuals, avoiding the simultaneous action of both on the target channel and the resulting repeated compensation overshoot.

[0029] In step 103, in parallel with the slow-channel estimator, the motion state data, the environmental perception data, and the desired control force are input into the fast-channel observer to estimate the lumped time-varying disturbance with the second convergence time constant and generate the disturbance compensation force, and to calculate the steady-state flow feedforward force based on the environmental perception data.

[0030] In parallel with the slow-channel estimator, the motion state data, the environmental perception data, and the desired control force are input into the fast-channel observer to estimate the lumped time-varying disturbance with a second convergence time constant and generate the disturbance compensation force. This can be achieved through the following steps: In parallel with the slow channel estimator, the desired control force is used as the system input and the motion state data is used as the system output observation. The nominal response channel of the fast channel observer is constructed by combining the robot's nominal hydrodynamic model, and the residual between the nominal motion output and the actual motion output is calculated. The residual is input into the fast-channel observer, and the lumped time-varying disturbance is recursively estimated using the second convergence time constant. The environmental perception data is used to help distinguish between the component of the lumped time-varying disturbance caused by water flow and the component caused by unmodeled hydrodynamic coupling effects. Based on the lumped time-varying disturbance obtained by recursion estimation, a disturbance compensation force is generated.

[0031] In specific implementation, in parallel with the slow-channel estimator, the desired control force is used as the system input, and the motion state data is used as the system output observation. The nominal response channel of the fast-channel observer is constructed using the robot's nominal hydrodynamic model, and the residual between the nominal motion output and the actual motion output is calculated. That is, within each control cycle, the fast-channel observer and the slow-channel estimator start running simultaneously without waiting for each other. The fast-channel observer first obtains the desired control force output by the basic control law in the current control cycle, substitutes the desired control force into the pre-calibrated robot nominal hydrodynamic model, which includes the nominal values ​​of the robot's added mass coefficient, linear damping coefficient, and nonlinear damping coefficient. Numerical integration is used to calculate the theoretically expected nominal motion output that the robot should produce under the current desired control force. The nominal motion output includes nominal linear acceleration and nominal angular acceleration. Simultaneously, from the... The actual linear acceleration and actual angular acceleration generated by the robot in the current control cycle are extracted from the motion state data. The nominal linear acceleration is subtracted from the actual linear acceleration component by component, and the nominal angular acceleration is subtracted from the actual angular acceleration component by component to obtain a six-degree-of-freedom residual vector. The six-degree-of-freedom residual vector is the residual between the nominal motion output and the actual motion output. This residual reflects the motion deviation that cannot be explained by the robot's nominal dynamics model, including deviations caused by external environmental disturbances and deviations caused by hydrodynamic coupling effects not modeled in the robot's nominal hydrodynamic model. Since the fast-channel observer and the slow-channel estimator operate in parallel, the calculation process of the residual is not affected by the working state of the slow-channel estimator. Even during the control cycle in which the slow-channel estimator pauses parameter updates due to unmet steady-state conditions, the fast-channel observer still independently completes the residual calculation and continues to advance the subsequent disturbance estimation process.

[0032] The residual is input into the fast-channel observer, and the lumped time-varying disturbance is recursively estimated using the second convergence time constant. The environmental perception data is used to help distinguish between the component of the lumped time-varying disturbance caused by water flow and the component caused by unmodeled hydrodynamic coupling effects. Specifically, the residual is used as the driving signal input to the fast-channel observer, which employs an extended state observer structure. In each control cycle, based on the magnitude and direction of the input residual, the estimated value of the lumped time-varying disturbance is recursively updated using the observer bandwidth corresponding to the second convergence time constant. The second convergence time constant is much smaller than the first convergence time constant; for example, when the first convergence time constant is on the order of 10 seconds, the second convergence time constant is on the order of 0.5 seconds to 1 second, enabling the fast-channel observer to quickly track the transient changes of the lumped time-varying disturbance. This recursion... During the estimation process, the environmental perception data is simultaneously introduced as auxiliary discrimination information. Specifically, the rate of change of water flow velocity at the current moment is obtained from the environmental perception data. When the rate of change of water flow velocity exceeds a preset water flow mutation threshold, the increment of the estimated value of the lumped time-varying disturbance in the current control period is preferentially attributed to the high-frequency mutation disturbance component of the water flow. When the rate of change of water flow velocity is lower than the water flow mutation threshold but the residual still has a significant amplitude, the estimated value of the lumped time-varying disturbance in the current control period is attributed to the unmodeled hydrodynamic coupling effect component. The unmodeled hydrodynamic coupling effect component includes, for example, the cross-coupling damping force generated between the degrees of freedom of motion when the robot performs a high angle of attack maneuver. This makes the high-frequency mutation disturbance component of the water flow and the unmodeled hydrodynamic coupling effect component in the estimated value of the lumped time-varying disturbance output by the fast-channel observer have distinguishable physical meanings.

[0033] Based on the lumped time-varying disturbance obtained through recursive estimation, a disturbance compensation force is generated. Specifically, from the lumped time-varying disturbance estimate updated recursively during the current control cycle of the fast-channel observer, the disturbance force and torque components of six degrees of freedom are directly extracted. The disturbance force components are inverted to serve as the disturbance compensation force components for the corresponding translational channel, and the disturbance torque components are inverted to serve as the disturbance compensation torque components for the corresponding rotational channel, thus forming a complete six-degree-of-freedom disturbance compensation force vector for subsequent superposition steps. The physical meaning of the inversion operation lies in the disturbance... The dynamic compensation force should act in the opposite direction to the lumped time-varying disturbance to counteract the deviation effect of the lumped time-varying disturbance on the robot's motion. The disturbance compensation force is updated once in each control cycle, and the update frequency is consistent with the operating frequency of the fast-channel observer. This allows the disturbance compensation force to respond promptly to the instantaneous changes in water flow abrupt changes and unmodeled hydrodynamic coupling effects, providing the robot with high-frequency disturbance rejection capability. Meanwhile, the feedforward compensation force generated by the slow-channel estimator is updated at a slower pace under the steady-state conditions. The two complement each other on the time scale.

[0034] In a preferred embodiment, reference Figure 3 As shown in the figure, this is a schematic flowchart of calculating the steady-state water flow feedforward force in some embodiments of this application. In this embodiment, the calculation of the steady-state water flow feedforward force based on the environmental perception data specifically includes: In step 1031, the water flow velocity and direction are extracted from the environmental sensing data; In step 1032, the steady-state force acting on the robot under the current water flow conditions is determined by combining the robot's frontal area and water flow resistance characteristics, and is used as the steady-state water flow feedforward force.

[0035] In specific implementation, the water flow velocity and direction are extracted from the environmental perception data. Specifically, within each control cycle, the scalar value of the water flow velocity and the angle value of the water flow direction, after time synchronization processing, are read from the environmental perception data at the current moment. The scalar value of the water flow velocity represents the magnitude of the water flow relative to the geodetic coordinate system, and the angle value of the water flow direction represents the direction of the water flow relative to the geodetic coordinate system. This data is directly output by the flow velocity measurement device mounted on the robot or indirectly calculated by the position drift inversion algorithm. For the case where the flow velocity measurement device directly outputs the data, the scalar value of the water flow velocity and the angle value of the water flow direction are read directly. For the case of position drift inversion, the equivalent water flow velocity and equivalent water flow direction, which are obtained in advance by the robot's hovering drift analysis, are read from the environmental perception data as the water flow velocity and direction, thereby completing the extraction of the water flow velocity and direction.

[0036] Furthermore, combining the robot's frontal area and water flow resistance characteristics, the steady-state force acting on the robot under the current water flow conditions is determined as the steady-state water flow feedforward force. Specifically, the frontal area of ​​the robot in three orthogonal directions is obtained. This frontal area is pre-calibrated based on the robot's shape design data. For example, for a typical frame-type underwater robot, its frontal area is approximately 0.3 square meters, its lateral area is approximately 0.5 square meters, and its vertical area is approximately 0.4 square meters. Based on the water flow velocity and direction, combined with the robot's current heading angle, the water flow velocity vector in the geodetic coordinate system is projected onto the three axes of the robot's coordinate system to obtain the frontal velocity components for each axis. Then, for each axis, using the frontal area and the corresponding water flow resistance coefficient, according to the relationship that fluid resistance is proportional to the square of the frontal velocity... The system calculates the steady-state force components of the water flow acting on the robot along the axis. The water flow drag coefficients are pre-calibrated based on the robot's shape through pool towing tests or computational fluid dynamics simulations. For example, the forward drag coefficient is approximately 1.0 to 1.2, and the lateral drag coefficient is approximately 0.8 to 1.0. Finally, the steady-state force components of the water flow along the three axes are synthesized into a complete six-degree-of-freedom steady-state water flow feedforward force vector. The translational force component is directly composed of the drag components along the three axes, while the rotational torque component is calculated based on the cross product of the lever arm length from the robot's geometric center to the geometric center of each oncoming surface and the corresponding drag component. The steady-state water flow feedforward force is used in subsequent superposition stages to directly feedforward compensate for the continuous steady-state pushing effect of the water flow on the robot. This allows the fast-channel observer to focus only on estimating high-frequency abrupt disturbances in the water flow and unmodeled hydrodynamic coupling effects, thereby achieving frequency-band coordinated compensation of the water flow disturbances.

[0037] In step 104, the feedforward compensation force, the disturbance compensation force, the steady-state water flow feedforward force, and the feedback output of the basic control law are superimposed to generate control commands for each thruster.

[0038] Preferably, the control commands for each thruster are generated by superimposing the feedforward compensation force, the disturbance compensation force, the steady-state water flow feedforward force, and the feedback output of the basic control law. Specifically, this includes: The feedforward compensation force, the disturbance compensation force, the steady-state flow feedforward force, and the feedback output of the basic control law are vector-superimposed on six degrees of freedom to obtain the comprehensive control force vector. Based on the comprehensive control force vector, thrust is distributed according to the robot thruster configuration matrix, and control commands for each thruster are generated.

[0039] Specifically, the feedforward compensation force, the disturbance compensation force, the steady-state water flow feedforward force, and the feedback output of the basic control law are vector-superimposed on six degrees of freedom to obtain the comprehensive control force vector. This is achieved by: within each control cycle, acquiring the feedforward compensation force generated by the slow-channel estimator, the disturbance compensation force generated by the fast-channel observer, the steady-state water flow feedforward force calculated from the environmental perception data, and the feedback control force vector calculated and output by the basic control law in real time based on the motion state error. All four sets of force vectors are expressed in a six-degree-of-freedom form in the robot's body coordinate system. Each set of force vectors includes three translational force components along the sway axis, lateral sway axis, and vertical sway axis, as well as a force vector around the sway axis. The three torque components of the roll axis, pitch axis, and yaw axis are used. The feedforward compensation force, the disturbance compensation force, the steady-state flow feedforward force, and the feedback control force vector are algebraically added component by component in their respective six degrees of freedom to obtain the comprehensive control force vector. For channels that have performed integral term attenuation, the integral contribution of the corresponding channel in the feedback output of the basic control law has approached zero. The feedforward compensation force naturally fills the quasi-static compensation function of the channel, so that the superimposed comprehensive control force of the channel will not be overcompensated due to the simultaneous existence of the integral term and the feedforward compensation force. The final comprehensive control force vector includes four control contributions from feedback regulation, quasi-static compensation, high-frequency disturbance rejection compensation, and steady-state flow feedforward compensation.

[0040] Further, based on the integrated control force vector and combined with the robot thruster configuration matrix, thrust is allocated, and control commands for each thruster are generated. Specifically, a thruster configuration matrix is ​​pre-established based on the robot thruster's installation position, thrust axis direction, and rotation direction. Each column of the thruster configuration matrix corresponds to the six-degree-of-freedom generalized force component contributed by a thruster when outputting a unit thrust in the robot's body coordinate system. For example, for an underwater robot with four horizontal vector thrusters and two vertical thrusters, the thruster configuration matrix is ​​a six-row, six-column mapping matrix. In each control cycle, the pseudo-inverse matrix of the thruster configuration matrix is ​​multiplied by the integrated control force vector to map the six-degree-of-freedom force components and torque components contained in the integrated control force vector to the target thrust value of each thruster. If the number of robot thrusters is more than six... There are multiple feasible solutions for the degrees of freedom. In this case, an optimization criterion with the goal of minimizing the total power consumption of the thrusters is introduced in the pseudo-inverse solution to determine the unique solution. After obtaining the target thrust value of each thruster, the target thrust value of each thruster is converted into a pulse width modulation duty cycle or analog voltage control signal according to the pre-calibrated correspondence curve between thruster thrust and control signal. This signal is then sent to the thruster driver as the control command for each thruster. In this process, if the target thrust value of a thruster exceeds its thrust saturation upper limit, the target thrust value of that thruster is limited to the thrust saturation upper limit, and the excess is redistributed to other unsaturated thrusters according to the pseudo-inverse relationship of the thruster configuration matrix. Similarly, if the target thrust value is lower than the thrust saturation lower limit, the lower limit is limited and the residual is redistributed to ensure that the execution accuracy of the integrated control force vector is not weakened by the saturation of a single thruster.

[0041] Furthermore, in another aspect of this application, in some embodiments, this application provides a marine intelligent robot, which includes an adaptive underwater motion control unit, with reference to... Figure 4 The figure is a schematic diagram of the structure of an adaptive underwater motion control unit according to some embodiments of this application. The adaptive underwater motion control unit includes: an acquisition module 401, a processing module 402, and an execution module 403, which are described below: Acquisition module 401, in this application, is mainly used to acquire the robot's motion state data and environmental perception data; Processing module 402, in this application, is mainly used to input the motion state data and the expected control force currently output by the basic control law into the slow channel estimator when it is determined that the robot meets the steady state condition, and to recursively estimate the quasi-static parameter perturbation with the first convergence time constant and generate the feedforward compensation force, while controlling the integral term in the basic control law corresponding to the action direction of the feedforward compensation force to decay toward zero. The processing module 402 described in this application is also used in parallel with the slow channel estimator to input the motion state data, the environmental perception data and the desired control force into the fast channel observer, estimate the lumped time-varying disturbance amount with the second convergence time constant and generate the disturbance compensation force, and calculate the steady-state water flow feedforward force based on the environmental perception data. The execution module 403 in this application is mainly used to superimpose the feedforward compensation force, the disturbance compensation force, the steady-state water flow feedforward force and the feedback output of the basic control law to generate control commands for each thruster.

[0042] Each module in the aforementioned adaptive underwater motion control unit can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, so that the processor can call and execute the corresponding operations of each module.

[0043] In another embodiment, this application provides a computer device, which may be a server, and its internal structure diagram may be as follows. Figure 5 As shown, the computer device includes a processor, memory, and a network interface connected via a system bus. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores adaptive underwater motion control data for the marine intelligent robot. The network interface communicates with external terminals via a network connection. When the computer program is executed by the processor, it implements an adaptive underwater motion control method for a marine intelligent robot.

[0044] Those skilled in the art will understand that Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0045] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described embodiment of the adaptive underwater motion control method for a marine intelligent robot.

[0046] In one embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described embodiment of the adaptive underwater motion control method for a marine intelligent robot.

[0047] In one embodiment, a computer program product or computer program is provided, comprising computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the steps described in the above embodiment of the adaptive underwater motion control method for a marine intelligent robot.

[0048] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0049] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0050] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. An adaptive underwater motion control method for a marine intelligent robot, characterized in that, The method includes the following steps: Acquire robot motion state data and environmental perception data; When the robot is determined to meet the steady-state condition, the motion state data and the expected control force currently output by the basic control law are synchronously input into the slow channel estimator. The perturbation of the quasi-static parameter is recursively estimated using the first convergence time constant, and a feedforward compensation force is generated. At the same time, the integral term in the basic control law corresponding to the direction of action of the feedforward compensation force is controlled to decay toward zero. In parallel with the slow-channel estimator, the motion state data, the environmental perception data, and the desired control force are input into the fast-channel observer to estimate the lumped time-varying disturbance with the second convergence time constant and generate the disturbance compensation force, and to calculate the steady-state flow feedforward force based on the environmental perception data. The feedforward compensation force, the disturbance compensation force, the steady-state water flow feedforward force, and the feedback output of the basic control law are superimposed to generate control commands for each thruster.

2. The method as described in claim 1, characterized in that, The motion state data includes the robot's posture angle, angular velocity, linear velocity, and depth, while the environmental perception data includes water flow velocity and direction.

3. The method as described in claim 1, characterized in that, When the robot is determined to meet the steady-state condition, the motion state data and the desired control force currently output by the basic control law are synchronously input into the slow-channel estimator. The quasi-static parameter perturbation is recursively estimated using the first convergence time constant, and the feedforward compensation force is generated, including: Based on the depth change rate and attitude angular rate in the motion state data, determine whether the robot currently meets the steady-state conditions; When it is determined that the robot meets the steady-state condition, the motion state data and the desired control force are used as identification input data pairs, and the quasi-static parameter perturbation is recursively estimated using the first convergence time constant. Based on the perturbation of the quasi-static parameters obtained by recursive estimation, a feedforward compensation force is generated.

4. The method as described in claim 3, characterized in that, The steady-state condition is that the depth change rate is lower than a preset depth change rate threshold and the attitude angular rate is lower than a preset attitude angular rate threshold, and this state is maintained for a preset number of consecutive judgment cycles; the quasi-static parameter perturbation refers to the restoring force deviation experienced by the robot body that exceeds a preset limit on a time scale.

5. The method as described in claim 1, characterized in that, Simultaneously, the decay of the integral term in the basic control law corresponding to the direction of the feedforward compensation force toward zero includes: Based on the motion channel where the non-zero component of the feedforward compensation force is located, determine the target channel for which integral term attenuation needs to be performed; On the target channel, the current accumulated value of the corresponding integral term in the basic control law is decayed towards zero by either decaying cycle by cycle or being directly cleared to zero.

6. The method as described in claim 1, characterized in that, In parallel with the slow-channel estimator, the motion state data, the environmental perception data, and the desired control force are input into the fast-channel observer to estimate the lumped time-varying disturbance with a second convergence time constant and generate the disturbance compensation force, including: In parallel with the slow channel estimator, the desired control force is used as the system input and the motion state data is used as the system output observation. The nominal response channel of the fast channel observer is constructed by combining the robot's nominal hydrodynamic model, and the residual between the nominal motion output and the actual motion output is calculated. The residual is input into the fast-channel observer, and the lumped time-varying disturbance is recursively estimated using the second convergence time constant. The environmental perception data is used to help distinguish between the component of the lumped time-varying disturbance caused by water flow and the component caused by unmodeled hydrodynamic coupling effects. Based on the lumped time-varying disturbance obtained by recursion estimation, a disturbance compensation force is generated.

7. The method as described in claim 6, characterized in that, The second convergence time constant is at least one order of magnitude smaller than the first convergence time constant.

8. The method as described in claim 1, characterized in that, The calculation of steady-state water flow feedforward force based on the environmental perception data includes: Extract water flow velocity and direction from the environmental perception data; Then, by combining the robot's frontal area and water flow resistance characteristics, the steady-state force acting on the robot under the current water flow conditions is determined, which serves as the steady-state water flow feedforward force.

9. The method as described in claim 1, characterized in that, The basic control law is a PID-type feedback control law that includes proportional, integral, and derivative terms.

10. A marine intelligent robot, comprising an adaptive underwater motion control unit, characterized in that, The adaptive underwater motion control unit includes: The acquisition module is used to acquire the robot's motion state data and environmental perception data; The processing module is used to synchronously input the motion state data and the desired control force currently output by the basic control law into the slow channel estimator when it is determined that the robot meets the steady-state conditions. The quasi-static parameter perturbation is estimated recursively with the first convergence time constant and a feedforward compensation force is generated. At the same time, the integral term in the basic control law corresponding to the direction of action of the feedforward compensation force is controlled to decay toward zero. The processing module is also used in parallel with the slow channel estimator to input the motion state data, the environmental perception data and the desired control force into the fast channel observer, estimate the lumped time-varying disturbance amount with the second convergence time constant and generate the disturbance compensation force, and calculate the steady-state flow feedforward force based on the environmental perception data. The execution module is used to superimpose the feedforward compensation force, the disturbance compensation force, the steady-state water flow feedforward force, and the feedback output of the basic control law to generate control commands for each thruster.

Images