Cable-controlled robot intelligent control system applicable to seabed multi-sample collection

Abstract

A cable-controlled robot intelligent control system suitable for seabed multi-sample collection is characterized in that a coordinate system is established, an ROV linear velocity matrix under a body coordinate system is converted into a spatial position change rate matrix under an inertial coordinate system, and a first conversion relation model is constructed; establishing a second conversion relation model between the angular velocity and the attitude angle change rate through the geometrical relation between the angular velocity and the attitude angle; converting the first relation model and the second relation model to obtain a kinematic model containing the spatial position, the attitude angle, the linear velocity and the angular velocity of the ROV; introducing an A * algorithm based on a kinematic model and a dynamical model of the ROV; an artificial potential field method is introduced based on a kinematic model and a kinetic model of the ROV, and an ROV obstacle avoidance path is planned; simulation is performed by using python programming based on ROV local path planning of an artificial potential field algorithm, and the ROV can complete local path planning in an unknown environment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of underwater robot control, and particularly relates to a cable-controlled robot intelligent control system applicable to seabed multi-sample collection. BACKGROUND

[0002] With the development of ocean exploration and intelligent control technology, cable-controlled underwater robot (ROV) technology appears. The cable-controlled underwater robot has the characteristics of large working depth, strong environmental adaptability and remote precise control, can carry various sensors and operation tools, and can perform seabed topographic mapping, marine ecological monitoring, marine resource exploration and other diversified tasks in the underwater complex environment, thereby promoting the in-depth research on the autonomous navigation and efficient task execution capability of the ROV in the field of marine scientific research. Among them, the global path planning of the cable-controlled underwater robot as one of the core technologies to realize the autonomous navigation and efficient task execution needs to plan an optimal or collision-free path for the cable-controlled underwater robot from the starting point to the target point in the three-dimensional marine environment based on the prior map or real-time perception data, and at the same time needs to meet the safety, energy efficiency and time cost balance under multiple constraint conditions.

[0003] In related technologies, for the global path planning of the ROV, some rule-based algorithms such as Dijkstra algorithm and A* algorithm are usually used for path planning in two-dimensional or simplified three-dimensional environment. However, the current way based on the simple rule algorithm can realize path planning to a certain extent, but often ignores the dynamics of the ROV, the dynamic changes of the marine environment and the comprehensive optimization under multiple constraint conditions, resulting in the problems of high energy consumption, long time and easy collision of the planned path in actual application. Therefore, it is necessary to provide a navigation control method of the cable-controlled underwater robot to meet the demand of efficient and safe operation of the cable-controlled underwater robot in the field of modern marine scientific research. SUMMARY

[0004] Therefore, it is necessary to provide a cable-controlled robot intelligent control system applicable to seabed multi-sample collection to solve the problems of high energy consumption, long time and easy collision of the traditional ROV global path planning method, and difficult to meet the demand of efficient and safe operation of the cable-controlled underwater robot.

[0005] The technical scheme adopted by the present application is as follows:

[0006] A cable-controlled robot intelligent control system applicable to seabed multi-sample collection comprises the following steps:

[0007] S1. Establish an inertial coordinate system E-xyz for describing the motion trajectory and motion posture of the ROV, and establish a body coordinate system O-x b y b zb ;

[0008] S2. Using the conversion matrix, the ROV linear velocity matrix in the body coordinate system is converted into a spatial position change rate matrix in the inertial coordinate system, thereby constructing a first conversion relationship model;

[0009] A second conversion relationship model between angular velocity and attitude angle change rate is established through the geometric relationship between angular velocity and attitude angle;

[0010] The first transformation relationship model and the second transformation relationship model are used to obtain a kinematic model including the spatial position, attitude angle, linear velocity and angular velocity of the ROV;

[0011] S3. Assuming that the ROV is a rigid body, according to the rigid body motion theorem and the moment of inertia theorem, the first dynamic model of the ROV can be expressed as:

[0012]

[0013] Among them, M RB is the mass matrix of ROV; C RB (V) is the Coriolis force and centripetal force matrix of ROV, which is related to the object's speed, and C RB (V)=-C RB T (V); F represents the resultant force (torque) acting on the ROV.

[0014] S4: Based on the kinematic and dynamic models of ROV, the A* algorithm is introduced;

[0015] In the A* algorithm, the corner constraint function p(θ) is introduced to make the ROV turn at an obtuse angle;

[0016] Increase the interval step size to reduce the number of calculation nodes of the optimal path;

[0017] Enhance the effect of the estimated cost with the target node, so that the ROV always has a tendency to search towards the end point;

[0018] S5: Based on the kinematic and dynamic models of the ROV, the artificial potential field method is introduced to perform ROV obstacle avoidance path planning;

[0019] S6: The ROV local path planning based on the artificial potential field algorithm was simulated by using Python programming. Under the action of the designed algorithm, the ROV was able to complete local path planning in an unknown environment.

[0020] Its further technical solution is:

[0021] In S2, the kinematic model of ROV is obtained as follows:

[0022]

[0023] The linear velocity and angular velocity of the origin O of the ROV body coordinate system can be expressed in matrix form in the body coordinate system as follows: V = [V1 V2] T wherein the matrix expression of the linear velocity of the ROV movement is: V1 = [u v w] T and the matrix expression of the angular velocity is: V2 = [p q r] T ; the expression matrix of the movement position and posture of the ROV in the inertial coordinate system is: η = [p θ] T wherein the matrix expression of the spatial position of the ROV is: p = [x y z] T and the attitude angle of the ROV is: θ = [φ θ ψ] T ;

[0024] wherein, is the conversion matrix from the ROV body coordinate system to the inertial coordinate system, and

[0025] which is related to the attitude angle and can be expressed as:

[0026]

[0027] In S3, considering the static force, propeller thrust and hydrodynamic force acting on the ROV during navigation, the dynamic model of the ROV is obtained based on the first dynamic model:

[0028]

[0029] wherein: M RB is the mass matrix of the ROV; C RB (V) is the Coriolis force and centripetal force matrix of the ROV, V is the velocity of the ROV, is the acceleration of the ROV, F I is the inertial hydrodynamic force, F D is the viscous hydrodynamic force, F R is the moment generated by the combined action of the gravity W and the buoyancy B of the object, and F Thr is the moment of the ROV propeller.

[0030] In S4, the specific steps of applying the A* algorithm to the ROV path planning are as follows:

[0031] A* algorithm will create two lists for storing nodes during the search process, open list and closed list. Open list stores nodes to be considered, each node has an estimated cost value, which represents the estimated cost of the path from the start node to the current node. In each step of the search process, A* algorithm will choose the node with the lowest estimated cost in the open list to expand. Closed list stores nodes that have been considered, once a node is added to the closed list, it will not be expanded again, which can avoid considering the same node repeatedly in the search process and improve the efficiency of the algorithm. When a node is selected as the current node, it will be moved from the open list to the closed list, while the nodes adjacent to the current node but not yet considered will be added to the open list. By maintaining the open list and the closed list, A* algorithm can search the nodes in the grid in an orderly manner and find the shortest path.

[0032] The planning steps of A* algorithm are as follows:

[0033] Initialization: add the start node to the open list and set its estimated cost to 0;

[0034] Loop search: when the open list is not empty;

[0035] Path backtracking: when the target node is found, the shortest path can be determined by tracing back to the start node along the parent node of each node.

[0036] The execution steps of loop search are as follows:

[0037] Select the node with the lowest estimated cost in the open list and mark it as the current node;

[0038] Remove the current node from the open list and add it to the closed list;

[0039] If the current node is the target node, it means that the shortest path has been found, end the algorithm;

[0040] Iterate through the adjacent nodes of the current node;

[0041] If the adjacent node is in the closed list, ignore it; if the adjacent node is not in the open list, add it to the open list and calculate its estimated cost, actual cost and heuristic estimate; if the adjacent node is already in the open list, check if the path through the current node to reach the node is shorter, if so, update the parent node and cost value of the node.

[0042] The beneficial effects of the present application are as follows:

[0043] The present invention takes into account the static force, propulsion system thrust and hydrodynamic force (including inertial hydrodynamic model and viscous hydrodynamic model) suffered by ROV during navigation, which can enhance the adaptability to complex underwater environment. In the marine environment, the external force suffered by ROV is complex and changeable, and hydrodynamic disturbance (such as flow velocity variation, eddy current, additional inertia) has a significant impact on its attitude and speed. Taking the complete hydrodynamic model into consideration, it is helpful for the control system to accurately estimate the disturbance effect according to the current speed and acceleration, thereby improving the robustness of the system. For example, the inertial hydrodynamic term can characterize the additional mass effect of the fluid on the acceleration, and the viscous damping term can reflect the changing trend of the resistance suffered by ROV under different speed states. It can also realize more reasonable propulsion system distribution and energy consumption optimization. The introduction of the propeller thrust model helps to reasonably distribute thrust in a multi-propeller system, avoids the energy waste caused by thrust overload or excessive redundancy. Considering static force (such as the torque of buoyancy and gravity) can assist in maintaining attitude balance, alleviate the burden of attitude adjustment on the propeller, thereby reducing energy consumption and improving the continuity of task execution. Furthermore, to improve the consistency between simulation results and the real system, more accurate dynamic modeling is introduced during the path planning and simulation verification phases to make the simulation environment more closely resemble the real operating environment. This helps evaluate the effectiveness and stability of planning algorithms under complex real-world conditions and shortens the debugging cycle from simulation to actual deployment.

[0044] The present invention can provide path feasibility constraints (the role of the kinematic model). The kinematic model describes the relationship between the position, attitude, and velocity of the ROV, reflecting its posture evolution law and non-holonomic constraints (such as the no-sideslip constraint and the maximum turning radius). In global path planning, it mainly plays the following roles:

[0045] (1) Restricting the path search space: ensuring that the path is physically reachable and avoiding generating trajectories that violate the ROV’s motion characteristics (such as sudden turns, sideways sliding, etc.).

[0046] (2) Constrained turning radius and maximum speed: Assisted planning generates a feasible path that meets the ROV's maneuverability

[0047] This invention can evaluate the cost and safety of path execution (the role of the dynamic model). The dynamic model describes the motion response of the ROV under the action of forces, including factors such as propulsion, hydrodynamics, buoyancy, mass, and added mass, reflecting the system's dynamic response capability and control load. It has the following functions in path planning:

[0048] (1) Evaluate the “controllability” and energy consumption cost of different paths: The dynamic model can be used to estimate the propulsion force and energy consumption required for each section along the path, thereby adding energy efficiency indicators to the path cost function to achieve path optimization.

[0049] (2) Avoiding uncontrollable trajectories: Although some paths are kinematically feasible, the ROV may not be able to track them stably due to factors such as hydrodynamic interference and acceleration exceeding the limit. The dynamic model can identify this risk in advance and avoid planning unfeasible trajectories. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 These are the coordinate system and ROV motion parameters of the present invention.

[0051] Figure 2 This is the seabed topographic map of the operating area of ​​the present invention.

[0052] Figure 3 This is the grid map of the present invention.

[0053] Figure 4 This is the force model of the artificial potential field method of the present invention.

[0054] Figure 5 This is the ROV local path planning based on the dynamic potential field method of the present invention.

[0055] Figure 6 This is a distance change diagram between the ROV and the dynamic obstacle of the present invention. DETAILED DESCRIPTION

[0056] The specific embodiments of the present invention will be described below with reference to the accompanying drawings.

[0057] like Figures 1-6 As shown, the intelligent control system of a cable-controlled robot applicable to collecting multiple samples on the seabed described in this embodiment mainly includes the following steps:

[0058] S1. Establish an inertial coordinate system E-xyz to describe the ROV motion trajectory and motion posture, and establish a body coordinate system Ox to describe the ROV force condition b y b z b ;

[0059] S2. Using the conversion matrix, the ROV linear velocity matrix in the body coordinate system is converted into a spatial position change rate matrix in the inertial coordinate system, thereby constructing a first conversion relationship model;

[0060] A second conversion relationship model between angular velocity and attitude angle change rate is established through the geometric relationship between angular velocity and attitude angle;

[0061] The first transformation relationship model and the second transformation relationship model are used to obtain a kinematic model including the spatial position, attitude angle, linear velocity and angular velocity of the ROV;

[0062] S3. Assuming the ROV as a rigid body, according to the rigid body motion theorem and the inertia distance theorem, the first dynamic model of the ROV can be expressed as:

[0063]

[0064] wherein M RB is the mass matrix of the ROV; C RB (V) is the Coriolis force and centripetal force matrix of the ROV, which is related to the motion speed of the object, and C RB (V) = -C RB T (V); F represents the resultant moment of force acting on the ROV;

[0065] S4: Based on the kinematic model and the dynamic model of the ROV, the A* algorithm is introduced;

[0066] In the A* algorithm, the corner constraint function p(θ) is introduced, so that the ROV turns in an obtuse angle;

[0067] The interval step length is increased, thereby reducing the number of calculation nodes of the optimal path;

[0068] The effect of the estimated cost of the target node is enhanced, thereby making the ROV always have a tendency to search in the direction of the end point;

[0069] S5: Based on the kinematic model and the dynamic model of the ROV, the artificial potential field method is introduced to perform the ROV obstacle avoidance path planning;

[0070] S6: The simulation of the ROV local path planning based on the artificial potential field algorithm is performed by using python programming, and under the action of the designed algorithm, the ROV can complete the local path planning in the unknown environment.

[0071] In S2, the kinematic model of the ROV is obtained as follows:

[0072]

[0073] The linear velocity and the angular velocity of the origin O of the body coordinate system of the ROV in the matrix form in the body coordinate system can be expressed as: V = [V1 V2] T , wherein the matrix expression of the linear velocity of the ROV motion is: V1 = [u v w] T , the matrix expression of the angular velocity is: V2 = [p q r] T ; the expression matrix of the motion position and the attitude of the ROV in the inertial coordinate system is: η = [p θ] T , wherein the matrix expression of the spatial position of the ROV is: p = [x y z] T , and the attitude angle of the ROV is: θ = [φ θ ψ]T ;

[0074] wherein, is the transformation matrix from the ROV body coordinate system to the inertial coordinate system, and

[0075] The relationship with the attitude angle can be expressed as:

[0076]

[0077] In S3, considering the static force, the propulsion system thrust and the hydrodynamic force acting on the ROV during navigation, the dynamic model of the ROV is obtained based on the first dynamic model:

[0078]

[0079] wherein: M RB is the mass matrix of the ROV; C RB (V) is the Coriolis force and centripetal force matrix of the ROV, V is the velocity of the ROV, is the acceleration of the ROV, F I is the inertial hydrodynamic force, F D is the viscous hydrodynamic force, F R is the moment of the combined action of the gravity W and the buoyancy B on the object, F Thr is the moment of the ROV propeller.

[0080] In S4, the specific steps of applying the A* algorithm to the ROV path planning are as follows:

[0081] In the running process of the A* algorithm, two lists for storing nodes in the search process are created, which are the open list and the closed list. The open list stores nodes to be considered, and each node has an estimated cost value, which represents the estimated path cost from the starting node to the current node. In each step of the search process, the A* algorithm selects the node with the lowest estimated cost in the open list for expansion. The closed list stores nodes that have been considered. Once a node is added to the closed list, it will not be expanded again, which can avoid repeatedly considering the same node in the search process and improve the efficiency of the algorithm. When a node is selected as the current node, it is moved from the open list to the closed list. Nodes adjacent to the current node but not yet considered are added to the open list. Through the maintenance of the open list and the closed list, the A* algorithm can orderly search the nodes in the grid and find the shortest path.

[0082] The planning steps of the A* algorithm are as follows:

[0083] Initialization: add the starting node to the open list and set its estimated cost to 0;

[0084] Loop search: when the open list is not empty;

[0085] Path backtracking: when the target node is found, the shortest path can be determined by tracing back to the starting node along the parent node of each node.

[0086] The execution steps of loop search are as follows:

[0087] Select the node with the lowest estimated cost in the open list and mark it as the current node;

[0088] Remove the current node from the open list and add it to the closed list;

[0089] If the current node is the target node, it means that the shortest path has been found, and the algorithm ends;

[0090] Iterate through the adjacent nodes of the current node;

[0091] If the adjacent node is in the closed list, ignore it; if the adjacent node is not in the open list, add it to the open list and calculate its estimated cost, actual cost and heuristic estimate; if the adjacent node is already in the open list, check if the path through the current node to reach this node is shorter, if so, update the parent node and cost value of this node.

[0092] In actual work:

[0093] Firstly:

[0094] In order to better describe the motion state and force state of ROV, it is necessary and important to establish a suitable coordinate system. In this study, the system recommended by the International Towing Tank Conference (ITTC) and the Society of Naval Architects and Marine Engineers (SNAME) is adopted to establish two coordinate systems: one is the inertial coordinate system, which is used to describe the motion trajectory and attitude of ROV; the other is the body coordinate system, which is used to describe the force state of ROV. Figure 1 As shown in the figure)

[0095] The inertial coordinate system E-xyz is also called the fixed coordinate system (abbreviated as "fixed system"). The inertial coordinate system is fixed at an arbitrary point E on the earth, the positive direction of Ez axis points to the center of the earth, Ex and Ey axes are perpendicular to each other in the horizontal plane, so that the coordinate system E-xyz becomes a right-handed coordinate system.

[0096] The body coordinate system O-x b y b z b is also called the moving coordinate system (abbreviated as "moving system"). The body coordinate system is fixed at a point O on the ROV, Ox b axis is positive along the longitudinal direction of the ROV; Oy bThe axis is positive for the right side of the ROV; Oz b The axis is perpendicular to Ox b y b The axis is positive for the downward direction of the plane.

[0097] The linear velocity and angular velocity of the origin O of the ROV body coordinate system can be expressed in the matrix form as follows: V = [V1 V2] T wherein the matrix expression of the linear velocity of the ROV movement is: V1 = [u v w] T and the matrix expression of the angular velocity is: V2 = [p q r] T ; the expression matrix of the movement position and posture of the ROV in the inertial coordinate system is: η = [p θ] T wherein the matrix expression of the spatial position of the ROV is: p = [x y z] T and the posture angle of the ROV is: θ = [φ θ ψ] T ; the expression matrix of the force and moment of the ROV in the body coordinate system is: F = [F1 F2] T wherein the force matrix is: F1 = [X Y Z] T and the moment matrix of the force acting on the ROV is: F2 = [K M N] T All the movement parameters are positive in the direction of the positive coordinate axis or in the direction of the positive coordinate axis according to the right-hand rule, and vice versa. The specific movement definition of the ROV in different degrees of freedom is shown in Table 1.

[0098] Table 1 Movement parameter definition of ROV

[0099]

[0100] Secondly,

[0101] The ROV body coordinate system can coincide with the inertial coordinate system through three rotations around the coordinate axes. Therefore, the conversion relationship between the linear velocity of the ROV body coordinate system and the spatial position of the ROV relative to the inertial coordinate system is:

[0102]

[0103] wherein, is the conversion matrix from the ROV body coordinate system to the inertial coordinate system, and which is related to the posture angle and can be expressed as:

[0104]

[0105] The relationship between the linear velocity of the ROV and the spatial posture can be obtained as follows:

[0106]

[0107] Similarly, the relationship between the angular velocity of the ROV and the attitude angle is:

[0108]

[0109] where the transformation matrix is

[0110] Therefore, the formula can be expanded as:

[0111]

[0112] Further, the kinematic model of the ROV is as follows:

[0113]

[0114] Subsequently:

[0115] In order to facilitate the study of the problem, the following settings and assumptions are made for the ROV. It is assumed that the ROV is a rigid body, its geometric shape does not change, and its mass and center of gravity are constant. The effects of the umbilical cable and other operating tools are not considered, nor is the effect of the seabed on the hydrodynamics of the ROV. The center of gravity of the ROV is taken as the origin of the body coordinate system, and it is approximately considered that the three coordinate axes of the ROV are the central principal axes of inertia.

[0116] According to the rigid body motion theorem and the inertia distance theorem, the dynamics model of the ROV can be expressed as:

[0117]

[0118] where M RB is the mass matrix of the ROV; C RB (V) is the Coriolis force and centripetal force matrix of the ROV, which is related to the motion speed of the object, and C RB (V) = -C RB T (V); F represents the resultant force (moment) acting on the ROV. For the ROV with the center of gravity as the coordinate origin and assuming that the three coordinate axes are the central principal axes of inertia, the mass matrix can be simplified as:

[0119]

[0120] where m is the mass of the ROV, I g is the inertia matrix of the ROV, and has

[0121] I g = diag[I x I y I z ]

[0122]

[0123]

[0124] Coriolis force and centripetal force matrix C RB (V) can be simplified as:

[0125]

[0126] S is the vector multiplication operator,

[0127]

[0128] M RB , C RB By substituting the expression of (V) into the equation and expanding it, we can obtain the general equation of ROV's force motion:

[0129]

[0130] ROV force analysis:

[0131] In general, the forces (torques) acting on the ROV can be roughly divided into several categories: static forces (torques) F generated by the combined action of the gravity W and buoyancy B on the object; R ; Hydrodynamic force (torque) F caused by the movement of ROV in the flow field H ROV thruster thrust (torque) F thr Accurate calculation and analysis of the above torques are the basis for ROV motion control and manipulation simulation.

[0132] Hydrodynamic force (torque) F acting on ROV H It is closely related to the geometric shape, movement speed, acceleration and movement direction of the ROV. In the case of infinitely deep and wide still water, that is, without considering the influence of flow field boundaries, currents and internal waves, for a certain ROV with a fixed and unchanging shape, its hydrodynamic force (torque) depends only on the ROV's speed V and acceleration. ROV's hydrodynamic force (torque) F H According to whether it is caused by steady motion or unsteady hydrodynamics, it can be divided into inertial hydrodynamics F I and viscous hydrodynamic force F D There are two types of inertial hydrodynamics. ROV inertial hydrodynamics is not only related to ROV acceleration, but also to the speed of movement. In order to facilitate calculation and research, the inertial hydrodynamics related to speed is incorporated into the viscous hydrodynamics F D Therefore, the hydrodynamic force can be expressed as:

[0133]

[0134] The kinematic model of ROV can be expressed as:

[0135]

[0136] Static force is a restoring force due to the combined action of gravity and buoyancy of ROV. When the center of gravity is the origin of the body coordinate system, according to the conversion relationship between the inertial coordinate system and the body coordinate system, the static force (torque) generated by the combined action of gravity and buoyancy in the body coordinate system of ROV is expressed as:

[0137]

[0138] Static force (torque) F k (η) is related to the motion posture of ROV. Wherein, r b represents the vector radius of the ROV center of buoyancy relative to the body coordinate system, r b =[x b , y b , z b ] T When the gravity and buoyancy of ROV are equal in size, the static water (torque) F k (η) of ROV can be further simplified as:

[0139]

[0140] Based on the seafloor topographic data in the ISA DeepData database, a three-dimensional map is constructed, and a rectangular region with a latitude of 21.405° to 21.445° north, a longitude of 159.645° to 159.600° east, and a depth of 1.280km to 1.340km is selected as the benchmark test scene of the static global path planning algorithm.

[0141] Based on Figure 2 topographic data, a passability map is generated, which is then simplified to a binary matrix. The discrete depth value under the region is represented as a matrix D, the relative depth is calculated and normalized by the following formula, and then the dimensionless quantity N is obtained:

[0142]

[0143] Where max(D) and min(D) represent the maximum depth and the minimum depth respectively, and the dimensionless quantity N has a value range of [0, 1]. Define a threshold value ε∈(0, 1), and get

[0144]

[0145] This is the passability classifier. Assuming that the binary matrix of the passability map is M0, and the obstacle avoidance band matrix is M rafe , by adjusting the number of loops of the following operation, the radial width of the obstacle avoidance band can be changed:

[0146]

[0147] M rafe =M rfan -M0

[0148] The processed passability map is rasterized to obtain the following output. The black area represents the forbidden area, the blue area is the obstacle avoidance zone, and the white area is the feasible region. The side length of each square is equal to the scale value of the actual map. The red and green squares represent the starting point and the end point of the path, respectively.

[0149] Algorithm design:

[0150] Based on the above grid map model, the decision-making process of the underwater robot can be modeled as a Markov Decision Process (MDP). First, define the basic variables:

[0151] 1) State space The two-dimensional coordinates of the agent in the grid map are unknown, and each state s t = (x t , y t ) ∈ S corresponds to the row and column index of the map cell.

[0152] 2) Action space A = {0, 1,..., 7} defines eight possible moving directions, including four basic directions (up, right, down, left) and four diagonal directions.

[0153] 3) Environmental dynamics are described by transition probabilities P(s t+1 | s t , a t ) and immediate reward function r(s t , a t ), where obstacle collision will result in negative rewards, reaching the target position will result in positive rewards, and other cases will impose a time penalty.

[0154] The goal of the agent is to learn the optimal strategy π * : S → A through interaction with the environment, so that the expected maximum of the cumulative discounted reward is maximized:

[0155]

[0156] where γ ∈ [0, 1] is the discount factor. To achieve this goal, this paper uses the Double Deep Q-Network (Double DQN) algorithm, whose mathematical model can be divided into the following key links:

[0157] (1) Q function approximation and Bellman equation

[0158] Traditional Q-learning approximates the optimal policy by iteratively updating the action-value function Q(s, a) following the Bellman optimality equation:

[0159]

[0160] However, in high-dimensional state space, it is difficult to efficiently store and update the table-form Q-function. To this end, DQN introduces a deep neural network Q(s, a: θ) as the function approximator, where θ is the network parameter. The update target of Q-function is then transformed to minimize the temporal difference error:

[0161]

[0162] where θ - is the target network parameter, which is periodically synchronized from the online network to stabilize the training process.

[0163] (2) Double DQN improvement mechanism

[0164] In traditional DQN, both action selection and value estimation rely on the target network when computing the target Q-value, which leads to overestimation bias. Double DQN decouples action selection from value estimation to solve this problem:

[0165]

[0166] That is, the optimal action of the next state is selected by the current online network θ, while the corresponding Q-value is evaluated by the target network θ

[0167] (3) Experience replay and network update

[0168] The transition samples (s t , a t , r t , s t+1 , d t ) accumulated by the agent during exploration are stored in the experience replay buffer where d t represents the episode termination flag. During training, a small batch of samples is randomly sampled to update the parameters, breaking the temporal correlation and improving data efficiency. The online network parameter θ is updated by gradient descent to minimize the loss function:

[0169]

[0170] The target network adopts a soft update strategy to gradually track the online network: θ - ← τθ + (1 - τ)θ - where τ << 1 is the smoothing coefficient.

[0171] (4) Balancing Exploration and Exploitation

[0172] To balance between exploring new actions and exploiting existing knowledge, the algorithm adopts an e-greedy policy to select actions. For any action a∈A, the probability of its selection is:

[0173]

[0174] where δ(·) is an indicator function that takes 1 when the condition in the parentheses is met, and 0 otherwise; ∈ represents the exploration rate, which varies according to the following decay rate:

[0175] ∈(k) = ∈ end + (∈ start - ∈ end ) · e -k∫decay_steps

[0176] This formula explicitly expresses two sources of probability:

[0177] 1) Exploration: All actions obtain a base probability of ∈, ensuring at least a total probability of ∈ for random exploration.

[0178] 2) Exploitation: The optimal action additionally obtains a probability bonus of (1-∈), ensuring that the strategy preferentially selects the current optimal action when exploiting.

[0179] (5) Network Architecture Design

[0180] The Q network adopts a fully connected structure, with the input layer dimension corresponding to the state space dimension |S = 2, and the output layer dimension matching the action space |A| = 8. To improve the function approximation capability, the network introduces a residual connection module:

[0181] ResBlock(x) = x + W2σ(W1x + b1) + b2

[0182] where σ is the Leaky ReLU activation function, and W1, W2 are trainable weight matrices.

[0183] The parameter initialization adopts an orthogonal initialization method:

[0184]

[0185] where n l is the input dimension of the l-th layer. This initialization strategy, combined with the gain coefficient adjustment of the activation function, can ensure the stability of the signal amplitude in forward propagation.

[0186] (6) Path Planning Specific Design​

[0187] For the grid environment characteristics, the reward function is designed as:

[0188]

[0189] Sparse reward setting forces the agent to quickly find the path to the goal, while the time penalty term -0.1 encourages the shortest path selection. The action space is expanded to 8 directions, allowing diagonal movement, which can reduce the path length by about 29% compared to 4-direction movement.

[0190] (7) Training dynamic analysis

[0191] Define local average reward As a performance indicator, where N is the size of the sliding window. When Consecutive K times do not improve more than the threshold δ Trigger early stop mechanism, save the current optimal model parameters. Constrain the parameter update amplitude by gradient clipping Where ξ is the clipping threshold.

[0192] The specific implementation process of A* algorithm is as follows: first, determine the positions of the starting node and the target node as the starting point and the end point of path planning; then, initialize two lists: open list Open for storing nodes to be examined, and closed list Close for recording nodes that have been examined. Then, A* algorithm starts running, traverses the nodes around the starting node, and selects the node with the minimum cost as the next search target according to certain evaluation criteria. This process is repeated as the path nodes are determined step by step, and the algorithm will continuously select nodes from the open list, examine the nodes around them, and update the open list and the closed list, until it finally extends to the target position, finding the optimal path from the starting node to the target node.

[0193] The traditional A* algorithm has problems such as sharp path turning points, weak search directionality, and excessive calculation of node numbers when planning paths. To solve these problems, an improved A* algorithm is proposed. First, the angle constraint condition of the evaluation function is introduced, which changes the right-angle turning to obtuse angle, making the path turning point more gentle. Second, by improving the distance function, the directionality of path search is enhanced, and redundant nodes on the path are removed, significantly reducing the path planning time. Finally, an improved step method is proposed to optimize the number of path nodes obtained by the improved A* algorithm.

[0194] The artificial potential field method, based on obstacle information within the planning space, can plan an optimal or satisfactory route for an ROV in advance or in real time, within given constraints and evaluation criteria. This algorithm eliminates the need to find a global path, offering fast and efficient performance. It is particularly well-suited for trajectory generation tasks requiring high real-time and safety requirements. While the resulting route may not be the shortest, it is relatively smooth and safe.

[0195] The artificial potential field method is a virtual method. Its basic idea is to represent the environment as a virtual potential field. The target point forms a gravitational potential field, and the obstacles form a repulsive potential field. The ROV moves towards the target under the combined action of the gravitational potential field of the target and the repulsive potential field around the obstacles. att ) and the repulsive force of the repulsive potential field (F rep ) produces a resultant force (F total ), the combined force drives the ROV to move towards the target.

[0196] In order to verify the effect of the artificial potential field algorithm in ROV local path planning, the ROV local path planning based on the artificial potential field algorithm was simulated by using Python programming. Figure 5 As shown, the blue solid line is the ROV local planning path based on the artificial potential field algorithm. Figure 6 The figure shows the distance change between the ROV and various obstacles during the dynamic obstacle avoidance process. From the above simulation results, it can be seen that under the action of the designed algorithm, the ROV can complete local path planning in an unknown environment.

[0197] The above description is an explanation of the present invention, not a limitation of the present invention. The scope of the present invention is defined in the claims. Any modifications may be made within the scope of protection of the present invention.

Claims

1. An intelligent control system for a cable-controlled robot applicable to collecting multiple samples from the seabed, characterized by: The steps include: S1. Establish an inertial coordinate system E-xyz to describe the ROV motion trajectory and motion posture, and establish a body coordinate system Ox to describe the ROV force condition b y b z b ; S2. Using the conversion matrix, the ROV linear velocity matrix in the body coordinate system is converted into a spatial position change rate matrix in the inertial coordinate system, thereby constructing a first conversion relationship model; A second conversion relationship model between angular velocity and attitude angle change rate is established through the geometric relationship between angular velocity and attitude angle; The first transformation relationship model and the second transformation relationship model are used to obtain a kinematic model including the spatial position, attitude angle, linear velocity and angular velocity of the ROV; S3. Assuming that the ROV is a rigid body, according to the rigid body motion theorem and the moment of inertia theorem, the first dynamic model of the ROV can be expressed as: Among them, M RB is the mass matrix of ROV; C RB (V) is the Coriolis force and centripetal force matrix of ROV, which is related to the object's speed, and C RB (V)=-C RB T (V); F represents the resultant force moment acting on the ROV; S4: Based on the kinematic and dynamic models of ROV, the A* algorithm is introduced; In the A* algorithm, the corner constraint function p(θ) is introduced to make the ROV turn at an obtuse angle; Increase the interval step size to reduce the number of calculation nodes of the optimal path; Enhance the effect of the estimated cost with the target node, so that the ROV always has a tendency to search towards the end point; S5: Based on the kinematic and dynamic models of the ROV, the artificial potential field method is introduced to perform ROV obstacle avoidance path planning; S6: The ROV local path planning based on the artificial potential field algorithm was simulated by using Python programming. Under the action of the designed algorithm, the ROV was able to complete local path planning in an unknown environment.

2. The intelligent control system for a cable-controlled robot applicable to collecting multiple seabed samples according to claim 1, characterized in that: In S2, the kinematic model of ROV is obtained as follows: The linear velocity and angular velocity of the origin O of the ROV body coordinate system can be expressed in matrix form as follows: V = [V1 V2] T , where the matrix representation of the linear velocity of the ROV motion is: V1 = [uvw] T , the matrix representation of angular velocity is: V2 = [pqr] T The expression matrix of the motion position and posture of ROV in the inertial coordinate system is: η=[pθ] T , where the matrix of the ROV spatial position is expressed as: p = [xyz] T , the attitude angle of ROV is: θ=[φθψ] T ; in, is the transformation matrix from the ROV body coordinate system to the inertial coordinate system, and Related to the attitude angle, it can be expressed as:

3. The intelligent control system for a cable-controlled robot applicable to collecting multiple seabed samples according to claim 1, characterized in that: In S3, the static force, propulsion system thrust, and hydrodynamic force exerted on the ROV during navigation are considered, and the dynamic model of the ROV is obtained based on the first dynamic model: Among them: M RB is the mass matrix of ROV; C RB (V) is the Coriolis force and centripetal force matrix of ROV, V is the velocity of ROV, is the acceleration of ROV, F I is the inertial hydrodynamic force, F D is the viscous hydrodynamic force, F R F is the torque generated by the combined action of the gravity W and buoyancy B on the object. Thr is the torque of the ROV thruster.

4. The intelligent control system for a cable-controlled robot applicable to collecting multiple seabed samples according to claim 1, characterized in that: In S4, the specific steps of applying the A* algorithm to ROV path planning are: During the operation of the A* algorithm, two lists are created to store nodes in the search process, namely the open list and the closed list. The open list stores the nodes to be considered. Each node has an estimated cost value, which represents the estimated cost of the path from the starting node to the current node. In each step of the search, the A* algorithm selects the node with the lowest estimated cost in the open list for expansion. The closed list stores the nodes that have been considered. Once a node is added to the closed list, it will not be expanded again, which can avoid repeated consideration of the same node during the search process and improve the efficiency of the algorithm. When a node is selected as the current node, it will be moved from the open list to the closed list, and the nodes adjacent to the current node but not yet considered will be added to the open list. By maintaining the open list and the closed list, the A* algorithm can search the nodes in the grid in an orderly manner and find the shortest path.

5. The intelligent control system for a cable-controlled robot applicable to collecting multiple samples on the seabed according to claim 4, characterized in that: The planning steps of the A* algorithm are as follows: Initialization: Add the starting node to the open list and set its estimated cost to 0; Loop search: while the open list is non-empty; Path backtracking: When the target node is found, the shortest path can be determined by tracing back to the starting node along the parent node of each node.

6. The intelligent control system for a cable-controlled robot applicable to collecting multiple seabed samples as claimed in claim 5, characterized in that: The execution steps of the loop search are as follows: Select the node with the lowest estimated cost in the open list and mark it as the current node; Remove the current node from the open list and add it to the closed list; If the current node is the target node, it means the shortest path has been found and the algorithm ends; Traverse the adjacent nodes of the current node; If the adjacent node is in the closed list, ignore it; If the adjacent node is not in the open list, it is added to the open list and its estimated cost, actual cost, and heuristic estimate are calculated; if the adjacent node is already in the open list, check whether the path to the node through the current node is shorter. If so, update the parent node and cost value of the node.

Images