Multi-working-condition-oriented controller online self-optimization system and method for underwater robot

Abstract

The invention discloses a multi-working-condition-oriented controller online self-optimization system and a multi-working-condition-oriented controller online self-optimization method for an underwater robot. The motion control experiment of the underwater robot is guided through a method based on multi-algorithm fusion, optimized parameters are provided, the precise tracking capability of a three-dimensional complex track under multiple working conditions can be improved, the online rapid debugging capability is achieved, the adaptation work of the underwater robot and the controller is rapidly completed, the experiment progress is accelerated, and the working efficiency is improved. The underwater robot can execute tasks as soon as possible, the problems that an underwater robot controller is poor in fixed parameter adaptability and cannot process response speed, control precision, meet the real-time requirement and the like are solved, meanwhile, the navigation state of the underwater robot can be learned and recorded, and the method has the advantage of being high in universality. And guidance is provided for the debugging work of the underwater robot with the same mother type.

Description

Technical Field

[0001] This invention belongs to the field of underwater robot technology, specifically a system and method for online self-optimization of controllers for underwater robots under multiple working conditions. Background Technology

[0002] Underwater robots are crucial tools for humanity to understand, explore, and develop the ocean, and are also a high-tech means for countries to safeguard national security and achieve sustainable development. Underwater robots play a vital role in underwater equipment maintenance, seabed survey and assessment, ocean search and rescue, and anti-submarine warfare. The various functions of underwater robots rely on a variety of onboard devices; however, the underwater environment is complex and the types of tasks are diverse, requiring underwater robots to adapt to various operating conditions depending on the task. Therefore, how to efficiently conduct multi-condition debugging experiments on multiple underwater robots to ensure their successful task execution is one of the important research issues.

[0003] The complexity of the deep-sea environment and the diversity of underwater robot types present challenges to equipment testing before underwater robots can perform tasks. The strongly coupled and nonlinear dynamic characteristics of underwater robots' six degrees of freedom (DOF), along with the impact of environmental disturbances (ocean currents, load changes), place extremely high demands on the robustness and adaptability of the control system. Traditional control schemes suffer from the following difficulties: poor coordination of cascaded control, limitations of single adaptive algorithms, weak robustness to environmental disturbances, and the inability of a single algorithm to balance the contradictions of "response speed, control accuracy, and computational cost." Furthermore, underwater robots require high-level navigation behavior during autonomous operations.

[0004] Therefore, in order to ensure that underwater robots can successfully complete their tasks and solve the problem of high-precision control of underwater robots under multiple working conditions, it is of great significance to accelerate the optimization of underwater robot controllers. There is an urgent need for a multi-cascaded controller parameter collaborative adaptive adjustment system with multi-algorithm fusion and disturbance resistance. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an online self-optimization system and method for underwater robot controllers operating under multiple conditions. Targeting the six-degree-of-freedom coupling characteristics, this invention designs a three-level regulating controller through multi-algorithm fusion to collaboratively adjust parameters, solving the problem that traditional fixed parameters cannot adapt to changing operating conditions. This accelerates underwater robot debugging and experimentation, while also demonstrating strong versatility, adapting to different types of underwater robots, and facilitating future upgrades and modifications.

[0006] The technical solution adopted by the present invention to achieve the above objectives is as follows:

[0007] An online self-optimization system for the controller of an underwater robot for multiple working conditions includes multi-source sensors and a controller mounted on the underwater robot;

[0008] The multi-source sensor is used to collect underwater environmental data and its own navigation data in real time according to the controller instructions during the underwater robot's navigation, and then send the data to the controller.

[0009] The controller has a built-in control backend and a frontend interface. The frontend interface is used for human-computer interaction to collect algorithm parameters and control commands input by the user, as well as to display online self-optimization results. The control backend is equipped with a data processing module, a multi-algorithm fusion decision module, and a multi-controller execution module. Based on the user's instructions, it processes and fuses the data uploaded by the multi-source sensors to generate control decisions. Finally, the calculated control commands for the propeller and rudder are output to the actuators, which adjust the robot's underwater posture to complete online self-optimization.

[0010] The multi-source sensors include, but are not limited to, integrated inertial navigation, depth gauge, and Doppler velocimeter, which collect underwater robot pose, velocity, and environmental data in real time, as well as various equipment parameters of the underwater robot, thereby obtaining underwater operating conditions.

[0011] The data processing module includes a filtering unit, a fusion unit, and a calculation unit, used to filter, fuse, and remove outliers from the underwater robot's position and attitude data, as well as calculate the stable time period, average value, and variance data.

[0012] The multi-algorithm fusion decision module includes:

[0013] The fuzzy logic unit outputs the initial adjustment amount Δk for the PID parameters of the heading controller, depth controller, and speed controller based on fuzzy rules. p fla Δk i fla Δk d fla ;

[0014] RBF neural network units are used to build and train the RBF-NN network structure, enabling it to output parameter corrections Δk for the heading controller, depth controller, and velocity controller. p nn Δk i nn Δk d nn ;

[0015] The model references the adaptive unit, calculates the system error e based on the second-order linear system model and adaptive law, and adjusts the PID parameters Δk of the heading controller, depth controller, and speed controller. p mrac Δk i mrac Δk d mrac ;

[0016] The fusion strategy unit fuses the above three parameters to calculate the final PID parameters: the heading controller k p new , the depth controller k i new , the speed controller k d new ;

[0017] Δk p [[ID=1--6]] p fla p +Δk nn p +Δk mrac

[0018] k p new p =k old p +Δk i

[0019] Similarly, calculate k new d , k new p .

[0020] The fuzzy logic unit executes the following method steps:

[0021] Define the input variables: the current error e of the controller, the error change rate ec; the output variables: the preliminary adjustment amounts Δk fla i , Δk fla d , Δk fla p ;

[0022] Fuzzy rule base:

[0023] If e is "positive large" and ec is "positive large", then Δk fla i is "positive large", Δk fla d is "negative large", Δk fla p is "medium";

[0024] If e is "small" and ec is "small", then Δk fla i is "small", Δk fla d is "small", Δk fla p is "small";

[0025] Reasoning and Defuzzification: The Mamdani reasoning method and the centroid method are used for fuzzification, and the output Δk is obtained. p fla Δk i fla Δk d fla .

[0026] The RBF neural network unit performs the following method steps:

[0027] Inputs: Controller current error e, error change rate ec, system state (velocity, heading angle, depth); Output: Parameter correction Δk p nn Δk i nn Δk d nn ;

[0028] RBF-NN network structure: 3 nodes in the input layer, 10 nodes in the hidden layer (Gaussian kernel function), and 3 nodes in the output layer; Training: Using simulation experimental data, the weights of the hidden layers are adjusted through the backpropagation algorithm to obtain the ideal model.

[0029] The model reference adaptive unit performs the following method steps:

[0030] Establish a second-order linear system model:

[0031]

[0032] Where ω n ζ is the natural frequency, and ζ is the damping ratio. The damping ratio is set according to the desired response, with overshoot <10% and settling time <2s.

[0033] Adaptive law: based on the actual output of the underwater robot's corresponding controller. Compared with the reference model output The error e= - Adjust the PID parameters:

[0034] Δk p mrac =Δk p0 mrac +e

[0035] Where Δk p0 mrac These are the initial parameters.

[0036] The multi-controller execution module includes a heading controller unit, a speed controller unit, and a depth controller unit, used to execute the PID parameters k of their respective heading controller, depth controller, and speed controller. p new k i new k d new The system calculates control and adjustment commands for the propeller and rudder based on the underwater robot's position and attitude information, and outputs control signals for the propeller and rudder to achieve online self-optimization of the robot's underwater posture.

[0037] An online self-optimization method for the controller of an underwater robot under multiple operating conditions includes the following steps:

[0038] S1. The front-end interface collects algorithm parameters and control commands input by the user through human-computer interaction and sends them to the control backend.

[0039] S2. The control backend controls the multi-source sensors to collect data according to user instructions;

[0040] S3. The control backend processes and fuses the received data to generate PID parameters for the heading controller, depth controller, and speed controller. Finally, the calculated control commands for the propeller and rudder are output as control decisions to the actuator, which adjusts the robot's underwater posture to complete online self-optimization.

[0041] S4. The control backend simultaneously builds a twin model and visualizes the calculation results and twin simulation process on the front-end interface.

[0042] The parameters that the user inputs to the front-end interface include:

[0043] Based on the water depth, altitude, latitude and longitude, set navigation parameters including but not limited to depth, altitude, speed, position and direction;

[0044] Based on the underwater robot's pressure resistance, actuator layout, and hydrodynamic coefficients, set navigation parameters including but not limited to depth, height, speed, position, and direction;

[0045] Based on the economic speed, maximum speed, and control precision required for the mission, set navigation parameters including but not limited to depth, altitude, speed, position, and direction.

[0046] The present invention has the following beneficial effects and advantages:

[0047] 1. The controller parameter online self-optimization system of the present invention can achieve cascaded cooperative adaptive control. Targeting the six-degree-of-freedom coupling characteristics, a three-stage regulating controller is designed with coordinated parameter adjustment, solving the problem that traditional fixed parameters cannot adapt to changes in operating conditions.

[0048] 2. This invention achieves a balance between multiple algorithms, enabling it to handle nonlinearity, approximate dynamics, ensure tracking performance, and balance response speed and control accuracy, thus meeting real-time requirements.

[0049] 3. The main design element of this invention is an underwater robot, which is highly versatile and adaptable to different types of underwater robots such as ROVs and AUVs. Migration is only required by adjusting the dynamic model parameters. It can continue to be used for future upgrades and modifications to underwater robots. Attached Figure Description

[0050] Figure 1 This is a flowchart of the method of the present invention.

[0051] Figure 2 This is a flowchart of a self-optimizing system.

[0052] Figure 3 This is a diagram of the underwater robot self-optimization system modules and units. Detailed Implementation

[0053] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Many specific details are set forth in the following description to provide a thorough understanding of the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar modifications without departing from the spirit of the invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0054] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.

[0055] like Figure 1 As shown, this invention provides an online self-optimization method for the controller of an underwater robot oriented towards multiple working conditions, comprising the following steps:

[0056] An online self-optimization method for the controller of an underwater robot under multiple operating conditions includes the following steps:

[0057] Step 1: Assemble the underwater robot in a pool, lake, or sea area that can ensure the underwater robot can complete the experiment, and then assemble the underwater robot.

[0058] Step 2: Based on the characteristics of the water area and the required debugging functions of the underwater robot, design navigation parameters including but not limited to depth, height, speed, position, and direction. Turn on the main control software, set the navigation parameters, and put the underwater robot into the preparation navigation state.

[0059] Step 3: Activate the self-optimization system and enter the initial values ​​of each parameter calculated based on previous navigation data.

[0060] Step 4: Turn on all equipment and sensors. The underwater robot enters the navigation phase. The main control software module executes the navigation task. The data processing module filters, fuses, and calculates the data error and error change rate of each control loop from the sensor data and transmits it to the self-optimization system.

[0061] Step 5: During navigation, the self-optimization system determines whether the parameters are appropriate by receiving the error and error change rate data of the control loop, and modifies the parameters online through the RBF neural network unit and the model reference adaptive unit.

[0062] Step Six: Once the underwater robot completes the navigation phase, upload all data and analyze and verify the results obtained from the collected data. If the navigation status does not meet the expectations, repeat steps two through six. Modify parameters offline using step three until all debugging tasks are completed.

[0063] A method for online self-optimization of the controller of an underwater robot for multiple working conditions was developed.

[0064] Furthermore, step 1 specifically includes the following steps:

[0065] Step 1.1: Confirm and record the depth topographic map and latitude / longitude region of the underwater robot physical verification platform;

[0066] Step 1.2: Confirm that all units and modules of the underwater robot have been assembled and tested successfully;

[0067] Furthermore, step 2 specifically includes the following steps:

[0068] Step 2.1: Based on the characteristics of the water area (depth, altitude, latitude and longitude), design and set navigation parameters including but not limited to depth, altitude, speed, position, and direction;

[0069] Step 2.2: Based on the characteristics of the underwater robot (pressure resistance, actuator arrangement, hydrodynamic coefficient), design navigation parameters including but not limited to depth, height, speed, position, and direction, and set the navigation parameters;

[0070] Step 2.3: Based on mission requirements (economic speed, maximum speed, control accuracy), design and set navigation parameters including but not limited to depth, altitude, speed, position, and direction.

[0071] Furthermore, step 3 specifically includes the following steps:

[0072] Step 3.1: Develop self-optimizing system software;

[0073] Step 3.2: Start the self-optimizing system software;

[0074] Step 3.3: Design according to the task requirements, navigation parameters, test waters, and previous navigation data, and fill in the initial values of each parameter in sequence;

[0075] Further, step 4 specifically includes the following steps:

[0076] Step 4.1: Turn on the main control software, set including but not limited to navigation mode, target depth, speed, heading, navigation time, etc., and turn on all devices;

[0077] Step 4.2: Collect sensor data, including but not limited to devices such as inertial navigation, Doppler velocimeter, depth gauge, etc.

[0078] Step 4.3: Filter and fuse the sensor data to eliminate outliers.

[0079] Step 4.4: Calculate the sensor data and output including but not limited to steady-state time, average value, variance, etc.

[0080] Step five specifically includes the following steps:

[0081] Step 5.1: Feed the sensor data to the self-optimizing system software for fuzzy logic adaptation (FLA) - to process non-linear optimization; (taking the heading controller as an example, the depth controller and speed controller are the same)

[0082] Define the input variables: current error e ψ , error change rate ec ψ ;

[0083] Output variable: preliminary adjustment amount Δk of PID parameters p fla , Δk i fla , Δk d fla .

[0084] Fuzzy rule base:

[0085] If e ψ is "positive large" and ec ψ is "positive large", then Δk p fla is "positive large", Δk i <00​​​​​​​​​​​p fla For "small", Δk i fla For "small", Δk d fla It is "small".

[0087] Reasoning and defuzzification: The Mamdani reasoning method and the centroid method are used for defuzzification, and the output Δk is obtained. p fla Parameters such as these.

[0088] Step 5.2: Sensor data is fed into the self-optimizing system software for RBF-NN correction—approximating the unknown and dynamically constructing the RBF-NN (taking the heading controller as an example, the same applies to the depth controller and velocity controller).

[0089] The input is "current error e" ψ Error change rate (ec) ψ "System status (velocity, pitch angle)"

[0090] The output is "parameter correction amount Δk" p nn Δk i nn Δk d nn .

[0091] Network structure: 3 nodes in the input layer, 10 nodes in the hidden layer (Gaussian kernel function), and 3 nodes in the output layer; Training: Using simulation experimental data, the weights of the hidden layers are adjusted through the backpropagation algorithm to approximate the unmodeled dynamics of the system (fluid dynamic changes).

[0092] Step 5.3: Sensor data is fed into the self-optimizing system software for MRAC tracking—ensuring response performance (taking the heading controller as an example; the same applies to the depth controller and velocity controller).

[0093] Design model (second-order linear system):

[0094]

[0095] Where ω n (Natural frequency) and ζ (damping ratio) are set according to the desired response (overshoot <10%, settling time <2s).

[0096] Adaptive law: based on the actual output of the underwater robot's corresponding controller. Compared with the reference model output error e ψ = - Adjust the PID parameters:

[0097] Δkp mrac =Δk p0 mrac +e ψ

[0098] Where Δk p0 mrac These are the initial parameters.

[0099] Step 5.4: The final PID parameters of the fusion strategy are adjusted as follows:

[0100] Δk p =Δk p fla +Δk p nn +Δk p mrac

[0101] k p new =k p old +Δk p

[0102] Similarly, calculate k i new k d new By combining the nonlinear processing capabilities of FLA, the dynamic approximation capabilities of RBF-NN, and the tracking performance of MRAC, a balance is struck between response speed and accuracy.

[0103] Step 5.5: Continuously adjust parameters online based on sensor data during the voyage until the voyage ends.

[0104] Furthermore, based on the recorded navigation data, the results are analyzed and verified. If the test objective is not achieved, steps two to five are repeated.

[0105] Step six specifically includes the following steps:

[0106] Step 6.1: The underwater robot completes the navigation phase and uploads all data, including but not limited to speed, target speed, depth, target depth, heading, and target heading.

[0107] Step 6.2: Adjust the controller parameters offline based on the mean and root mean square deviation of the stable segment of the speed and target speed.

[0108] Step 6.3: Adjust the controller parameters offline based on the mean and root mean square deviation of the stable segment of the depth and target depth.

[0109] Step 6.4: Adjust the controller parameters offline based on the mean and root mean square deviation of the stable sections of the heading and target heading.

[0110] like Figure 1-2 As shown, the present invention discloses an online self-optimization system and method for an underwater robot controller oriented to multiple operating conditions, comprising: an online self-optimization method for an underwater robot controller oriented to multiple operating conditions, and an online self-optimization system for an underwater robot controller oriented to multiple operating conditions.

[0111] like Figure 3 As shown, an underwater robot should include, but is not limited to, the following modules and units.

[0112] 1. Sensor module: including but not limited to inertial navigation, Doppler velocimeter, depth gauge and other equipment, used to collect position and attitude data of underwater robot.

[0113] 2. Data processing module: including but not limited to data filtering unit, data fusion unit, data calculation unit, etc., used to filter, fuse and calculate new data such as errors of underwater robot position and attitude data.

[0114] 3. Multi-algorithm fusion decision module: including fuzzy logic unit, RFB neural network unit, and model reference adaptive unit, used to calculate controller parameters based on data such as the underwater robot's position and attitude.

[0115] 4. Multi-controller execution module: including a heading controller unit, a depth controller unit, and a speed controller unit, used to calculate the control output of the propeller and rudder based on the controller parameters and the underwater robot's position and attitude information.

[0116] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should be considered within the scope of protection of the present invention.

Claims

1. An online self-optimization system for the controller of an underwater robot oriented to multiple working conditions, characterized in that, This includes multi-source sensors and controllers carried by underwater robots; The multi-source sensor is used to collect underwater environmental data and its own navigation data in real time according to the controller instructions during the underwater robot's navigation, and then send the data to the controller. The controller has a built-in control backend and a frontend interface. The frontend interface is used for human-computer interaction to collect algorithm parameters and control commands input by the user, as well as to display online self-optimization results. The control backend is equipped with a data processing module, a multi-algorithm fusion decision module, and a multi-controller execution module. Based on the user's instructions, it processes and fuses the data uploaded by the multi-source sensors to generate control decisions. Finally, the calculated control commands for the propeller and rudder are output to the actuators, which adjust the robot's underwater posture to complete online self-optimization.

2. The online self-optimization system for an underwater robot controller oriented towards multiple working conditions according to claim 1, characterized in that, The multi-source sensors include, but are not limited to, integrated inertial navigation, depth gauge, and Doppler velocimeter, which collect underwater robot pose, velocity, and environmental data in real time, as well as various equipment parameters of the underwater robot, thereby obtaining underwater operating conditions.

3. The online self-optimization system for an underwater robot controller oriented towards multiple working conditions according to claim 1, characterized in that, The data processing module includes a filtering unit, a fusion unit, and a calculation unit, used to filter, fuse, and remove outliers from the underwater robot's position and attitude data, as well as calculate the stable time period, average value, and variance data.

4. The online self-optimization system for an underwater robot controller oriented towards multiple working conditions according to claim 1, characterized in that, The multi-algorithm fusion decision module includes: The fuzzy logic unit outputs the initial adjustment amount Δk for the PID parameters of the heading controller, depth controller, and speed controller based on fuzzy rules. p fla Δk i fla Δk d fla ; RBF neural network units are used to build and train the RBF-NN network structure, enabling it to output parameter corrections Δk for the heading controller, depth controller, and velocity controller. p nn Δk i nn Δk d nn ; The model references the adaptive unit, calculates the system error e based on the second-order linear system model and adaptive law, and adjusts the PID parameters Δk of the heading controller, depth controller, and speed controller. p mrac Δk i mrac Δk d mrac ; The fusion strategy unit combines the above three parameters to calculate the final PID parameters: (Forward controller k) p new Depth controller k i new Speed ​​controller k d new ; Δk p =Δk p fla +Δk p nn +Δk p mrac k p new =k p old +Δk p Similarly, calculate k i new k d new .

5. The online self-optimization system for an underwater robot controller oriented towards multiple working conditions according to claim 4, characterized in that, The fuzzy logic unit performs the following method steps: Define input variables: the current error e and the rate of change of error ec of the controller; output variables: the initial adjustment amount Δk of the PID parameters of the heading controller, depth controller, and speed controller. p fla Δk i fla Δk d fla ; Fuzzy rule base: If e is "Positive Big" and ec is "Positive Big", then Δk p fla is "Positive Big", Δk i fla is "Negative Big", Δk d fla is "Medium"; If e is "small" and ec is "small", then Δk p fla For "small", Δk i fla For "small", Δk d fla For "small"; Reasoning and Defuzzification: The Mamdani reasoning method and the centroid method are used for fuzzification, and the output Δk is obtained. p fla Δk i fla Δk d fla .

6. The online self-optimization system for an underwater robot controller oriented towards multiple working conditions according to claim 4, characterized in that, The RBF neural network unit performs the following method steps: Inputs: Controller current error e, error change rate ec, system state (velocity, heading angle, depth); Output: Parameter correction Δk p nn Δk i nn Δk d nn ; RBF-NN network structure: 3 nodes in the input layer, a Gaussian kernel with 10 hidden nodes, and a 3-node output layer; Training: Using simulation data, the weights of the hidden layers are adjusted through the backpropagation algorithm to obtain the ideal model.

7. The online self-optimization system for an underwater robot controller oriented towards multiple working conditions according to claim 4, characterized in that, The model reference adaptive unit performs the following method steps: Establish a second-order linear system model: Where ω n ζ is the natural frequency, and ζ is the damping ratio. The damping ratio is set according to the desired response, with overshoot <10% and settling time <2s. Adaptive law: based on the actual output of the underwater robot's corresponding controller. Compared with the reference model output The error e= - Adjust the PID parameters: Δk p mrac =Δk p0 mrac +e Where Δk p0 mrac These are the initial parameters.

8. An online self-optimization system for an underwater robot controller oriented towards multiple working conditions according to claim 1, characterized in that, The multi-controller execution module includes a heading controller unit, a speed controller unit, and a depth controller unit, used to execute the PID parameters k of their respective heading controller, depth controller, and speed controller. p new k i new k d new The system calculates control and adjustment commands for the propeller and rudder based on the underwater robot's position and attitude information, and outputs control signals for the propeller and rudder to achieve online self-optimization of the robot's underwater posture.

9. A method for online self-optimization of an underwater robot controller oriented to multiple working conditions according to claim 1, characterized in that, Includes the following steps: S1. The front-end interface collects algorithm parameters and control commands input by the user through human-computer interaction and sends them to the control backend. S2. The control backend controls the multi-source sensors to collect data according to user instructions; S3. The control backend processes and fuses the received data to generate PID parameters for the heading controller, depth controller, and speed controller. Finally, the calculated control commands for the propeller and rudder are output as control decisions to the actuator, which adjusts the robot's underwater posture to complete online self-optimization. S4. The control backend simultaneously builds a twin model and visualizes the calculation results and twin simulation process on the front-end interface.

10. A method for online self-optimization of an underwater robot controller oriented to multiple operating conditions according to claim 9, characterized in that, The parameters that the user inputs to the front-end interface include: Based on the water depth, altitude, latitude and longitude, set navigation parameters including but not limited to depth, altitude, speed, position and direction; Based on the underwater robot's pressure resistance, actuator layout, and hydrodynamic coefficients, set navigation parameters including but not limited to depth, height, speed, position, and direction; Based on the economic speed, maximum speed, and control precision required for the mission, set navigation parameters including but not limited to depth, altitude, speed, position, and direction.

Images