AUV obstacle avoidance methods, systems, and devices based on vector fields and fuzzy rules

Abstract

This invention discloses an AUV obstacle avoidance method, system, computer device, and storage medium based on vector fields and fuzzy rules. By acquiring a sonar dataset, an obstacle recognition model is constructed using a YOLOv4 network and trained on the sonar dataset. The trained obstacle recognition model identifies obstacles and constructs an obstacle environment. A vector field algorithm and fuzzy rules are used to calculate the AUV's desired heading, guiding the AUV to avoid detected obstacles and move towards its destination. In addition to using the vector field algorithm for obstacle avoidance path planning, fuzzy rules for safety distance, unit momentum vector, and roll-up torque are introduced to ensure the safety and effectiveness of local obstacle avoidance, preventing the AUV from getting stuck in local terrain. Compared to traditional obstacle avoidance methods, this invention not only enables AUVs to effectively avoid unknown obstacles but also demonstrates good adaptability and obstacle avoidance capabilities for complex terrains such as parallel obstacles and square well obstacles.

Description

Technical Field

[0001] This invention relates to the field of autonomous underwater vehicle (AUV) technology, and in particular to an AUV obstacle avoidance method, system, computer device, and readable storage medium based on vector fields and fuzzy rules. Background Technology

[0002] Autonomous Underwater Vehicles (AUVs) are underwater vehicles integrating sensing, communication, navigation, control, power, and propulsion technologies. They possess advantages such as wide operating range, high maneuverability, high safety, and intelligence, and have gradually become important tools for completing various underwater missions. Path planning is a crucial component of the AUV's automatic control system, serving as the foundation for autonomous underwater navigation. Path planning includes global path planning based on prior environmental information and local path planning in uncertain environments based on sensor information. Typically, the marine environment is unknown or partially unknown. When an AUV navigates along a globally planned path, it inevitably encounters previously unknown obstacles such as ships, marine life, and protruding rocks on the seabed. Therefore, local path planning, i.e., real-time obstacle avoidance, is a problem that must be considered in the design of the AUV control system, and it has even greater application value in real-world marine scenarios.

[0003] Traditional obstacle avoidance methods for unknown obstacles include fast-search random tree algorithms, artificial potential field algorithms, and fuzzy logic algorithms. Unlike other robotic platforms on land or in the air, these methods primarily use sonar sensors to obtain the real-time distance and angle of surrounding obstacles and utilize local dynamic path planning techniques to complete obstacle avoidance tasks, effectively improving the autonomy of AUVs. However, these traditional algorithms still suffer from problems such as being prone to getting trapped in local minima and unstable and low-quality obstacle avoidance paths: the obstacle avoidance path of the fast-search random tree algorithm is composed of nodes, and the path length and smoothness are not optimal, resulting in poor path quality; in the artificial potential field algorithm, complex environments are prone to many locations where the net force is zero, i.e., local minima of the potential field. Because the net force is zero, the AUV will always be in a state of immobility, leading to path planning failure; the fuzzy logic algorithm relies on the researcher's experience to set, and its fuzzy rules may fail in complex environments, leading to decision errors and AUV obstacle avoidance failure.

[0004] When AUVs perform missions, the underwater environment is quite complex, so a real-time obstacle avoidance method that is effective against unknown obstacles and different terrains is needed. Summary of the Invention

[0005] This invention provides an AUV obstacle avoidance method, system, computer device, and readable storage medium based on vector fields and fuzzy rules to solve the technical problems of traditional obstacle avoidance methods for unknown obstacles, such as easy getting trapped in local minima and unstable and low-quality obstacle avoidance paths. This enables AUVs to adapt to various terrains and have more stable and smooth obstacle avoidance trajectories.

[0006] To address the aforementioned technical problems, in a first aspect, embodiments of the present invention provide an AUV obstacle avoidance method based on vector fields and fuzzy rules, the method comprising:

[0007] Acquire a sonar dataset, which includes forward-looking sonar images;

[0008] An obstacle recognition model was constructed using a YOLOv4 network, and the sonar dataset was input into the obstacle recognition model for training.

[0009] The trained obstacle recognition model identifies real-time forward-looking sonar images to determine whether there are obstacles around the AUV. If there are obstacles, the location of the obstacles is output, and the obstacle environment is constructed using a matrix grid method based on the detected obstacle locations.

[0010] The AUV's position is obtained using an AUV positioning system, and the desired heading of the AUV is calculated using a vector field algorithm and fuzzy rules based on the AUV's position, destination position, and the positions of obstacles in the obstacle environment.

[0011] The AUV is controlled to navigate according to the desired course, and a dead reckoning algorithm based on the AUV dynamics model is continuously used to determine whether the AUV has reached the destination. If the destination has been reached, the AUV navigation mission is completed.

[0012] In a further embodiment, the method for calculating the desired heading of an AUV using a vector field algorithm and fuzzy rules includes:

[0013] Calculate the azimuth angle between the AUV and the destination based on the AUV's current position and destination position:

[0014] θ end =atan2[(Y end -Y AUV ),(X end -X AUV )],

[0015] Among them, X AUV and Y AUV These are the X and Y position coordinates of the guide point based on the AUV, respectively. end and Y end These are the X and Y coordinates of the target endpoint, respectively.

[0016] Calculate the AUV's attraction vector based on the azimuth angle:

[0017]

[0018] Based on the detected obstacle location and AUV location, calculate the distance from the detected obstacle to the AUV:

[0019]

[0020] Among them, X obstacle and Y obstacle The X and Y coordinates of the detected obstacle;

[0021] The AUV's repulsion vector is calculated based on the AUV's location, the obstacle's location, and the detected distance from the obstacle to the AUV:

[0022]

[0023] Where, d zero This represents the maximum detection range of the forward-looking sonar.

[0024] Calculate the composite vector of the AUV based on its repulsion and attraction vectors:

[0025]

[0026] Where, k a and k r These are the attraction gain and repulsion gain constants;

[0027] Calculate the desired heading angle of the AUV based on the synthesized vector:

[0028] θ command =arctan( synthetic ,Y synthetic ),

[0029] The X and Y coordinates of the AUV-based guide point are:

[0030]

[0031] Among them, X A ′ UV and Y A ′ UV Let d be the X and Y position coordinates of the AUV body center. lookahead θ is the distance from the body center of the AUV to the guide point. AUV It is the current heading angle of the AUV.

[0032] In a further embodiment, the method for calculating the desired heading of an AUV using a vector field algorithm and fuzzy rules further includes:

[0033] The fuzzy rules for AUV navigation include setting a safety distance, introducing a unit momentum vector, and a roll-up torque;

[0034] The unit momentum vector is used to make the obstacle avoidance behavior of the AUV in the vector field approximate that of a conservative potential field.

[0035] The upwind torque is used to turn the AUV back in the direction it came from when it is stuck in extreme terrain, thereby leaving the extreme terrain.

[0036] In a further embodiment, the safe distance is:

[0037]

[0038] Where, k d k is the safety distance gain constant. d >0.5, where m is the mass of the AUV and F is the maximum thrust that the AUV thruster can provide;

[0039] When the AUV detects an obstacle within a safe distance, the AUV thrusters will reverse, causing the AUV to travel backward.

[0040] In a further embodiment, the unit momentum vector is:

[0041]

[0042] In a further embodiment, the winding torque is:

[0043] τ windup =θ end -θ AUV ,

[0044] Where, τ windup Restricted to Inside.

[0045] In a further embodiment, the synthesized vector of the AUV after correction according to the fuzzy rule is:

[0046]

[0047] The desired heading angle is:

[0048] θ cimmand =arctan(X) synthetic ,Y synthetic ),

[0049] The speed increment is:

[0050] dv=k θ (θ command -θ AUV )+k τ τ windup ,

[0051] Where, k θ and k τ These are the heading control gain and the upwind torque control gain, respectively.

[0052] Secondly, embodiments of the present invention provide an AUV obstacle avoidance system based on vector fields and fuzzy rules, the system comprising:

[0053] Data acquisition unit: used to acquire sonar datasets, the sonar datasets including forward-looking sonar images;

[0054] Model training unit: used to construct an obstacle recognition model using a YOLOv4 network and input the sonar dataset into the obstacle recognition model for training;

[0055] Obstacle Environment Construction Unit: This unit is used to identify real-time forward-looking sonar images using a trained obstacle recognition model, identify whether there are obstacles around the AUV, and if so, output the obstacle location. Based on the detected obstacle location, it constructs the obstacle environment using a matrix grid method.

[0056] Desired heading calculation unit: used to obtain the AUV's position using the AUV's positioning system, and to calculate the AUV's desired heading using a vector field algorithm and fuzzy rules based on the AUV's position, destination position, and the positions of obstacles in the obstacle environment;

[0057] Destination determination unit: Used to control the AUV to travel according to the desired course, and continuously determine whether the AUV has reached the destination position based on the dead reckoning algorithm of the AUV dynamic model. If the destination position is reached, the AUV navigation mission is completed.

[0058] Thirdly, embodiments of the present invention provide a computer device including a memory, a processor, and a transceiver, which are connected to each other via a bus; the memory is used to store a set of computer program instructions and data, and can transmit the stored data to the processor, and the processor can execute the program instructions stored in the memory to perform the method as described in any one of the preceding claims.

[0059] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing a computer program that, when executed, implements the method as described in any of the preceding claims.

[0060] This invention provides an AUV obstacle avoidance method, system, computer device, and computer storage medium based on vector fields and fuzzy rules. This application acquires a sonar dataset, constructs an obstacle recognition model using a YOLOv4 network, and trains it based on the sonar dataset. The trained obstacle recognition model identifies real-time forward-looking sonar images to construct an obstacle environment. The AUV's positioning system is used to obtain its position, and based on the obstacle positions, endpoint positions, and the AUV's position within the obstacle environment, a vector field algorithm and fuzzy rules are used to calculate the AUV's desired heading, guiding the AUV forward. The AUV adjusts its steering mechanism to change its heading in three-dimensional space, bringing it closer to the desired heading, thereby avoiding detected obstacles and moving towards the target point. In addition to using a vector field algorithm for obstacle avoidance path planning, the method also introduces fuzzy rules based on safety distance, unit momentum vector, and roll-up torque to ensure the safety and effectiveness of local obstacle avoidance, preventing the AUV from getting stuck in local terrain. Compared to traditional obstacle avoidance methods, this invention not only enables AUVs to effectively avoid unknown obstacles, but also has good adaptability and obstacle avoidance capabilities for complex terrains such as parallel obstacles and square well obstacles, thus ensuring the safe navigation of AUVs in unknown underwater environments. Attached Figure Description

[0061] Figure 1 This is a schematic diagram of an AUV obstacle avoidance method based on vector fields and fuzzy rules provided in an embodiment of the present invention;

[0062] Figure 2 This is a simplified flowchart of an AUV obstacle avoidance method based on vector fields and fuzzy rules provided in an embodiment of the present invention;

[0063] Figure 3 This is a schematic diagram of the synthesized vector in the AUV obstacle avoidance method based on vector fields and fuzzy rules provided in the embodiments of the present invention;

[0064] Figure 4 This is a simulation diagram of parallel obstacles in the AUV obstacle avoidance method based on vector fields and fuzzy rules provided in the embodiments of the present invention;

[0065] Figure 5 This is a simulation diagram of a square well in the AUV obstacle avoidance method based on vector fields and fuzzy rules provided in the embodiments of the present invention;

[0066] Figure 6 This is a schematic diagram of complex terrain simulation in the AUV obstacle avoidance method based on vector field and fuzzy rules provided in the embodiments of the present invention;

[0067] Figure 7 This is a schematic diagram of an AUV obstacle avoidance system based on vector fields and fuzzy rules provided in an embodiment of the present invention;

[0068] Figure 8 This is a schematic diagram of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0069] The embodiments of the present invention are described in detail below with reference to the accompanying drawings. The embodiments are provided for illustrative purposes only and should not be construed as limiting the scope of the invention. The accompanying drawings are for reference and illustration only and do not constitute a limitation on the scope of protection of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of this invention.

[0070] First Embodiment

[0071] Please see Figure 1 and Figure 2 In an embodiment of the present invention, an AUV obstacle avoidance method based on vector fields and fuzzy rules is provided, the method comprising the following steps:

[0072] S1. Obtain a sonar dataset, which includes forward-looking sonar images.

[0073] In the sonar dataset, the forward-looking sonar images are subject to significant noise interference, so preprocessing of the forward-looking sonar images is required before training.

[0074] Preprocessing begins with the two most basic morphological operations: erosion and dilation. Both erosion and dilation use a small convolutional kernel to traverse the entire image. In erosion, the kernel takes a local minimum after traversing the entire image, which manifests as the edges of the foreground image (brighter areas) being eroded. Erosion can be used to remove minor noise and separate connected objects. In dilation, the kernel takes a local maximum after traversing the entire image, which manifests as the foreground image being expanded. Dilation can be used to remove gaps within objects and connect separated adjacent objects. Erosion and dilation are not reversible; performing erosion and dilation sequentially does not restore the original image. This leads to the opening and closing operations in morphological processing. The opening operation (erosion followed by dilation) can be used to eliminate minor noise in the background; the closing operation (dilation followed by erosion) can be used to repair noise within objects. The advantage of opening and closing operations is that they do not significantly affect the shape and size of objects in the overall image. Noise in forward-looking sonar images often appears in the background region. Therefore, this method uses an opening operation, i.e., erosion followed by dilation, to remove background noise.

[0075] Finally, to further reduce noise in the image, median filtering was used. Median filtering is a non-linear smoothing technique, and its image processing is somewhat similar to erosion and dilation: a small convolutional kernel is used to traverse the entire image. For each pixel, the small convolutional kernel calculates the median value obtained after sorting the pixels within its convolution range, and finally assigns this median value to the pixel. Median filtering is effective in removing salt-and-pepper noise.

[0076] S2. An obstacle recognition model is constructed using a YOLOv4 network, and the sonar dataset is input into the obstacle recognition model for training.

[0077] After preprocessing, an obstacle recognition model is constructed using the YOLOv4 network. The main architecture of the YOLOv4 network includes a feature fusion network consisting of a backbone network, a cross-stage partial darknet53 (CSPDarknet53), a spatial pyramid pooling network (SPP), and a path aggregation network (PANet), as well as the final prediction part.

[0078] The CSPDarknet53 network serves as the backbone. After inputting a fixed-size image, it continuously performs convolutions. The width and height of the image are compressed and the number of channels is expanded, eventually forming many feature layers of different scales. The last three feature layers with higher semantic information are then fed into the subsequent feature fusion network.

[0079] The feature fusion network, located between the backbone network and the prediction output, can further enhance the diversity and robustness of features. YOLOv4 utilizes the SPP module to fuse feature maps of different scales and employs top-down and bottom-up feature pyramids to strengthen the network's feature extraction capabilities. The last three feature layers, which have higher semantic information, are fed into the SPP module for max pooling after three convolutions. After pooling, the layers are stacked and then subjected to three more convolutions. Subsequently, they are fed into the PANet network along with the other two feature layers for continuous upsampling, downsampling, and stacking, thus achieving feature fusion and ultimately forming the YOLOv4 feature pyramid structure.

[0080] After the three feature layers are determined, YOLOv4 determines a detection box of a certain size and shape in each cell. Finally, the three feature layers are fed into the prediction part for prediction and output the final result.

[0081] After the obstacle recognition model is built, the sonar dataset is input into the obstacle recognition model for training.

[0082] S3. The trained obstacle recognition model is used to identify the real-time forward-looking sonar images to identify whether there are obstacles around the AUV. If there are obstacles, the location of the obstacles is output, and the obstacle environment is constructed using the matrix grid method based on the detected obstacle locations.

[0083] The forward-looking sonar acquires real-time images of the AUV. It then identifies obstacles in the AUV's vicinity. If an obstacle is present in the forward-looking sonar image, the trained model outputs the obstacle's coordinates. A polar coordinate system is constructed using the forward-looking sonar's position as the origin. A pixel with coordinates (x, y) in the image represents the polar coordinate system with the direction as y. Distance is The acoustic reflection intensity information at the location is obtained, where φ and R represent the horizontal opening angle and slant range of the forward-looking sonar, respectively, and w and h represent the horizontal and vertical dimensions of the image, respectively. This information is then used to output information such as the angle and distance of the obstacle.

[0084] The obstacle environment is constructed using the matrix grid method. The obstacles in the field are represented by a two-dimensional digital matrix consisting of 0 and 1, where 1 represents the obstacle area and 0 represents the obstacle-free area. A Cartesian coordinate system is established in the field.

[0085] S4. The AUV's position is obtained using the AUV's positioning system, and the desired heading of the AUV is calculated using a vector field algorithm and fuzzy rules based on the AUV's position, destination position, and the position of obstacles in the obstacle environment.

[0086] In the process of AUV obstacle avoidance trajectory planning, not only must the vector field algorithm be followed, but also the constraints of fuzzy rules must be met. Fuzzy rules include setting a safety distance, introducing a unit momentum vector and a roll-up torque.

[0087] Initially, the AUV travels in a designated direction with a certain initial velocity based on the coordinates of the destination position. During travel, it makes local obstacle avoidance decisions based on the position and distance of obstacles detected by the forward-looking sonar within a certain field of view. When obstacles are not within the detection range of the forward-looking sonar, the AUV is only affected by the attraction vector of the destination and travels towards the target destination with the attraction vector as the desired heading. When an obstacle enters the detection range of the forward-looking sonar, if the obstacle is within a safe distance that is difficult to avoid in time, the AUV's thrusters will reverse until there are no obstacles within the safe distance and the path will be replanned to avoid collision with the obstacle; if the obstacle is outside the safe distance, the AUV will be affected by both the attraction vector of the destination and the repulsion vector of the obstacle. If the resultant vector of the two is zero, the AUV will avoid the zero vector point by introducing a unit momentum vector and a roll-up torque to avoid being trapped in a dead point of the vector field; otherwise, the AUV will use the resultant vector as the desired heading and control its motion state according to the desired heading to avoid obstacles and head towards the destination. Figure 3 This is a schematic diagram of the synthesized vector in the AUV obstacle avoidance method based on vector fields and fuzzy rules.

[0088] During AUV operation, at each specific time interval, the attraction vector of the destination to the AUV and the repulsion vector of each obstacle within the forward-looking sonar detection range are continuously calculated. The desired heading is then determined based on the arctangent of the vector sum. A simple proportional controller is used to control the heading by the deviation between the current heading and the desired heading.

[0089] In this embodiment of the invention, the desired heading of the AUV is calculated using a vector field algorithm and fuzzy rules, including the following steps:

[0090] First, calculate the azimuth angle from the AUV to the destination based on the AUV's current position and destination position:

[0091] θ end =atan2[(Y end -Y AUV ),(X end -X AUV )],

[0092] Among them, X AUV and Y AUV These are the X and Y position coordinates of the guide point based on the AUV, respectively. end and Y end These are the X and Y coordinates of the target endpoint, respectively.

[0093] Calculate the AUV's attraction vector based on the azimuth angle:

[0094]

[0095] Based on the detected obstacle location and AUV location, calculate the distance d from the obstacle detected by the forward-looking sonar to the AUV. obstacle :

[0096]

[0097] Among them, X obstacle and Y obstacle These are the X and Y coordinates of the detected obstacle.

[0098] Next, the AUV's repulsion vector is calculated based on the AUV's position, the obstacle's position, and the detected distance from the obstacle to the AUV:

[0099]

[0100] Where, d zero This is the maximum detection range of the forward-looking sonar, which is the distance at which the obstacle no longer generates a repulsion vector.

[0101] It is worth noting that, in order to ensure the stability of the AUV when navigating between parallel obstacles, the repulsive and attractive vectors experienced by the AUV are calculated based on the AUV's guide point rather than its body center. That is, in the calculation of the attractive and repulsive vectors:

[0102]

[0103] Where, X′ AUV and Y′ AUV Let d be the X and Y position coordinates of the AUV body center. lookahead θ is the distance from the body center of the AUV to the guide point. AUV It is the current heading angle of the AUV.

[0104] like Figure 3 The schematic diagram of the synthesized vector in the AUV obstacle avoidance method based on vector fields and fuzzy rules of this invention is shown. By combining the repulsive and attractive vectors of the AUV, the synthesized vector can be calculated:

[0105]

[0106] Where, k a and k r The attraction gain and repulsion gain constants can be determined through tuning.

[0107] Furthermore, the expected heading angle of the AUV can be calculated:

[0108] θ command =arctan(X) synthetic ,Y synthetic ),

[0109] After obtaining the desired heading, the AUV's drive can be controlled based on its output, and the AUV's coordinates can be updated. If the destination is not reached, steps S3-S4 are repeated to detect the surrounding environment and obstacles, and the obstacle avoidance decision is updated again. In this way, the local path planning of the AUV when facing unknown obstacles is achieved.

[0110] In addition to using a vector field algorithm to calculate the desired heading to guide obstacle avoidance, this invention also introduces fuzzy rules to ensure the safety and effectiveness of local obstacle avoidance.

[0111] Firstly, a safe distance is set. If an obstacle is detected within this safe distance, making timely obstacle avoidance difficult, the AUV's thrusters will reverse to propel the vehicle backward until there are no obstacles within the safe distance, at which point the path will be replanned to avoid collisions. This safe distance is calculated as follows:

[0112]

[0113] Where, k d k is the safety distance gain constant. d >0.5, where m is the mass of the AUV and F is the maximum thrust that the AUV's thrusters can provide. Introducing a safety distance can further prevent the AUV from colliding with obstacles that are difficult to avoid in time.

[0114] Secondly, by introducing a unit momentum vector, the obstacle avoidance method based on vector fields can cause the AUV to get stuck in a local minimum of the potential field. The unit momentum vector makes the obstacle avoidance behavior of the AUV in the vector field closer to that of a conservative potential field, thus preventing the AUV from getting stuck in a dead point of the vector field. The unit momentum vector is calculated as follows:

[0115]

[0116] Third, an uplift moment is introduced. After leaving the dead point of the vector field, AUVs can still easily get trapped in terrain with extreme concavity (such as square wells). Introducing an uplift moment changes the AUV's heading, causing it to turn back to the entrance direction it came from, thus leaving the extreme terrain. The uplift moment is defined as follows:

[0117] τ windup =θ end -θ AUV ,

[0118] Where, τ windup Restricted to Inside.

[0119] Based on fuzzy rules and by setting a gain constant, the AUV can avoid dead points in the vector field and is not trapped in extreme terrain. The corrected synthetic vector calculation is as follows:

[0120]

[0121] The desired heading angle is calculated as follows:

[0122] θ command =arctan(X) synthetic ,Y synthetic ),

[0123] The speed increment is calculated as follows:

[0124] dv=k θ (θ command -θ AUV )+k τ τ windup ,

[0125] Where, k θ and k τ These are the heading control gain and the upwind torque control gain, respectively, obtained through debugging.

[0126] S4. Control the AUV to sail according to the desired course, and continuously determine whether the AUV has reached the destination position based on the dead reckoning algorithm of the AUV dynamic model. If the destination position is reached, the AUV sailing mission is completed.

[0127] By using the aforementioned desired course and speed increments as guidance for AUV motion control, safe navigation of AUVs in unknown underwater environments can be achieved.

[0128] During AUV navigation, a dead reckoning algorithm based on the AUV dynamics model is needed to continuously determine whether the AUV has reached its destination. If it is determined that the destination has not been reached, steps S3-S4 are repeated at predetermined time intervals to continuously control the AUV's course until the destination is reached and the AUV navigation mission is completed.

[0129] Next, we will verify the AUV obstacle avoidance method based on vector fields and fuzzy rules of this invention. Before verification, it is necessary to construct a model of the AUV in a simulation environment. Four commonly used models for AUV navigation in an external flow field include: particle kinematics model, particle dynamics model, rigid body kinematics model, and rigid body dynamics model. In the model, the inertial (translational and rotational) velocity of the AUV is represented by the flow field and the relative velocity of the AUV. This representation emphasizes the effect of relative flow; fluid dynamic effects such as lift and drag also depend on the motion of the AUV relative to the surrounding fluid. Another equivalent method is to write the equation of motion using the inertial velocity of the AUV. In this case, the consistency between the dynamic model and Newton's second law is more obvious, but considering relative effects becomes more difficult.

[0130] After comprehensive consideration, this invention uses a rigid body dynamics model as the dynamic model of the AUV. In addition to the flow effects caused by added mass and inertia, the external forces and torques acting on the AUV consist of the following:

[0131] Gravity and buoyancy (f g / b and f g / b );

[0132] Viscous effect (f v and m v );

[0133] Propulsion and control (f ctrl and m ctrl ).

[0134] The forces and torques used for propulsion and control are typically generated by the AUV propeller and control rudders, altering the total external forces and torques acting on the AUV. This invention calculates the required propeller speed and control rudder angle based on the speed increment and desired heading angle, and then modifies the control force f. ctrl and control torque m ctrl To achieve motion control of AUV.

[0135] In algorithm verification, using the above dynamic model and obstacle avoidance method, the AUV can continuously update its position in the Cartesian coordinate system during navigation, thus providing guidance for the obstacle avoidance process. If the updated Cartesian coordinates reach the destination range, the mission is completed; if the updated Cartesian coordinates do not reach the destination range, steps S3-S4 are repeated until the target destination is reached, completing the mission. Figure 4-6 The following are schematic diagrams illustrating simulation results for different terrains. Figure 4 This is a simulation diagram of parallel obstacles in an AUV obstacle avoidance method based on vector fields and fuzzy rules. Figure 5 A simulation diagram of a square well in an AUV obstacle avoidance method based on vector fields and fuzzy rules; Figure 6 This is a simulation diagram of complex terrain in the AUV obstacle avoidance method based on vector fields and fuzzy rules. The simulation results verify the effectiveness of the AUV obstacle avoidance method based on vector fields and fuzzy rules.

[0136] In this embodiment of the invention, an AUV obstacle avoidance method based on vector fields and fuzzy rules is proposed to address the obstacle avoidance problem of unknown obstacles. The method involves acquiring a sonar dataset, constructing an obstacle recognition model using a YOLOv4 network, and training the model on the sonar dataset. The trained obstacle recognition model then identifies real-time forward-looking sonar images to construct an obstacle environment. The AUV's positioning system is used to obtain its position, and based on the AUV's position, destination position, and obstacle positions within the obstacle environment, a vector field algorithm and fuzzy rules are used to calculate the AUV's desired heading, guiding the AUV forward. The AUV adjusts its steering mechanism to change its heading in three-dimensional space, bringing it closer to the desired heading, thereby avoiding detected obstacles and moving towards the target point. In addition to using a vector field algorithm for obstacle avoidance path planning, the method also introduces fuzzy rules based on safety distance, unit momentum vector, and roll-up torque to ensure the safety and effectiveness of local obstacle avoidance, preventing the AUV from getting stuck in local terrain. Compared to traditional obstacle avoidance methods, this invention not only enables AUVs to effectively avoid unknown obstacles, but also has good adaptability and obstacle avoidance capabilities for complex terrains such as parallel obstacles and square well obstacles, thus ensuring the safe navigation of AUVs in unknown underwater environments.

[0137] Second Embodiment

[0138] Accordingly, such as Figure 7 As shown, based on an AUV obstacle avoidance method based on vector fields and fuzzy rules, this embodiment of the invention also provides an AUV obstacle avoidance system based on vector fields and fuzzy rules, the system comprising:

[0139] Data acquisition unit 1: used to acquire sonar datasets, the sonar datasets including forward-looking sonar images;

[0140] Model training unit 2: used to construct an obstacle recognition model using a YOLOv4 network and input the sonar dataset into the obstacle recognition model for training;

[0141] Obstacle Environment Construction Unit 3: It is used to identify the real-time acquired forward-looking sonar images through the trained obstacle recognition model, identify whether there are obstacles around the AUV, and if there are, output the obstacle position, and construct the obstacle environment using the matrix grid method based on the detected obstacle position.

[0142] Desired heading calculation unit 4: used to obtain the AUV's position using the AUV's positioning system, and to calculate the AUV's desired heading using a vector field algorithm and fuzzy rules based on the AUV's position, destination position, and obstacle positions in the obstacle environment;

[0143] Terminal Judgment Unit 5: Used to control the AUV to sail according to the desired course, and continuously judge whether the AUV has reached the destination position based on the dead reckoning algorithm of the AUV dynamic model. If the destination position is reached, the AUV sailing mission is completed.

[0144] For specific limitations regarding an AUV obstacle avoidance system based on vector fields and fuzzy rules, please refer to the above-described limitations regarding an AUV obstacle avoidance method based on vector fields and fuzzy rules, which will not be repeated here. Those skilled in the art will recognize that the various modules and steps described in conjunction with the embodiments disclosed in this application can be implemented in hardware, software, or a combination of both. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0145] Third Embodiment

[0146] like Figure 8 As shown in the figure, an embodiment of the present invention provides a computer device including a memory, a processor, and a transceiver, which are connected to each other via a bus; the memory is used to store a set of computer program instructions and data, and can transmit the stored data to the processor; the processor can execute the program instructions stored in the memory to perform the steps of the above-described AUV obstacle avoidance method based on vector fields and fuzzy rules.

[0147] The memory may include volatile memory or non-volatile memory, or both; the processor may be a central processing unit, a microprocessor, an application-specific integrated circuit, a programmable logic device, or a combination thereof. By way of example, but not limitation, the programmable logic device described above may be a complex programmable logic device, a field-programmable gate array, a general-purpose array logic, or any combination thereof.

[0148] In addition, memory can be a physically independent unit or integrated with the processor.

[0149] Those skilled in the art will understand that Figure 8 The structure shown is merely a block diagram of a portion of the structure related to the present invention application and does not constitute a limitation on the computer device to which the present invention application is applied. A specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have the same component arrangement.

[0150] Fourth embodiment

[0151] In one embodiment, a computer-readable storage medium is provided for storing one or more computer programs, the one or more computer programs including program code, which, when run on a computer, is used to execute the above-described AUV obstacle avoidance method based on vector fields and fuzzy rules.

[0152] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., SSD), etc.

[0153] 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 computer-readable storage medium, and when the computer program is executed, it can include the processes of the embodiments of the above methods.

[0154] This embodiment presents an AUV obstacle avoidance method, system, computer device, and computer storage medium based on vector fields and fuzzy rules to address the technical problems of traditional obstacle avoidance methods for unknown obstacles, such as easy getting trapped in local minima and unstable and low-quality obstacle avoidance paths. This application acquires a sonar dataset and trains an obstacle recognition model using the YOLOv4 algorithm. The forward-looking sonar identifies the presence of obstacles around the AUV based on the trained obstacle recognition model and constructs an obstacle environment. The AUV's positioning system obtains its position. Based on the AUV's position, destination position, and the positions of obstacles within the obstacle environment, the vector field algorithm and fuzzy rules are used to calculate the AUV's desired heading to guide its movement. The AUV adjusts its steering mechanism to change its heading in three-dimensional space, bringing it closer to the desired heading, thereby avoiding detected obstacles and moving towards the target point. In addition to using the vector field algorithm for obstacle avoidance path planning, the method also introduces fuzzy rules based on safety distance, unit momentum vector, and roll-up torque to ensure the safety and effectiveness of local obstacle avoidance and prevent the AUV from getting stuck in local terrain. Compared to traditional obstacle avoidance methods, this invention not only enables AUVs to effectively avoid unknown obstacles, but also has good adaptability and obstacle avoidance capabilities for complex terrains such as parallel obstacles and square well obstacles, thus ensuring the safe navigation of AUVs in unknown underwater environments.

[0155] The embodiments described above are merely preferred embodiments of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be understood that this invention is illustrative rather than restrictive, and the method is not limited to the specific configurations described above. Although the accompanying drawings illustrate one or more examples of the invention, any modifications, equivalent substitutions, or improvements made by those skilled in the art without departing from the spirit and principles of the invention should be included within the scope of the claims. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An AUV obstacle avoidance method based on vector fields and fuzzy rules, characterized in that, The method includes: Acquire a sonar dataset, which includes forward-looking sonar images; An obstacle recognition model was constructed using a YOLOv4 network, and the sonar dataset was input into the obstacle recognition model for training. The trained obstacle recognition model identifies real-time forward-looking sonar images to determine whether there are obstacles around the AUV. If there are obstacles, the location of the obstacles is output, and the obstacle environment is constructed using a matrix grid method based on the detected obstacle locations. The AUV's position is obtained using an AUV positioning system, and the desired heading of the AUV is calculated using a vector field algorithm and fuzzy rules based on the AUV's position, destination position, and the positions of obstacles in the obstacle environment. The method for calculating the desired heading of an AUV using vector field algorithms and fuzzy rules includes: Calculate the azimuth angle between the AUV and the destination based on the AUV's current position and destination position: ， in, and These are the X and Y position coordinates of the guide point based on the AUV. and These are the X and Y coordinates of the target endpoint, respectively. Calculate the AUV's attraction vector based on the azimuth angle: ， Based on the detected obstacle location and AUV location, calculate the distance from the detected obstacle to the AUV: ， in, and The X and Y coordinates of the detected obstacle; The AUV's repulsion vector is calculated based on the AUV's location, the obstacle's location, and the detected distance from the obstacle to the AUV: ， in, This represents the maximum detection range of the forward-looking sonar. Calculate the composite vector of the AUV based on its repulsion and attraction vectors: , in, and These are the attraction gain and repulsion gain constants; Calculate the desired heading angle of the AUV based on the synthesized vector: ， The X and Y coordinates of the AUV-based guide point are: ， in, and Let X and Y be the coordinates of the body center of the AUV. The distance from the AUV's body center to the guide point. This is the current heading angle of the AUV; The method for calculating the desired heading of an AUV using vector field algorithms and fuzzy rules also includes: The fuzzy rules for AUV navigation include setting a safety distance, introducing a unit momentum vector, and a roll-up torque; The unit momentum vector is used to make the obstacle avoidance behavior of the AUV in the vector field approximate that of a conservative potential field. The upwind torque is used to turn the AUV back to the direction it came from when it is stuck in extreme terrain, thereby leaving the extreme terrain. The AUV is controlled to navigate according to the desired course, and a dead reckoning algorithm based on the AUV dynamics model is continuously used to determine whether the AUV has reached the destination. If the destination has been reached, the AUV navigation mission is completed.

2. The AUV obstacle avoidance method based on vector fields and fuzzy rules as described in claim 1, its features... The characteristic is that the safe distance is: ， in, The safety distance gain constant, , For AUV quality, The maximum thrust that an AUV propulsion unit can provide; When the AUV detects an obstacle within a safe distance, the AUV thrusters will reverse, causing the AUV to travel backward.

3. The AUV obstacle avoidance method based on vector fields and fuzzy rules as described in claim 1, characterized in that, The unit momentum vector is: 。 4. The AUV obstacle avoidance method based on vector fields and fuzzy rules as described in claim 3, characterized in that, The upward winding torque is: ， in, Restricted to Inside.

5. The AUV obstacle avoidance method based on vector fields and fuzzy rules as described in claim 4, characterized in that, The synthesized vector of the AUV after correction according to the fuzzy rule is: ， in, To attract gain, the desired heading angle is: ， The speed increment is: ， in, and These are the heading control gain and the upwind torque control gain, respectively.

6. An AUV obstacle avoidance system based on vector fields and fuzzy rules, characterized in that, The system includes: Data acquisition unit: used to acquire sonar datasets, the sonar datasets including forward-looking sonar images; Model training unit: used to construct an obstacle recognition model using a YOLOv4 network and input the sonar dataset into the obstacle recognition model for training; Obstacle Environment Construction Unit: This unit is used to identify real-time forward-looking sonar images using a trained obstacle recognition model, identify whether there are obstacles around the AUV, and if so, output the obstacle location. Based on the detected obstacle location, it constructs the obstacle environment using a matrix grid method. Desired heading calculation unit: used to obtain the AUV's position using the AUV's positioning system, and to calculate the AUV's desired heading using a vector field algorithm and fuzzy rules based on the AUV's position, destination position, and the positions of obstacles in the obstacle environment; The method for calculating the desired heading of an AUV using vector field algorithms and fuzzy rules includes: Calculate the azimuth angle between the AUV and the destination based on the AUV's current position and destination position: ， in, and These are the X and Y position coordinates of the guide point based on the AUV. and These are the X and Y coordinates of the target endpoint, respectively. Calculate the AUV's attraction vector based on the azimuth angle: ， Based on the detected obstacle location and AUV location, calculate the distance from the detected obstacle to the AUV: ， in, and The X and Y coordinates of the detected obstacle; The AUV's repulsion vector is calculated based on the AUV's location, the obstacle's location, and the detected distance from the obstacle to the AUV: ， in, This represents the maximum detection range of the forward-looking sonar. Calculate the composite vector of the AUV based on its repulsion and attraction vectors: , in, and These are the attraction gain and repulsion gain constants; Calculate the desired heading angle of the AUV based on the synthesized vector: ， The X and Y coordinates of the AUV-based guide point are: ， in, and Let X and Y be the coordinates of the body center of the AUV. The distance from the AUV's body center to the guide point. This is the current heading angle of the AUV; The method for calculating the desired heading of an AUV using vector field algorithms and fuzzy rules also includes: The fuzzy rules for AUV navigation include setting a safety distance, introducing a unit momentum vector, and a roll-up torque; The unit momentum vector is used to make the obstacle avoidance behavior of the AUV in the vector field approximate that of a conservative potential field. The upwind torque is used to turn the AUV back to the direction it came from when it is stuck in extreme terrain, thereby leaving the extreme terrain. Destination determination unit: Used to control the AUV to travel according to the desired course, and continuously determine whether the AUV has reached the destination position based on the dead reckoning algorithm of the AUV dynamic model. If the destination position is reached, the AUV navigation mission is completed.

7. A computer device, characterized in that: It includes a memory, a processor, and a transceiver, which are connected to each other via a bus; the memory is used to store a set of computer program instructions and data, and can transfer the stored data to the processor, and the processor can execute the program instructions stored in the memory to perform the method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program that, when executed, implements the method as described in any one of claims 1 to 5.

Images