A coordinated control method for multiple underwater robot systems using distributed prediction

Abstract

The present invention discloses a coordinated control method for a multi-underwater robot system using distributed prediction, which relates to the field of robot technology. Step S1 is used to sense and collect various data of an underwater environment in real time, including water flow speed, direction, terrain features and obstacle positions. Step S2 is used to obtain a collision factor Pzyz of an obstacle in a travel route and an obstacle avoidance offset factor Pyyz, and the collision risk and avoidance requirements of the travel route are evaluated accordingly. Step S3 is used to dynamically correct an avoidance offset grade evaluation scheme of the robot's travel route based on an environmental impact coefficient Yxxs, and generate a collaborative correction grade evaluation scheme. Step S4 is used to sense the operating status in real time, and the system can continuously monitor the adjustment effect, and feedback information is fed back to step S2 for closed-loop adaptive adjustment, so as to significantly improve the detection efficiency and coverage, enhance the ability to cope with complex dynamic environments, and ensure the reliability and safety of task execution.

Description

Technical Field

[0001] The invention relates to the field of robot technology, and in particular to a coordinated control method for a multi-underwater robot system using distributed prediction. Background Art

[0002] With the global emphasis on marine resources and environmental protection, marine exploration and environmental monitoring tasks have become increasingly important. However, traditional single underwater robot systems have limited efficiency in the vast ocean and are difficult to meet the needs of large-scale exploration. A single robot can usually only cover a limited area, resulting in slow data collection and insufficient coverage, which cannot meet the extensive demand for marine data in scientific research, resource exploration, and environmental protection. In addition, the operation of a single robot requires frequent human intervention, making it difficult to achieve long-term, continuous autonomous work.

[0003] The underwater environment is extremely complex and uncertain, including factors such as changes in ocean currents, complex terrain, and various obstacles. Changes in ocean currents may cause the robot to deviate from the planned route, and complex terrain and unknown obstacles increase the difficulty of robot navigation. In this environment, it is difficult for a single robot to respond to all emergencies autonomously, and it is often necessary to frequently adjust the path and mission plan, which not only increases the complexity and execution time of the task, but may also lead to mission failure or damage to the robot.

[0004] In addition, due to the limited ability to identify possible environmental changes in advance and make corresponding adjustments, it is difficult for a single robot to work efficiently and accurately in a complex environment. Errors may occur in the collaborative work content or travel routes of multiple robots, further affecting the completion of the task. Traditional single-robot systems cannot respond and adjust quickly in a dynamically changing environment, and lack flexibility and adaptability. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention provides a coordinated control method for a multi-underwater robot system using distributed prediction, which solves the problems mentioned in the background technology.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: a coordinated control method for a multi-underwater robot system using distributed prediction, comprising the following steps:

[0007] S1. The robot uses the underwater detection sensor integrated in the robot to perceive the surrounding environment in real time, collect relevant data information including water flow speed, direction, terrain characteristics and obstacle location detection, and form an underwater real-time perception information group;

[0008] S2. Predict the obstacle position using a recursive neural network model based on the collected underwater real-time perception information group, and obtain: the route obstacle collision factor Pzyz and the obstacle avoidance offset factor Pyyz, and synchronously obtain the robot route collision level evaluation plan and the robot route avoidance offset level evaluation plan based on the route obstacle collision factor Pzyz and the obstacle avoidance offset factor Pyyz results after matching;

[0009] S3. According to the robot's route collision assessment scheme and the robot's route avoidance offset assessment scheme, the collected underwater real-time perception information group is evaluated for underwater environment, and the underwater environment impact travel coefficient Yxxs is obtained to correct the robot's route avoidance offset level assessment scheme and obtain a collaborative correction level assessment scheme.

[0010] S4. Coordinated control is performed on several underwater robot systems that perform tasks according to the content of the collaborative correction level assessment plan, including route deviation, planning and reformulation, and real-time perception of the operating status of the coordinated controlled underwater robots is fed back to step S2.

[0011] Preferably, the integrated underwater detection sensor includes sonar, multi-beam sonar, laser radar, current meter and inertial measurement unit, and the sonar is used to obtain the position and distance of the detected obstacle and generate the acoustic wave image of the surrounding environment; the multi-beam sonar is used to provide a high-resolution underwater topographic map and detect the seabed characteristics; the current meter is used to measure the water flow speed and direction to provide environmental dynamic information; the inertial measurement unit is used to detect the robot's posture, acceleration and angular velocity to assist navigation and stable control; and the integrated underwater detection sensor is used to obtain the underwater real-time perception information group;

[0012] Among them, the underwater real-time perception information group includes: robot coordinate position, water flow speed, distance from the bottom of the water, obstacle size and obstacle coordinate position.

[0013] Preferably, the collected underwater real-time perception information group is subjected to real-time normalization and standardization processing, and then the processed underwater real-time perception information group is subjected to a recursive neural network model to establish an underwater robot trajectory model, and after model training, the collision information between the underwater obstacle and the underwater robot trajectory is obtained, including the real-time coordinate position information of the robot: the machine coordinate r t (x r ,y r ,z r ) and the machine speed vector v r (v rx ,v ry ,v rz ) and the real-time coordinate position information of the obstacle: obstacle coordinate o t(x o ,y o ,z o ) and the obstacle velocity vector v o (v ox ,v oy ,v oz ), and then according to the time series: time t, perform trajectory simulation training steps on the collision information to obtain: the route obstacle collision factor Pzyz and the obstacle avoidance offset factor Pyyz, and synchronously obtain the robot route collision level evaluation plan and the robot route avoidance offset level evaluation plan after matching according to the route obstacle collision factor Pzyz and the obstacle avoidance offset factor Pyyz results.

[0014] Preferably, the route obstacle collision factor Pzyz is obtained through the following trajectory simulation training steps:

[0015] Step 1: Pass The calculation formula obtains the relative position of the robot and the obstacle at time t: Relative position ;

[0016] Step 2: Pass The calculation formula obtains the relative speed between the robot and the obstacle: relative speed ;

[0017] Step 3: By relative position and relative speed Use vector modulus The formula calculates the Euclidean distance to obtain the predicted closest distance between the robot and the obstacle: Predicted closest distance ;

[0018] Step 4: Get the predicted shortest distance by solving The smallest time ,pass Calculation formula to obtain the minimum distance between the robot and the obstacle ;

[0019] Step 5: Based on the minimum distance Minimum collision distance with the preset machine: Minimum collision distance Calculate the following formula , obtain: the obstacle collision factor Pzyz of the travel route.

[0020] Preferably, the robot travel route collision level assessment scheme is obtained by matching in the following manner:

[0021] The collision factor of the obstacle on the route Pzyz is less than 0, and the first-level collision assessment result is obtained. The current robot's route trajectory will not collide with the obstacle position or moving trajectory;

[0022] The collision factor of the obstacle on the route Pzyz≥1, and the second-level collision assessment result is obtained. The current robot's route trajectory will collide with the obstacle position or moving trajectory.

[0023] Preferably, the obstacle avoidance offset factor Pyyz is obtained through the following trajectory simulation training steps:

[0024] Avoidance step 1: When the robot's route collision level assessment scheme is a second-level collision assessment result, execute avoidance step 2;

[0025] Avoidance step 2: When the robot avoids obstacles, it should Move in the direction perpendicular to the direction, and the avoidance direction vector should be consistent with the relative speed. =v r -v o The vector is perpendicular, and then we get: avoidance direction vector d: , the relative speed is represented by the avoidance direction vector d The components of the vector on the y-axis and x-axis, where d y and d x They represent the y-axis moving speed and x-axis moving speed of the robot relative to the obstacle respectively;

[0026] Among them, the relative speed =v r -v o Specifically: ;

[0027] Avoidance step 3: Normalize the avoidance direction vector d , and based on the minimum distance obtained , calculate the offset : ;

[0028] Avoidance step 4: Based on the offset Calculate the obstacle avoidance offset factor Pyyz: .

[0029] Preferably, the robot's route avoidance deviation level evaluation scheme is obtained by matching in the following manner:

[0030] If the obstacle avoidance offset factor Pyyz is less than 1, the robot's avoidance route does not meet the safe avoidance effect. Increase the robot's avoidance angle and related avoidance adjustment parameters by 5%, and execute the trajectory simulation training step of the obstacle avoidance offset factor Pyyz again. Repeat until the obstacle avoidance offset factor Pyyz is greater than or equal to 1 to achieve the effect that the avoidance route can be safely avoided.

[0031] The obstacle avoidance offset factor Pyyz is less than 1, and the robot's avoidance route satisfies the effect of safe avoidance, and the robot's avoidance angle and related avoidance adjustment parameters are not adjusted.

[0032] Preferably, when the robot travel route collision assessment scheme and the robot travel route avoidance deviation assessment scheme are executed, the real-time fluctuation perception of the underwater environment assessment is triggered, the water flow velocity and the distance from the water bottom are formatted and normalized, and the water flow velocity value Slz and the distance from the water bottom value Sdz are obtained, and they are associated with the time t to obtain the underwater environment impact travel coefficient Yxxs;

[0033] The underwater environment influence travel coefficient Yxxs is obtained by the following calculation formula:

[0034] ;

[0035] Wherein, Yxxs represents the underwater environment impact travel coefficient, t represents the total number of detections within a time period, i represents the i-th detection within a time period, Slz represents the water velocity value, and Sdz represents the distance from the water bottom. By counting the absolute value information of the difference between the water velocity value Slz and the distance value Sdz from the water bottom recorded in the adjacent time periods, the underwater environment fluctuation information expression is obtained: the underwater environment impact travel coefficient Yxxs.

[0036] Preferably, the collaborative correction level evaluation scheme is obtained by the following matching method:

[0037] The underwater environment impact travel coefficient Yxxs is less than 1, and the first collaborative correction level is obtained, and no collaborative correction control is performed on the robot executing the robot travel route collision assessment plan and the robot travel route avoidance deviation assessment plan;

[0038] The underwater environment affects the travel coefficient Yxxs≥1, and the second collaborative correction level is obtained. The robots that execute the robot travel route collision assessment plan and the robot travel route avoidance offset assessment plan are collaboratively corrected and controlled, including adjusting the robot avoidance angle and increasing the relevant avoidance adjustment parameter values ​​to 110%.

[0039] Preferably, the detection tasks of the robots executing the robot route collision assessment plan, the robot route avoidance offset assessment plan and the collaborative correction level assessment plan are synchronously sent to the adjacent robots executing the detection tasks, and the detection areas are proportionally allocated, including allocating the detection tasks originally undertaken by the avoidance robot to other robots and adjusting their routes. At the same time, the operating status and environmental data of the adjusted robots are collected in real time, and the perception data is fed back to the S2 step for further optimization of the model and robot adjustment.

[0040] The present invention provides a method for coordinated control of multiple underwater robot systems using distributed prediction, which has the following beneficial effects:

[0041] (1) Step S1 is used to sense and collect various underwater environment data in real time, including water flow speed, direction, terrain features and obstacle positions. Step S2 is used to obtain the obstacle collision factor Pzyz and obstacle avoidance offset factor Pyyz of the route, and the collision risk and avoidance requirements of the route are evaluated accordingly. Step S3 is used to dynamically correct the avoidance offset level evaluation scheme of the robot's route based on the environmental impact coefficient Yxxs, and generate a collaborative correction level evaluation scheme. Step S4 is used to sense the operating status in real time. The system can continuously monitor the adjustment effect, and use the feedback information for further optimization and adjustment, as well as feedback to step S2 for closed-loop adaptive adjustment. Through effective communication and data sharing, each robot can understand the status of the entire system in real time, make the best decision, and ensure the smooth progress and efficient completion of the task.

[0042] (2) By obtaining the obstacle collision factor Pzyz and the obstacle avoidance offset factor Pyyz of the route, and obtaining the robot's route collision level assessment plan and the robot's route avoidance offset level assessment plan after matching, accurate obstacle collision prediction is achieved, providing quantitative risk assessment and multi-level collision risk assessment, helping the robot to make efficient decisions and optimize the path.

[0043] (3) By calculating the underwater environment impact travel coefficient Yxxs and obtaining the collaborative correction level evaluation scheme, the collaborative working ability of the multi-robot system is improved. Through dynamic task allocation and adjustment, the overall task completion efficiency and quality are improved, as well as the system's adaptability and stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is a schematic diagram of the steps of a coordinated control method for a multi-underwater robot system using distributed prediction according to the present invention. DETAILED DESCRIPTION

[0045] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0046] Example 1

[0047] The present invention provides a method for coordinated control of multiple underwater robot systems using distributed prediction. Figure 1 , including the following steps:

[0048] S1. The robot uses the underwater detection sensor integrated in the robot to perceive the surrounding environment in real time, collect relevant data information including water flow speed, direction, terrain characteristics and obstacle location detection, and form an underwater real-time perception information group;

[0049] S2. Predict the obstacle position using a recursive neural network model based on the collected underwater real-time perception information group, and obtain: the route obstacle collision factor Pzyz and the obstacle avoidance offset factor Pyyz, and synchronously obtain the robot route collision level evaluation plan and the robot route avoidance offset level evaluation plan based on the route obstacle collision factor Pzyz and the obstacle avoidance offset factor Pyyz results after matching;

[0050] S3. According to the robot's route collision assessment scheme and the robot's route avoidance offset assessment scheme, the collected underwater real-time perception information group is evaluated for underwater environment, and the underwater environment impact travel coefficient Yxxs is obtained to correct the robot's route avoidance offset level assessment scheme and obtain a collaborative correction level assessment scheme.

[0051] S4. Coordinated control is performed on several underwater robot systems that perform tasks according to the content of the collaborative correction level assessment plan, including route deviation, planning and reformulation, and real-time perception of the operating status of the coordinated controlled underwater robots is fed back to step S2.

[0052] In this embodiment, various data of the underwater environment, including water flow speed, direction, terrain features and obstacle positions, are sensed and collected in real time through step S1, and the obstacle collision factor Pzyz and obstacle avoidance offset factor Pyyz of the route are obtained through step S2, and the collision risk and avoidance requirements of the route are evaluated accordingly. The avoidance offset level evaluation scheme of the robot's route is dynamically corrected based on the environmental impact coefficient Yxxs through step S3, and a collaborative correction level evaluation scheme is generated. The operating status is sensed in real time through step S4, and the system can continuously monitor the adjustment effect, and the feedback information is used for further optimization and adjustment, as well as feedback to step S2 for closed-loop adaptive adjustment, so as to significantly improve the detection efficiency and coverage, enhance the ability to cope with complex dynamic environments, and ensure the reliability and safety of task execution. This method overcomes the shortcomings of a single robot system, and realizes efficient task allocation and execution through collaborative work and dynamic adjustment.

[0053] Example 2

[0054] This embodiment is explained in Example 1, please refer to Figure 1 Specifically: the integrated underwater detection sensors include sonar, multi-beam sonar, lidar, current meter and inertial measurement unit. The sonar is used to obtain the position and distance of the detected obstacles and generate acoustic wave images of the surrounding environment; the multi-beam sonar is used to provide high-resolution underwater topographic maps and detect seabed features; the current meter measures the speed and direction of the water flow and provides dynamic environmental information; the inertial measurement unit detects the robot's posture, acceleration and angular velocity to assist navigation and stable control; and the integrated underwater detection sensor is used to obtain underwater real-time perception information groups;

[0055] Among them, the underwater real-time perception information group includes: robot coordinate position, water flow speed, distance from the bottom of the water, obstacle size and obstacle coordinate position.

[0056] According to the collected underwater real-time perception information group, real-time normalization and standardization are performed, and then the underwater robot trajectory model is established using a recursive neural network model for the processed underwater real-time perception information group. After model training, the collision information between the underwater obstacles and the underwater robot trajectory is obtained, including the real-time coordinate position information of the robot: the machine coordinate r t (x r ,y r ,z r ) and the machine speed vector v r (v rx ,v ry ,v rz ) and the real-time coordinate position information of the obstacle: obstacle coordinate o t (x o ,yo ,z o ) and the obstacle velocity vector v o (v ox ,v oy ,v oz ), and then according to the time series: time t, perform trajectory simulation training steps on the collision information to obtain: the route obstacle collision factor Pzyz and the obstacle avoidance offset factor Pyyz, and synchronously obtain the robot route collision level evaluation plan and the robot route avoidance offset level evaluation plan after matching according to the route obstacle collision factor Pzyz and the obstacle avoidance offset factor Pyyz results.

[0057] Example 3

[0058] This embodiment is explained in Example 1, please refer to Figure 1 Specifically, the obstacle collision factor Pzyz of the travel route is obtained through the following trajectory simulation training steps:

[0059] Step 1: Pass The calculation formula obtains the relative position of the robot and the obstacle at time t: Relative position ;

[0060] Step 2: Pass The calculation formula obtains the relative speed between the robot and the obstacle: relative speed ;

[0061] Step 3: By relative position and relative speed Use vector modulus The formula calculates the Euclidean distance to obtain the predicted closest distance between the robot and the obstacle: Predicted closest distance ;

[0062] Step 4: Get the predicted shortest distance by solving The smallest time ,pass Calculation formula to obtain the minimum distance between the robot and the obstacle ;

[0063] Step 5: Based on the minimum distance Minimum collision distance with the preset machine: Minimum collision distance Calculate the following formula , obtain: the obstacle collision factor Pzyz of the travel route.

[0064] The robot travel route collision level assessment scheme is obtained by matching in the following way:

[0065] The collision factor of the obstacle on the route Pzyz is less than 0, and the first-level collision assessment result is obtained. The current robot's route trajectory will not collide with the obstacle position or moving trajectory;

[0066] The collision factor of the obstacle on the route Pzyz≥1, and the second-level collision assessment result is obtained. The current robot's route trajectory will collide with the obstacle position or moving trajectory.

[0067] The obstacle avoidance offset factor Pyyz is obtained through the following trajectory simulation training steps:

[0068] Avoidance step 1: When the robot's route collision level assessment scheme is a second-level collision assessment result, execute avoidance step 2;

[0069] Avoidance step 2: When the robot avoids obstacles, it should Move in the direction perpendicular to the direction, and the avoidance direction vector should be consistent with the relative speed. =v r -v o The vector is perpendicular, and then we get: avoidance direction vector d: , the relative speed is represented by the avoidance direction vector d The components of the vector on the y-axis and x-axis, where d y and d x They represent the y-axis moving speed and x-axis moving speed of the robot relative to the obstacle respectively;

[0070] Among them, the relative speed =v r -v o Specifically: ;

[0071] Avoidance step 3: Normalize the avoidance direction vector d , and based on the minimum distance obtained , calculate the offset : ;

[0072] Avoidance step 4: Based on the offset Calculate the obstacle avoidance offset factor Pyyz: .

[0073] The robot route avoidance deviation level evaluation scheme is obtained by matching in the following way:

[0074] If the obstacle avoidance offset factor Pyyz is less than 1, the robot's avoidance route does not meet the safe avoidance effect. Increase the robot's avoidance angle and related avoidance adjustment parameters by 5%, and execute the trajectory simulation training step of the obstacle avoidance offset factor Pyyz again. Repeat until the obstacle avoidance offset factor Pyyz is greater than or equal to 1 to achieve the effect that the avoidance route can be safely avoided.

[0075] The obstacle avoidance offset factor Pyyz is less than 1, and the robot's avoidance route satisfies the effect of safe avoidance, and the robot's avoidance angle and related avoidance adjustment parameters are not adjusted.

[0076] In this embodiment, accurate obstacle collision prediction is achieved by obtaining the route obstacle collision factor Pzyz and the obstacle avoidance offset factor Pyyz, and obtaining the robot route collision level assessment plan and the robot route avoidance offset level assessment plan content after matching, providing quantitative risk assessment and multi-level collision risk assessment, and helping the robot to make efficient decisions and optimize the path.

[0077] Example 4

[0078] This embodiment is explained in Example 1, please refer to Figure 1 Specifically: when the robot's route collision assessment scheme and the robot's route avoidance offset assessment scheme are executed, the real-time fluctuation perception of the underwater environment assessment is triggered, and the water flow velocity and the distance from the bottom of the water are formatted and normalized to obtain: the water flow velocity value Slz and the distance from the bottom of the water value Sdz, and are associated with the time t to obtain: the underwater environment impact travel coefficient Yxxs;

[0079] The underwater environment influence travel coefficient Yxxs is obtained by the following calculation formula:

[0080] ;

[0081] Wherein, Yxxs represents the underwater environment impact travel coefficient, t represents the total number of detections within a time period, i represents the i-th detection within a time period, Slz represents the water velocity value, and Sdz represents the distance from the water bottom. By counting the absolute value information of the difference between the water velocity value Slz and the distance value Sdz from the water bottom recorded in the adjacent time periods, the underwater environment fluctuation information expression is obtained: the underwater environment impact travel coefficient Yxxs.

[0082] The collaborative correction level evaluation scheme is obtained by the following matching method:

[0083] The underwater environment impact travel coefficient Yxxs is less than 1, and the first collaborative correction level is obtained, and no collaborative correction control is performed on the robot executing the robot travel route collision assessment plan and the robot travel route avoidance deviation assessment plan;

[0084] The underwater environment affects the travel coefficient Yxxs≥1, and the second collaborative correction level is obtained. The robots that execute the robot travel route collision assessment plan and the robot travel route avoidance offset assessment plan are collaboratively corrected and controlled, including adjusting the robot avoidance angle and increasing the relevant avoidance adjustment parameter values ​​to 110%.

[0085] The detection tasks of the robots that execute the robot route collision assessment plan, the robot route avoidance offset assessment plan and the collaborative correction level assessment plan are synchronously sent to the adjacent robots that execute the detection tasks, and the detection areas are proportionally allocated, including allocating the detection tasks originally undertaken by the avoidance robot to other robots and adjusting their routes. At the same time, the operating status and environmental data of the adjusted robots are collected in real time, and the perception data is fed back to the S2 step for further optimization of the model and robot adjustment.

[0086] In this embodiment, the collaborative working ability of the multi-robot system is improved by calculating the underwater environment impact travel coefficient Yxxs and obtaining the collaborative correction level evaluation scheme. Through dynamic task allocation and adjustment, the overall task completion efficiency and quality are improved, and the system's adaptability and stability are improved.

[0087] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A coordinated control method for a multi-underwater robot system using distributed prediction, characterized in that: The following steps are involved: S1. The robot uses the underwater detection sensor integrated in the robot to perceive the surrounding environment in real time, collect relevant data information including water flow speed, direction, terrain characteristics and obstacle location detection, and form an underwater real-time perception information group; S2. Predict the obstacle position using a recursive neural network model based on the collected underwater real-time perception information group, and obtain: the route obstacle collision factor Pzyz and the obstacle avoidance offset factor Pyyz, and synchronously obtain the robot route collision level evaluation plan and the robot route avoidance offset level evaluation plan based on the route obstacle collision factor Pzyz and the obstacle avoidance offset factor Pyyz results after matching; The obstacle collision factor Pzyz of the travel route is obtained through the following trajectory simulation training steps: Step 1: Pass The calculation formula obtains the relative position of the robot and the obstacle at time t: Relative position ; Step 2: Pass The calculation formula obtains the relative speed between the robot and the obstacle: relative speed ; Step 3: By relative position and relative speed Use vector modulus The formula calculates the Euclidean distance to obtain the predicted closest distance between the robot and the obstacle: Predicted closest distance ; Step 4: Get the predicted shortest distance by solving The smallest time ,pass Calculation formula to obtain the minimum distance between the robot and the obstacle ; Step 5: Based on the minimum distance Minimum collision distance with the preset machine: Minimum collision distance Calculate the following formula , obtain: the obstacle collision factor Pzyz of the travel route; The obstacle avoidance offset factor Pyyz is obtained through the following trajectory simulation training steps: Avoidance step 1: When the robot's route collision level assessment scheme is a second-level collision assessment result, execute avoidance step 2; Avoidance step 2: When the robot avoids obstacles, it should Move in the direction perpendicular to the direction, and the avoidance direction vector should be consistent with the relative speed. =v r -v o The vector is perpendicular, and then we get: avoidance direction vector d: , the relative speed is represented by the avoidance direction vector d The components of the vector on the y-axis and x-axis, where d y and d x They represent the y-axis moving speed and x-axis moving speed of the robot relative to the obstacle respectively; Among them, the relative speed =v r -v o Specifically: ; Avoidance step 3: Normalize the avoidance direction vector d , and based on the minimum distance obtained , calculate the offset : ; Avoidance step 4: Based on the offset Calculate the obstacle avoidance offset factor Pyyz: ; S3. According to the robot's route collision assessment scheme and the robot's route avoidance offset assessment scheme, the collected underwater real-time perception information group is evaluated for underwater environment, and the underwater environment impact travel coefficient Yxxs is obtained to correct the robot's route avoidance offset level assessment scheme and obtain a collaborative correction level assessment scheme. When the robot's route collision assessment scheme and the robot's route avoidance offset assessment scheme are executed, the real-time fluctuation perception of the underwater environment assessment is triggered, and the water flow velocity and the distance from the bottom of the water are formatted and normalized to obtain: the water flow velocity value Slz and the distance from the bottom of the water value Sdz, and then associated with the time t to obtain: the underwater environment impact travel coefficient Yxxs; The underwater environment influence travel coefficient Yxxs is obtained by the following calculation formula: ； In the formula, Yxxs represents the underwater environment impact travel coefficient, t represents the total number of detections within a time period, i represents the i-th detection within a time period, Slz represents the water flow velocity value, and Sdz represents the distance from the bottom of the water. By counting the absolute value information of the difference between the water flow velocity value Slz and the distance value Sdz from the bottom of the water recorded in the adjacent time period, the underwater environment fluctuation information is obtained: the underwater environment impact travel coefficient Yxxs; S4. Coordinated control is performed on several underwater robot systems that perform tasks according to the content of the collaborative correction level assessment plan, including route deviation, planning and reformulation, and real-time perception of the operating status of the coordinated controlled underwater robots is fed back to step S2.

2. The method for coordinated control of multiple underwater robot systems using distributed prediction according to claim 1, characterized in that: The integrated underwater detection sensors include sonar, multi-beam sonar, lidar, current meter and inertial measurement unit. Sonar is used to obtain the position and distance of detected obstacles and generate acoustic wave images of the surrounding environment; multi-beam sonar is used to provide high-resolution underwater topographic maps and detect seabed features; current meters are used to measure water flow speed and direction to provide environmental dynamic information; inertial measurement units are used to detect the robot's posture, acceleration and angular velocity to assist navigation and stable control; and integrated underwater detection sensors are used to obtain real-time underwater perception information groups; Among them, the underwater real-time perception information group includes: robot coordinate position, water flow speed, distance from the bottom of the water, obstacle size and obstacle coordinate position.

3. The method for coordinated control of multiple underwater robot systems using distributed prediction according to claim 1, characterized in that: According to the collected underwater real-time perception information group, real-time normalization and standardization are performed, and then the underwater robot trajectory model is established using a recursive neural network model for the processed underwater real-time perception information group. After model training, the collision information between the underwater obstacles and the underwater robot trajectory is obtained, including the real-time coordinate position information of the robot: the machine coordinate r t (x r ,y r ,z r ) and the machine speed vector v r (v rx ,v ry ,v rz ) and the real-time coordinate position information of the obstacle: obstacle coordinate o t (x o ,y o ,z o ) and the obstacle velocity vector v o (v ox ,v oy ,v oz ), and then according to the time series: time t, perform trajectory simulation training steps on the collision information to obtain: the route obstacle collision factor Pzyz and the obstacle avoidance offset factor Pyyz, and synchronously obtain the robot route collision level evaluation plan and the robot route avoidance offset level evaluation plan after matching according to the route obstacle collision factor Pzyz and the obstacle avoidance offset factor Pyyz results.

4. The method for coordinated control of multiple underwater robot systems using distributed prediction according to claim 1, characterized in that: The robot travel route collision level assessment scheme is obtained by matching in the following way: The collision factor of the obstacle on the route Pzyz is less than 0, and the first-level collision assessment result is obtained. The current robot's route trajectory will not collide with the obstacle position or moving trajectory; The collision factor of the obstacle on the route Pzyz≥1, and the second-level collision assessment result is obtained. The current robot's route trajectory will collide with the obstacle position or moving trajectory.

5. The method for coordinated control of multiple underwater robot systems using distributed prediction according to claim 1, characterized in that: The robot route avoidance deviation level evaluation scheme is obtained by matching in the following way: If the obstacle avoidance offset factor Pyyz is less than 1, the robot's avoidance route does not meet the safe avoidance effect. Increase the robot's avoidance angle and related avoidance adjustment parameters by 5%, and execute the trajectory simulation training step of the obstacle avoidance offset factor Pyyz again. Repeat until the obstacle avoidance offset factor Pyyz is greater than or equal to 1 to achieve the effect that the avoidance route can be safely avoided. The obstacle avoidance offset factor Pyyz is less than 1, and the robot's avoidance route satisfies the effect of safe avoidance, and the robot's avoidance angle and related avoidance adjustment parameters are not adjusted.

6. The method for coordinated control of multiple underwater robot systems using distributed prediction according to claim 1, characterized in that: The collaborative correction level evaluation scheme is obtained by the following matching method: The underwater environment impact travel coefficient Yxxs is less than 1, and the first collaborative correction level is obtained, and no collaborative correction control is performed on the robot executing the robot travel route collision assessment plan and the robot travel route avoidance deviation assessment plan; The underwater environment affects the travel coefficient Yxxs≥1, and the second collaborative correction level is obtained. The robots that execute the robot travel route collision assessment plan and the robot travel route avoidance offset assessment plan are collaboratively corrected and controlled, including adjusting the robot avoidance angle and increasing the relevant avoidance adjustment parameter values ​​to 110%.

7. The method for coordinated control of multiple underwater robot systems using distributed prediction according to claim 1, characterized in that: The detection tasks of the robots that execute the robot route collision assessment plan, the robot route avoidance offset assessment plan and the collaborative correction level assessment plan are synchronously sent to the adjacent robots that execute the detection tasks, and the detection areas are proportionally allocated, including allocating the detection tasks originally undertaken by the avoidance robot to other robots and adjusting their routes. At the same time, the operating status and environmental data of the adjusted robots are collected in real time, and the perception data is fed back to the S2 step for further optimization of the model and robot adjustment.

Images