An Energy Consumption Optimal Underwater Area Coverage Method Based on a Bilevel Programming Framework under the Influence of Ocean Currents

Abstract

The present invention provides an energy consumption optimal underwater area coverage method based on a bilevel programming framework under the influence of ocean currents, including: processing the known map of the sea area to be operated to construct an environmental map; formulating this problem as a bilevel programming model by using the coverage path and the energy consumption index as decision variables; the upper-layer algorithm uses a bio-inspired neural network algorithm to solve the coverage path and transmits it to the lower-layer algorithm; the lower layer uses the ant colony optimization algorithm to solve the energy consumption problem, calculates its energy consumption fitness value with the ocean current as a constraint and returns it to the upper-layer algorithm. If there is no feasible path, it searches for an escape path and returns it to the upper-layer algorithm; the upper-layer algorithm makes a path decision based on the result of the lower-layer algorithm, iterates cyclically until the task area coverage is completed, and outputs the optimal coverage path and the AUV propulsion speed. The present invention can select a more beneficial coverage direction and propulsion speed under the influence of ocean currents, and reduce the energy consumption of AUV underwater detection operations while completing the task.

Description

Technical Field

[0001] The present invention belongs to the field of path planning and relates to an energy consumption optimal underwater area coverage method based on a two-layer planning framework under the influence of ocean currents. Background Art

[0002] To manage the ocean, it is necessary to understand the ocean first, which must be achieved through various underwater robots and acoustic systems for ocean exploration to obtain accurate ocean data such as underwater maps, geological resource distributions, and ecological distributions. Therefore, in order for underwater robots to safely and effectively explore a predetermined area, a suitable coverage path planning strategy is necessary.

[0003] Early research on coverage path planning methods relied on heuristic methods. Such methods might work very well in implementation, but could not ensure that the robot successfully covered the entire target area. However, one of the most important criteria for evaluating whether a coverage path planning strategy can be successfully applied is that the robot can completely cover the target area. Additionally, graph traversal algorithms convert the traversal of the working area into the traversal of all nodes in the graph, which can ensure that each node is covered, but a large number of repeated paths are traversed, and the shortest covering distance cannot be guaranteed. Moreover, compared with onshore and aerial carriers such as unmanned aerial vehicles and unmanned vehicles, the deployment and recovery of underwater robots are more cumbersome, and the battery capacity is limited. The actual detectable area size of underwater robots is limited by energy. Therefore, it is very valuable to search for the optimal path through optimization algorithms and reasonably plan the energy consumption of underwater robots so that the underwater robots can complete the full coverage of the task area with a shorter operation path and lower energy consumption.

[0004] In addition, ocean currents naturally exist in the ocean environment and will interfere with the navigation of underwater robots. If the influence of ocean currents is not considered, the planned propulsion speed of the underwater robot is the same as the actual propulsion speed relative to the seabed. In fact, ocean currents will affect the accuracy of the implementation control of the planning algorithm. Especially in strong ocean currents, infeasible paths may even occur. Therefore, how to operate along the direction of ocean currents as much as possible and reasonably utilize the energy of ocean currents is the key to improving the efficiency of underwater area detection operations and is a challenging task in ocean exploration. Summary of the Invention

[0005] To solve the above problems, the present invention discloses an energy consumption optimal underwater area coverage method based on a bilevel programming framework under the influence of ocean currents. Taking an Autonomous Underwater Vehicle (AUV) as the research carrier of the present invention, it realizes the full coverage of the underwater mission area, comprehensively considers multiple indicators such as ocean current conditions, energy consumption, turning angle, coverage distance, and time, and plans a full coverage mission path with optimal energy consumption; divides this area coverage task into path planning and path optimization parts, formulates the problem as a bilevel programming framework, and solves it through a hybrid algorithm of a bio-inspired neural network algorithm and an ant colony algorithm; the present invention can select a more beneficial coverage direction and propulsion speed under the influence of ocean currents, and reduce the energy consumption of AUV underwater detection operations while completing the task.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] An energy consumption optimal underwater area coverage method based on a bilevel programming framework under the influence of ocean currents, specifically including the following steps:

[0008] Step 1: Process the known map of the sea area to be operated, and construct an environmental map, which includes separating the obstacle area and the area to be operated, binary processing the image to construct a grid map, and dilating the obstacle processing in three parts;

[0009] Step 2: Further, in step 2, divide this area coverage task into path planning and path optimization parts, formulate the problem as a bilevel programming framework, use a bio-inspired neural network algorithm in the upper-layer algorithm to solve the coverage path, calculate the feasible path and transfer the path list to the lower-layer algorithm;

[0010] Step 3: The lower layer uses the ant colony optimization algorithm to solve the energy consumption problem, takes the ocean current in each grid as a constraint condition, searches for the optimal propulsion speed of each feasible path in the feasible path list in step 2, and calculates its energy consumption as the fitness value and returns it to the upper-layer algorithm. If there is no feasible path, find an escape path and return it to the upper-layer algorithm;

[0011] Step 4: The upper-layer algorithm updates the planned path based on the results of the lower-layer algorithm, and iterates in this way until the algorithm termination condition is met, that is, the mission area is completely covered, and the optimal coverage path is output;

[0012] Further, in step 1, the process of processing the known map of the sea area to be operated and constructing an environmental map is as follows:

[0013] Step 1.1 Separate the obstacle area and the area to be operated, and the specific steps are as follows:

[0014] For the underwater area to be detected, it is usually necessary to obtain its topographic map in advance through a multibeam device, perform two-dimensional processing on the selected depth d of the topographic map to obtain a plan view of the area where the AUV is to operate;

[0015] Perform obstacle recognition on the two-dimensional map. Among them, the part greater than the depth value is the ocean, which is a navigable area, and the part less than the depth value is the underwater terrain, which is an obstacle area;

[0016] Step 1.2 Perform binary processing on the image to construct a grid map:

[0017] Assign values to the navigable area and the obstacle area respectively, and process them into a 0-1 map that can be recognized by a computer. According to the operation radius R of the AUV task Draw a grid map with a grid side length of L, as Figure 1 shown in the left figure;

[0018] Step 1.3 Perform dilation processing on the obstacles. The specific steps are as follows:

[0019] For the convenience of designing the algorithm, the present invention regards the AUV to be operated as a mass point that does not occupy space, and appropriately ignores the maneuverability of the AUV in the actual sea area, such as turning radius, driving inertia, etc. If the designed path planning algorithm is directly added to the final AUV control system, it is very likely to cause the AUV to collide with obstacles;

[0020] Based on this, the present invention performs dilation processing on the obstacles: there are N*N pixel blocks in each grid. If the number of obstacle pixel blocks in the grid is greater than N d pieces, then set this grid as an obstacle grid;

[0021] After this step, the obstacle grids completely cover the obstacle area of the sea area to be operated, and the environmental sea chart used for the final path planning is obtained;

[0022] Furthermore, in step 2, the task of covering this area is divided into two parts: path planning and path optimization, and the problem is expressed as a two-layer planning framework. The upper-layer algorithm uses a bio-inspired neural network algorithm to solve the coverage path, calculates the feasible path and transmits the path list to the lower-layer algorithm. The specific process is as follows:

[0023] Step 2.1 Construct the specific implementation framework of this solution. The specific steps are as follows:

[0024] The core goal of this task is to achieve full coverage of the underwater task area through the AUV. This process needs to comprehensively consider multiple indicators such as sea current conditions, energy consumption, turning angle, coverage distance, and time, which is a complex multi-objective optimization problem; considering the battery capacity limit of the task AUV, energy consumption is used as another main measurement target to plan the full-coverage task path with the optimal energy consumption;

[0025] The task of covering this area is divided into two parts: path planning and path optimization. The problem is formulated as a bilevel programming framework and solved by a hybrid algorithm of a bio-inspired neural network algorithm and an ant colony algorithm, as Figure 2 shown;

[0026] The upper-layer algorithm uses a bio-inspired neural network algorithm to plan each step of the path. In this process, the list of optional paths for each step needs to be passed to the lower-layer algorithm, and the fitness value is obtained from the lower-layer algorithm to calculate the path. Finally, a path that can completely cover the task area is obtained;

[0027] The lower-layer algorithm uses an ant colony optimization algorithm to search for an optimal path within the list of feasible paths, including the magnitude and direction of the AUV propulsion speed in each grid area, and calculates its energy consumption as the fitness value to return to the upper-layer algorithm;

[0028] The upper-layer algorithm updates the planned path based on this fitness value and iterates in this way until the algorithm termination condition is met, and then outputs the optimal result path;

[0029] Its technical route is as Figure 2 shown;

[0030] Step 2.2 uses a bio-inspired neural network algorithm to plan a path that can completely cover the task area. The specific steps are as follows:

[0031] Each grid G in the map i corresponds one-to-one with the neurons in the bio-inspired neural network structure. The activity value of the neuron represents the state of the corresponding grid, as Figure 1 shown in the right figure of

[0032] Let x i be the activity value of the i-th neuron. The differential equation of the dynamic neuron activity output value of the i-th neuron can be obtained as:

[0033]

[0034] where the parameters A, B, and D represent the passive decay rate, the upper and lower limits of neural excitation, is the stimulus input of the neuron, is the inhibitory input of the neuron, and I i is the external input of the i-th neuron:

[0035]

[0036] E is a positive constant much larger than B, which can ensure that the neural activity output values of the neurons representing the target point and the neurons representing the obstacles are at the peak and valley values respectively in the entire neural network structure; ω ij represents the connection weight between neuron i and neuron j, ω ij = f(|ij|), and the function f(|ij|) is a decay function, defined as:

[0037]

[0038] j is the grid around i, and its serial number is as shown in the left figure of Figure 1 ; u is a positive constant. The bio-inspired neural network algorithm ensures that the activity value of the target point can spread in the entire activity space, while the neurons at the obstacles can only act locally;

[0039] Step 2.3 makes a path decision according to the neuron activity under specific constraints. The specific steps are as follows:

[0040] Define the planned path list as P; for the operation goal of optimal energy consumption, the AUV should consider factors such as taking the shortest path, reducing the number of turns, and saving energy as much as possible along the ocean current direction when making a path decision. Therefore, in this model, the selection of the next position point of the robot is determined by the neuron activity value of this point and the previous position of the AUV, that is:

[0041] P n →x n = max{x j + ay j + bz j + μe cj , j = 1, 2,..., k}

[0042] Here ay j is the turning angle constraint, a is a positive constant, y j = 1 - Δθ j / π is a monotonically decreasing function of the direction difference between the current position and the next position of the AUV, Δθ j is the turning angle between the current moving direction and the next moving direction. If the current position coordinates of the robot are P c = (x pc , y pc ), the previous position coordinates P p = (x pp , y pp ), and the next moment position coordinates are P j = (x pj , y pj ), then there is:

[0043] Δθ j = |θ j-θ c |

[0044] = arctan(y pj -y pc ,x pj -x pc ) - arctan(y pc -y pp ,x pc -x pp )

[0045] Similarly, bz j is the ocean current constraint, b is a positive constant, z j = 1 - β / π is a monotonically decreasing function of the difference between the next position of the AUV and the ocean current direction within the next position. β is the angle between the forward direction of the AUV in the next grid area and the ocean current direction in the next grid area; Let the ocean current vector within each grid be Its magnitude is c, and the angle between its direction and the north direction is α. In the ocean environment, the vertical current is weaker relative to the cruising speed of the AUV and the horizontal current. In the AUV path planning, the influence of the vertical current on the AUV movement is usually ignored. c x ,c y are the magnitudes of the ocean current on the x and y axes respectively. Then there is:

[0046] β = |θ j - α j |

[0047] = arctan(y pj - y pc ,x pj - x pc ) - arctan(c x ,c y )

[0048] In particular, μe cj is the constraint of the decision vector of the lower-layer algorithm on the upper-layer algorithm. μ is a positive constant, and e cj is the energy consumption from grid i to grid j, which is transmitted back from the lower-layer ant colony algorithm to the upper layer;

[0049] First, calculate the active output values of the neurons in the four directions with serial numbers {2, 4, 6, 8} adjacent to the current neuron, and then find the neuron with the maximum active value. The AUV then starts moving to the position of this neuron with the maximum active output value. When the AUV sails from the current position to the next position, the next position will become the new current position. At this time, the AUV will still select the path according to the strategy given by this formula. The previous position becomes the covered area;

[0050] Further, in step 3, the ant colony optimization algorithm is used to solve the energy consumption problem for the lower layer. Taking the ocean current in each grid as a constraint condition, the optimal propulsion speed of each feasible path is searched in the feasible path list of step 2, and its energy consumption is calculated as the fitness value and returned to the upper layer algorithm. If there is no feasible path, an escape path is searched and returned to the upper layer algorithm. The specific steps are as follows:

[0051] Step 3.1 Determine whether there is a feasible path in the path list of step 2. The specific steps are as follows:

[0052] If there is an optional path in the path selection list transmitted by the upper layer algorithm at this time, go to step 3.2;

[0053] If there is no optional path in the path selection list transmitted by the upper layer algorithm at this time, it means that the AUV is trapped in a dead zone. At this time, go to step 3.3;

[0054] Step 3.2 Search for the optimal propulsion speed in the path list of step 2 through the ant colony optimization algorithm. The specific steps are as follows:

[0055] Step 3.2.1 Establish the objective function:

[0056] Define the energy consumption e cj as the objective function, which is related to the propulsion power P VEH of the underwater robot and the propulsion duration t cj as follows:

[0057]

[0058] In the formula, P VEH is proportional to the cube of the AUV propulsion speed. k is the drag coefficient, and its value is determined by the AUV design; here, there is only movement in the up, down, left, and right directions, so the propulsion path is 1; is the propulsion speed of the AUV during sub-path navigation, is the speed of the AUV relative to the seabed during sub-path navigation; can be obtained through and the ocean current by vector synthesis

[0059]

[0060] The ocean current in adjacent grids does not change violently. Therefore, in the present invention, the ocean current on a section of the path is replaced by so as to reduce the number of changes in the AUV propulsion speed;

[0061] Limit the upper and lower boundaries of the AUV propulsion speed: is 0.3 m / s, is 0.6 m / s; when the AUV actually sails on the sub-path, it has only four moving directions of {2, 4, 6, 8}. Let v g，x , v g，y be the magnitudes of the actual sailing speed on the x-axis and y-axis respectively. Other constraints include

[0062]

[0063] Step 3.2.2 performs parameter optimization through the ant colony optimization algorithm:

[0064] Take the propulsion speed of the AUV when sailing on the sub-path obtained by calculating the energy consumption as the variable to be optimized, which includes two parameters: magnitude and direction;

[0065] For the convenience of using the ant colony algorithm, these two parameter values are abstractly represented on the XOY plane. Limit its magnitude to a total of 7 values of {0.3, 0.35,..., 0.6}, and limit its angle to a total of 360 values of {0°, 1°,..., 359°};

[0066] Draw two equally spaced line segments perpendicular to the x-axis. Divide the first line segment l 1 into six equal parts to obtain 7 nodes, representing the magnitude of the propulsion speed; Divide the second line segment l 2 into 359 equal parts to obtain 360 nodes, representing the direction of the propulsion speed; Let the symbol K(x i , y j ) represent a node, x i is the abscissa of the line segment l i , y j is the ordinate of the line segment l i . K(x 0 , y 0 ) represents the origin of coordinates, i.e., the starting point. Then K(x 1 , y 3 ) represents that the magnitude of the propulsion speed is 0.4 m / s;

[0067] Suppose an ant crawls from any x i to x i+1 in equal time, regardless of the distance between nodes. All ants start from the origin of coordinates K(x 0 , y 0 ), then they will reach the line segment l 1 simultaneously, and reach the end point on the line segment l 2 simultaneously, completing one cycle;

[0068] Suppose τ ij represents at the node K(x i , y jThe amount of information remaining on it. At the initial moment, the amount of information on each node is the same. Let denote the probability that the k-th ant crawls from any point x i-1 to node K(x i , y j ), which is expressed as

[0069]

[0070] In the formula, λ is the information heuristic factor, indicating the role of pheromone in the movement of ants; μ is the expected heuristic factor, indicating the role of heuristic information in the movement of ants; η ij denotes the heuristic information of node K(x i , y j ), which is defined here as

[0071]

[0072] In the formula, takes values as follows: In the first loop, In subsequent loops, is the ordinate value corresponding to the optimal path (optimal performance index) generated in the previous loop;

[0073] The amount of information τ i , y j ) on the path node K(x at time t + 1 ij (t + 1) can be adjusted according to the following rules:

[0074] τ ij (t + 1) = (1 - ρ)·τ ij (t) + Δτ ij (t)

[0075]

[0076] In the formula, ρ represents the pheromone evaporation coefficient, then 1 - ρ represents the pheromone residue factor. To prevent the infinite accumulation of information, the value range of ρ is: Δτ ij (t) represents the information increment on the path node K(x i , y j ) in this loop. At the initial moment, Δτ ij (0) = 0; denotes the amount of information left by the k-th ant on the path node K(x i , y j ) in this loop, which can be expressed as

[0077]

[0078] Among them, C represents the pheromone intensity at (i, j), and E k represents the objective function value of the k-th ant in this cycle. After the ants complete a cycle, the pheromones on all paths are updated. When all the ants have converged to obtain an optimal path, the algorithm ends, and the optimal path of the ant colony algorithm, its corresponding optimal propulsion speed magnitude and direction are output;

[0079] Calculate the optimal propulsion speed magnitude and direction corresponding to the optional paths in the upper-layer algorithm path selection list in sequence, and send this optimal propulsion speed magnitude and direction as well as the corresponding energy consumption list back to the upper-layer algorithm;

[0080] Step 3.3 queries the path points that have not been covered in the path list through the ant colony optimization algorithm, calculates the escape path with the minimum fitness value from the current path point to the uncovered path point through the ant colony optimization algorithm, and returns this path sequence and the propulsion speed list to the upper-layer algorithm. The specific steps are as follows:

[0081] Similarly, take the path energy consumption as its objective function. At this time is the length of the escape path, expressed as the sum of the lengths of each section of the path; set that when searching for the escape path, the covered area can move from the current grid to the surrounding 8 grids. When moving in the four directions of {2, 4, 6, 8}, the distance is 1 unit, and when moving in the four directions of {1, 3, 5, 7}, the distance is 1.4;

[0082] The transition probability of the defined ant k on the grid map is:

[0083]

[0084] Among them, T k represents the path point that the k-th ant can choose next, and η ij represents the distance between two adjacent grids:

[0085]

[0086] After one cycle ends, the amount of information on the ant movement path can be adjusted according to the following rules:

[0087] τ ij (t + 1) = (1 - ρ)·τ ij (t) + Δτ ij (t)

[0088]

[0089]

[0090] Among them, E kIt represents the objective function value of the k-th ant in this iteration. After the ant completes one iteration, the pheromone on all paths is updated. When all ants have converged to obtain an optimal path, the algorithm ends, and the optimal path of the ant colony algorithm, its corresponding escape path, and the list of propulsion speeds are output;

[0091] The escape path and the list of propulsion speeds corresponding to the path are transmitted back to the upper-layer algorithm;

[0092] Furthermore, in step 4, the upper-layer algorithm updates the planned path based on the results of the lower-layer algorithm, and iterates in this way until the task area is completely covered, and the optimal coverage path is output. The specific steps are as follows:

[0093] Step 4.1 The lower-layer algorithm transmits the list of fitness values. The specific steps are as follows:

[0094] The upper-layer bio-inspired neural network algorithm calculates the neuron activity value through the optimal propulsion speed magnitude and direction transmitted by the lower-layer ant colony algorithm and the corresponding energy consumption list, determines the next position, updates the planned path, adds this path to the coverage path list, and uses this position as the starting point for the next iteration;

[0095] Step 4.2 The lower-layer algorithm shows that the AUV is trapped in a dead zone. The specific steps are as follows:

[0096] If the AUV is trapped in a dead zone, the lower-layer algorithm does not have a corresponding list of fitness values, and what is transmitted to the upper layer is the escape path and the list of propulsion speeds corresponding to the path. At this time, the upper-layer algorithm adds this escape path to the coverage path list and uses the end point of the escape path as the starting point of the upper-layer algorithm for the next iteration;

[0097] When performing iteration, the upper-layer bio-inspired neural network algorithm does not need to calculate the neuron activity value, directly transmits the list of selectable paths at the current point to the lower-layer algorithm, and updates the coverage path list in the manner of step 4.1;

[0098] The planned path is updated through the orderly combination of step 4.1 and step 4.2, and iterates in this way until the algorithm termination condition is met, and the optimal coverage path is output.

[0099] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0100] 1. The present invention constructs a two-layer planning framework based on the hybrid solution of bio-inspired neural network algorithm and ant colony algorithm to solve the coverage path planning problem;

[0101] 2. The present invention establishes the interdependent relationship among the ocean current vector, turning angle, and energy consumption, thereby dealing with the interaction between the propulsion speed vector and the coverage path selection, and achieving the beneficial combination of the ocean current vector and the coverage task.

[0102] 3. The present invention enables the AUV to successfully complete the underwater area exploration task, while reducing the turning angle, making the best use of the widely existing ocean currents as much as possible, effectively saving energy, and realizing a more efficient and cost-effective coverage plan. BRIEF DESCRIPTION OF THE DRAWINGS

[0103] Figure 1 It is a schematic diagram of the grid map of the task area (left figure) and the structure of the bio-inspired neural network (right figure) provided by the present invention;

[0104] Figure 2 It is the technical route of the underwater area coverage task provided by the present invention;

[0105] Figure 3 It is the flowchart of the coverage algorithm based on the bilevel programming framework provided by the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0106] The present invention will be further clarified below in conjunction with the drawings and specific embodiments. It should be understood that the following specific embodiments are only used to illustrate the present invention and not to limit the scope of the present invention. It should be noted that the terms "front", "rear", "left", "right", "upper" and "lower" used in the following description refer to the directions in the drawings, and the terms "inner" and "outer" respectively refer to the directions towards or away from the geometric center of a specific component.

[0107] Embodiment 1: An energy consumption-optimal underwater area coverage method based on a bilevel programming framework under the influence of ocean currents according to the present invention, the specific implementation process of its algorithm is as Figure 3 shown, and it includes the following steps:

[0108] Step 1: Process the known map of the sea area to be operated, and construct an environmental map, which includes separating the obstacle area and the area to be operated, binarizing the image to construct a grid map, and dilating the obstacles.

[0109] Step 1.1 Separate the obstacle area and the area to be operated, and the specific steps are as follows:

[0110] For the underwater area to be detected, usually, its topographic map needs to be obtained in advance through a multibeam device, and the topographic map is two-dimensionally processed at a selected depth d to obtain a plan view of the area to be operated by the AUV;

[0111] Identify obstacles in the two-dimensional map. Among them, the part greater than the depth value is the ocean, which is a navigable area, and the part less than the depth value is the underwater terrain, which is the obstacle area;

[0112] Step 1.2 Binarize the image and construct a grid map:

[0113] Assign values to the navigable area and the obstacle area respectively, and process them into a 0-1 map recognizable by a computer. According to the operation radius R of the AUV task Draw a grid map with a grid side length of L, as Figure 1 shown in the left figure;

[0114] Step 1.3 Expand the obstacles. The specific steps are as follows:

[0115] For the convenience of designing the algorithm, the present invention regards the AUV to be operated as a mass point that does not occupy space, and appropriately ignores the maneuverability of the AUV in the actual sea area, such as turning radius, driving inertia, etc. If the designed path planning algorithm is directly added to the final AUV control system, it is very likely that the AUV will collide with obstacles;

[0116] Based on this, the present invention expands the obstacles: there are N*N pixel blocks in each grid. If the number of obstacle pixel blocks in the grid is greater than N d pieces, then set this grid as an obstacle grid;

[0117] After this step, the obstacle grids completely cover the obstacle area of the sea area to be operated, and the environmental sea chart used for the final path planning is obtained;

[0118] Step 2: Divide the regional coverage task into path planning and path optimization parts, formulate the problem as a two-layer planning framework, use a bio-inspired neural network algorithm in the upper-layer algorithm to solve the coverage path, calculate the feasible path and transmit the path list to the lower-layer algorithm;

[0119] Step 2.1 Construct the specific implementation framework of this solution. The specific steps are as follows:

[0120] The core goal of this task is to achieve full coverage of the underwater task area through the AUV. This process needs to comprehensively consider multiple indicators such as sea current conditions, energy consumption, turning angle, coverage distance, and time, which is a complex multi-objective optimization problem; considering the battery capacity limitation of the task AUV, take energy consumption as another main measurement goal and plan the full-coverage task path with the optimal energy consumption;

[0121] Divide the regional coverage task into path planning and path optimization parts, formulate the problem as a two-layer planning framework, and solve it through a hybrid of a bio-inspired neural network algorithm and an ant colony algorithm, as Figure 2 shown;

[0122] The upper - layer algorithm uses a bio - inspired neural network algorithm to plan each step of the path. In this process, the list of optional paths for each step needs to be passed to the lower - layer algorithm, and the fitness value is obtained from the lower - layer algorithm to calculate the path. Finally, a path that can completely cover the task area is obtained;

[0123] The lower - layer algorithm uses the ant colony optimization algorithm to search for an optimal path within the list of feasible paths, including the magnitude of the AUV propulsion speed and the driving direction in each grid area, and calculates its energy consumption as the fitness value to return to the upper - layer algorithm;

[0124] The upper - layer algorithm updates the planned path based on this fitness value, and iterates in this way until the algorithm termination condition is met, and outputs the optimal result path;

[0125] Its technical route is as Figure 2 shown;

[0126] Step 2.2 uses a bio - inspired neural network algorithm to plan a path that can completely cover the task area. The specific steps are as follows:

[0127] Each grid G in the map i corresponds one - to - one with the neurons in the bio - inspired neural network structure. The activity value of the neuron represents the state of the corresponding grid, as Figure 1 shown in the right figure in

[0128] Let x i be the activity value of the i - th neuron. The differential equation for the dynamic neuron activity output value of the i - th neuron can be obtained as:

[0129]

[0130] Among them, the parameters A, B, and D represent the passive attenuation rate, the upper and lower limits of neural excitation respectively, is the stimulus input of the neuron, is the inhibitory input of the neuron, I i is the external input of the i - th neuron:

[0131]

[0132] E is a positive constant much larger than B, which can ensure that the neural activity output values of the neurons representing the target point and the neurons representing the obstacles are at the peak and valley respectively in the entire neural network structure; ω ij represents the connection weight between neuron i and neuron j, ω ij = f(|ij|), and the function f(|ij|) is a decay function, defined as:

[0133]

[0134] j is the grid around i, and its serial number is as follows Figure 1 As shown in the middle left figure; u is a positive constant, the biologically inspired neural network algorithm ensures that the activity value of the target point can be propagated throughout the activity space, while the neurons at the obstacle can only act locally;

[0135] Step 2.3 makes path decisions based on the neuron activity under specific constraints. The specific steps are as follows:

[0136] Define the path list obtained by planning as P; in order to achieve the optimal energy consumption, the AUV should consider factors such as taking the shortest path, reducing the number of turns, and trying to save energy along the direction of the ocean current when making path decisions. Therefore, in this model, the robot's next position x n The choice of the neuron activity value x at that point j and the previous position of the AUV, namely:

[0137] P n →x n =max{x j +ay j +bz j +μe cj , j = 1, 2, ..., k}

[0138] Here j is the rotation constraint, a is a positive constant, y j =1-Δθ j / π is a monotonically decreasing function of the difference between the current position of the AUV and the direction of the next position, Δθ j It is the angle between the current moving direction and the next moving direction. If the current position coordinate of the robot is P c =(x pc ,y pc ), the previous position coordinate P p =(x pp ,y pp ), the position coordinate at the next moment is P j =(x pj ,y pj ), then we have:

[0139] Δθ j =|θ j -θ c |

[0140] =arctan(y pj -y pc , x pj -x pc )-arctan(y pc -y pp , xpc -x pp )

[0141] Similarly, bz j is the ocean current constraint, b is a positive constant, and z j = 1 - β / π is a monotonically decreasing function of the difference between the direction of the ocean current in the next position of the AUV and the next position, and β is the angle between the forward direction of the AUV in the next grid area and the direction of the ocean current in the next grid area; assume that the ocean current vector in each grid is its magnitude is c, and the angle between its direction and the north direction is α. In the ocean environment, the vertical current is weaker than the AUV cruising speed and the horizontal current. In the AUV path planning, the influence of the vertical current on the AUV movement is usually ignored. c x , c y are the magnitudes of the ocean current on the x and y axes respectively, then there are:

[0142] β = |θ j - α j |

[0143] = arctan(y pj - y pc , x pj - x pc ) - arctan(c x , c y )

[0144] Particularly, μe cj is the constraint of the lower-layer algorithm decision vector on the upper-layer algorithm, μ is a positive constant, and e cj is the energy consumption from grid i to grid j, which is transmitted back to the upper layer by the lower-layer ant colony algorithm;

[0145] First, calculate the active output values of the neurons in the four directions {2, 4, 6, 8} adjacent to the current neuron, and then find the neuron with the maximum active value. The AUV then starts moving to the position of this neuron with the maximum active output value. When the AUV sails from the current position to the next position, the next position will become the new current position. At this time, the AUV will still select the path according to the strategy given by this formula. The previous position becomes the covered area;

[0146] Step 3: The lower layer uses the ant colony optimization algorithm to solve the energy consumption problem. Taking the ocean current in each grid as a constraint condition, search for the optimal propulsion speed of each feasible path in the feasible path list in Step 2, and calculate its energy consumption as the fitness value and return it to the upper layer algorithm. If there is no feasible path, search for an escape path and return it to the upper layer algorithm;

[0147] Step 3.1 Determine whether there is a feasible path in the path list in Step 2. The specific steps are as follows:

[0148] If there is an optional path in the path selection list transmitted by the upper-layer algorithm at this time, go to step 3.2;

[0149] If there is no optional path in the path selection list transmitted by the upper-layer algorithm at this time, it means that the AUV has fallen into a dead zone, and go to step 3.3 at this time;

[0150] Step 3.2 searches for the optimal propulsion speed in the path list of step 2 through the ant colony optimization algorithm. The specific steps are as follows:

[0151] Step 3.2.1 Establish the objective function:

[0152] Define the energy consumption e cj as the objective function, which is related to the propulsion power P VEH of the underwater robot and the propulsion duration t cj as follows:

[0153]

[0154] In the formula, P VEH is proportional to the cube of the AUV propulsion speed. k is the drag coefficient, and its value is determined by the AUV design; here, there is only movement in the up, down, left, and right directions, so the propulsion path is 1; is the propulsion speed of the AUV during sub-path navigation, is the speed of the AUV relative to the seabed during sub-path navigation; can be obtained through and the ocean current by vector synthesis

[0155]

[0156] The ocean current in adjacent grids does not change violently, so the present invention uses to replace the ocean current on a section of the path, which can reduce the number of changes in the AUV propulsion speed;

[0157] Limit the upper and lower boundaries of the AUV propulsion speed: is 0.3 m / s, is 0.6 m / s; the AUV has only four moving directions of {2, 4, 6, 8} during actual sub-path navigation. Let v g，x , v g，y be the magnitudes of the actual navigation speed on the x and y axes respectively. Other constraint conditions include

[0158]

[0159] Step 3.2.2 Perform parameter optimization through the ant colony optimization algorithm:

[0160] The propulsion speed of the parameter AUV obtained by energy consumption during sub-path navigation As the variable to be optimized, it includes two parameters: magnitude and direction;

[0161] For the convenience of using the ant colony algorithm, these two parameter values are abstractly represented on the XOY plane. The magnitude is restricted to 7 values: {0.3, 0.35,..., 0.6}, and the angle is restricted to 360 values: {0°, 1°,..., 359°};

[0162] Draw two line segments with equal spacing and perpendicular to the X-axis. Divide the first line segment l 1 into six equal parts to obtain 7 nodes, representing the magnitude of the propulsion speed; divide the second line segment l 2 into 359 equal parts to obtain 360 nodes, representing the direction of the propulsion speed; let the symbol K(x i , y j ) represent a node, where x i is the abscissa of the line segment l i , and y j is the ordinate of the line segment l i . K(x 0 , y 0 ) represents the origin of coordinates, i.e., the starting point. Then K(x 1 , y 3 ) represents a propulsion speed of 0.4 m / s;

[0163] Suppose an ant crawls from any x i to x i+1 in equal time, regardless of the distance between nodes. All ants start from the origin of coordinates K(x 0 , y 0 ). Then they will reach the line segment l 1 simultaneously and reach the end point on the line segment l 2 simultaneously, completing one cycle;

[0164] Let τ ij represent the amount of information left at the node K(x i , y j ). At the initial moment, the amount of information at each node is the same. Let represent the probability that the k-th ant crawls from any x i-1 point to the node K(x i , y j ), which is expressed as

[0165]

[0166] In the formula, λ is the information heuristic factor, representing the role of pheromone in the movement of ants; μ is the expected heuristic factor, representing the role of heuristic information in the movement of ants; η ij represents the heuristic information of node K(x i , y j ), which is defined here as

[0167]

[0168] In the formula, takes values as follows: In the first loop, In subsequent loops, is the ordinate value corresponding to the optimal path (optimal performance index) generated in the previous loop;

[0169] The amount of pheromone τ i , y j ) on the path node K(x ij (t + 1) can be adjusted according to the following rules:

[0170] τ ij (t + 1) = (1 - ρ)·τ ij (t) + Δτ ij (t)

[0171]

[0172] In the formula, ρ represents the pheromone evaporation coefficient, then 1 - ρ represents the pheromone residue factor. To prevent the infinite accumulation of information, the value range of ρ is: Δτ ij (t) represents the information increment on the path node K(x i , y j ) in this loop. At the initial moment, Δτ ij (0) = 0; represents the amount of pheromone left by the k-th ant on the path node K(x i , y j ) in this loop, which can be expressed as

[0173]

[0174] Among them, C represents the pheromone intensity at (i, j), E k represents the objective function value of the k-th ant in this loop. After the ants complete a loop, the pheromone on all paths is updated. When all ants have converged to obtain an optimal path, the algorithm ends, and the optimal path of the ant colony algorithm and its corresponding optimal propulsion speed magnitude and direction are output;

[0175] Calculate the optimal propulsion speed magnitude and direction corresponding to the selectable paths in the upper-layer algorithm path selection list in sequence, and transmit this optimal propulsion speed magnitude and direction as well as the corresponding energy consumption list back to the upper-layer algorithm;

[0176] Step 3.3 queries the path points that have not been covered in the path list through the ant colony optimization algorithm, calculates the escape path with the minimum fitness value from the current path point to the uncovered path points through the ant colony optimization algorithm, and returns this path sequence and the propulsion speed list to the upper-layer algorithm. The specific steps are as follows:

[0177] Similarly, take the path energy consumption as its objective function. At this time is the escape path length, expressed as the sum of the lengths of each section of the path; set that the covered area can move from the current grid to the surrounding 8 grids when finding the escape path. When moving in the four directions of {2, 4, 6, 8}, the distance is 1 unit, and when moving in the four directions of {1, 3, 5, 7}, the distance is 1.4;

[0178] The transition probability of the defined ant k on the grid map is:

[0179]

[0180] where T k represents the path points that ant k can choose next, and η ij represents the distance between two adjacent grids:

[0181]

[0182] After one cycle ends, the information amount of the ant movement path can be adjusted according to the following rules:

[0183] τ ij (t + 1) = (1 - ρ)·τ ij (t) + Δτ ij (t)

[0184]

[0185]

[0186] where E k represents the objective function value of the kth ant in this cycle. After the ant completes one cycle, update the pheromone on all paths. When all ants have converged to obtain an optimal path, the algorithm ends, and output the optimal path of the ant colony algorithm and its corresponding escape path and propulsion speed list;

[0187] Transmit this escape path and the propulsion speed list of the corresponding path back to the upper-layer algorithm;

[0188] Step 4: The upper-layer algorithm updates the planned path based on the results of the lower-layer algorithm, and iterates in this way until the task area is completely covered, and outputs the optimal coverage path;

[0189] Step 4.1 The lower-layer algorithm transmits the fitness value list, and the specific steps are as follows:

[0190] The upper-layer bio-inspired neural network algorithm calculates the neuron activity value based on the optimal propulsion speed magnitude and direction transmitted by the lower-layer ant colony algorithm and the corresponding energy consumption list, obtains the next position, updates the planned path, adds this path to the coverage path list, and uses this position as the starting point for the next iteration;

[0191] Step 4.2 The lower-layer algorithm indicates that the AUV is in a dead zone, and the specific steps are as follows:

[0192] If the AUV is in a dead zone, the lower-layer algorithm has no corresponding fitness value list, and what is transmitted to the upper layer is the escape path and the propulsion speed list of the corresponding path. At this time, the upper-layer algorithm adds this escape path to the coverage path list and uses the end point of the escape path as the starting point of the upper-layer algorithm for the next loop iteration;

[0193] When iterating, the upper-layer bio-inspired neural network algorithm does not need to calculate the neuron activity value, directly transmits the list of selectable paths at the current point to the lower-layer algorithm, and updates the coverage path list in the manner of Step 4.1;

[0194] The planned path is updated through the orderly combination of Step 4.1 and Step 4.2, and iterates in this way until the algorithm termination condition is met, and the optimal coverage path is output.

[0195] The technical means disclosed in the solution of the present invention are not limited to the technical means disclosed in the above embodiments, but also include technical solutions composed of any combination of the above technical features.

Claims

1. An energy consumption optimal underwater area coverage method based on a bilevel programming framework under the influence of ocean currents, characterized in that, the method comprises the following steps: Step 1: Process the known map of the sea area to be operated to construct an environmental map, which includes separating the obstacle area and the area to be operated, binary processing the image to construct a grid map, and dilating the obstacle; Step 2: In Step 2, divide the area coverage task into path planning and path optimization, formulate the problem as a bilevel programming framework, use the bio-inspired neural network algorithm in the upper-level algorithm to solve the coverage path, calculate the feasible path and transmit the path list to the lower-level algorithm, specifically as follows, Step 2.1 Construct the specific implementation framework of this solution, Step 2.2 Use the bio-inspired neural network algorithm to plan a path that can completely cover the task area. Each grid G in the map i corresponds one-to-one with the neurons in the bio-inspired neural network structure. The activity value of the neuron represents the state of the corresponding grid. Step 2.3 Make path decisions according to the neuron activity under specific constraints, and the specific steps are as follows: Define the path list obtained by the planning as P; for the job objective of optimal energy consumption, the next position point x of the robot n is selected by the neuron activity value x of this point j and the previous position of the underwater robot, that is: P n →x n = max{x j + ay j + bz j + μe cj , j = 1, 2, …, k} Here ay j is the corner constraint, a is a positive constant, y j = 1 - Δθ j / π is a monotonically decreasing function of the difference in direction between the current position and the next position of the underwater robot, Δθ j is the angle between the current movement direction and the next movement direction. If the current position coordinates of the robot are P c = (x pc , y pc ), the previous position coordinates P p = (x pp , y pp ), and the position coordinates at the next moment are P j = (x pj , y pj ), then there is: Δθ j = |θ j - θ c | = arctan(y pj -y pc , x pj -x pc ) - arctan(y pc -y pp , x pc -x pp ) Similarly, bz j is the ocean current constraint, b is a positive constant, and z j = 1 - β / π is a monotonically decreasing function of the difference between the next position of the underwater robot and the ocean current direction within the next position. β is the angle between the forward direction of the underwater robot in the next grid area and the ocean current direction in the next grid area; assume that the ocean current vector within each grid is with magnitude c and the angle between its direction and the north direction being α. In the ocean environment, the vertical current is weaker compared to the cruising speed of the underwater robot and the horizontal current. In the path planning of the underwater robot, the influence of the vertical current on the movement of the underwater robot is usually ignored. c x , c y are the magnitudes of the ocean current in the x and y axes respectively, then we have: β = |θ j - α j | = arctan(y pj -y pc , x pj -x pc ) - arctan(c x , c y ) μe cj is the constraint of the lower-layer algorithm decision vector on the upper-layer algorithm. μ is a positive constant, and e cj is the energy consumption from grid i to grid j, which is passed back from the lower-layer ant colony algorithm to the upper layer; First, calculate the activity output values of the neurons in the four directions with serial numbers {2, 4, 6, 8} adjacent to the current neuron, then find the neuron with the maximum activity value, and the underwater robot starts to move to the position of this neuron with the maximum activity output value. When the underwater robot sails from the current position to the next position, the next position will become the new current position. At this time, the underwater robot will still select the path according to the strategy given by this formula, and the previous position becomes the covered area; Step 3: The lower layer uses the ant colony optimization algorithm to solve the energy consumption problem, takes the ocean current in each grid as a constraint condition, searches for the optimal propulsion speed of each feasible path in the feasible path list in Step 2, and calculates its energy consumption as the fitness value and returns it to the upper-level algorithm. If there is no feasible path, find an escape path and return it to the upper-level algorithm; specifically as follows, Step 3.1 Judge whether there is a feasible path in the path list in Step 2, and the specific steps are as follows: If there is an optional path in the path selection list transmitted by the upper-level algorithm at this time, go to Step 3.2; If there is no optional path in the path selection list transmitted by the upper-level algorithm at this time, it means that the underwater robot has fallen into a dead zone, and at this time, go to Step 3.3; Step 3.2 Search for the optimal propulsion speed in the path list in Step 2 through the ant colony optimization algorithm, Step 3.3 Query the path points that have not been covered in the path list through the ant colony optimization algorithm, calculate the escape path with the minimum fitness value from the current path point to the uncovered path point through the ant colony optimization algorithm, and return this path sequence and the propulsion speed list to the upper-level algorithm, Step 4: The upper-level algorithm updates the planned path based on the results of the lower-level algorithm, and iterates in this way until the algorithm termination condition is met, that is, the task area is completely covered, and the optimal coverage path is output.

2. The energy consumption optimal underwater area coverage method based on a bilevel programming framework under the influence of ocean currents according to claim 1, characterized in that, in Step 1, the process of processing the known map of the sea area to be operated to construct an environmental map is as follows: Step 1.1 Separate the obstacle area and the area to be operated, and the specific steps are as follows: For the underwater area to be detected, it is necessary to obtain its topographic map in advance through a multibeam device, perform two-dimensional processing on the selected depth d of the topographic map to obtain a plan view of the area where the underwater robot is to operate; Identify obstacles in the two-dimensional map. Among them, the part greater than the depth value is the ocean, which is a navigable area, and the part less than the depth value is the underwater terrain, which is an obstacle area; Step 1.2 Perform binary processing on the image to construct a grid map: Assign values to the navigable area and the obstacle area respectively, and process them into a 0-1 map recognizable by a computer. According to the operation radius R of the underwater robot task Draw a grid map with a grid side length of L Step 1.3 Perform dilation processing on the obstacles. The specific steps are as follows: Regarding the underwater robot to be operated as a point mass that does not occupy space, perform dilation processing on obstacles: there are N*N pixel blocks in each grid. If the number of obstacle pixel blocks in the grid is greater than N d pieces, then set this grid as an obstacle grid; After this step, the obstacle grid completely covers the obstacle area of the sea area to be operated, and an environmental sea chart for the final path planning is obtained.

3. According to the method for optimal energy consumption underwater area coverage based on a two-layer planning framework under the influence of ocean currents as described in claim 1, characterized in that, The specific steps of step 2.1 are as follows: Divide the regional coverage task into path planning and path optimization parts, formulate the problem as a two-layer planning framework, and solve it through a hybrid algorithm of a bio-inspired neural network algorithm and an ant colony algorithm. The upper-layer algorithm uses a bio-inspired neural network algorithm to plan each step of the path. In this process, the list of optional paths for each step needs to be passed to the lower-layer algorithm, and the fitness value is obtained from the lower-layer algorithm to calculate the path, and finally a path that can completely cover the task area is obtained; The lower-layer algorithm uses an ant colony optimization algorithm to search for an optimal path within the list of feasible paths, including the magnitude and driving direction of the underwater robot's propulsion speed in each grid area, and calculates its energy consumption as the fitness value and returns it to the upper-layer algorithm; The upper-layer algorithm updates the planned path based on this fitness value, and iterates in this way until the algorithm termination condition is met, and outputs the optimal result path; The specific steps of step 2.2 are as follows: Each grid G in the map i corresponds one-to-one spatially with the neurons in the bio-inspired neural network structure, and the activity value of the neuron represents the state of the corresponding grid. Let x i be the activity value of the i-th neuron. The differential equation for obtaining the dynamic neuron activity output value of the i-th neuron is as follows: Among them, the parameters A, B, and D respectively represent the passive attenuation rate, the upper and lower limits of neural excitation, is the stimulus input to the neuron, is the inhibitory input to the neuron, I i is the external input of the i-th neuron: E is a positive constant much larger than B, ensuring that the neural activity output values of the neurons representing the target point and the neurons representing the obstacles are at the peak and valley respectively in the entire neural network structure; ω ij represents the connection weight between neuron i and neuron j, ω ij = f(|ij|), and the function f(|ij|) is a decay function, defined as: j is the grid around i, and u is a positive constant.

4. According to the method for optimal energy consumption underwater area coverage based on a two-layer planning framework under the influence of ocean currents as described in claim 1, characterized in that, The specific steps of step 3.2 are as follows: Step 3.2.1 Establish the objective function: Define the energy consumption e cj as the objective function, which is related to the propulsion power P VEH of the underwater robot and the propulsion duration t cj as follows: where P VEH is proportional to the cube of the propulsion speed of the underwater robot. k is the drag coefficient, and its value is determined by the design of the underwater robot. Since there is only up, down, left, and right movement here, the propulsion path is unit 1; is the propulsion speed of the underwater robot during sub-path navigation, is the speed of the underwater robot relative to the seabed during sub-path navigation; can be obtained through and the ocean current by vector synthesis The ocean currents in adjacent grids do not change drastically, so the ocean currents along a path are represented by to reduce the number of changes in the propulsion speed of the underwater robot; Limit the upper and lower boundaries of the underwater robot's propulsion speed: is 0.3 m / s, is 0.6 m / s; when the underwater robot actually sails on the sub-path, there are only four moving directions {2, 4, 6, 8}, let v g,x , v g,y be the magnitudes of the actual sailing speed on the x and y axes respectively, and other constraint conditions include Step 3.2.2 Perform parameter optimization through the ant colony optimization algorithm: The propulsion speed of the underwater robot when obtaining the parameters of energy consumption during sub-path navigation As the variable to be optimized, it includes two parameters: magnitude and direction; For the convenience of using the ant colony algorithm, these two parameter values are abstractly represented on the XOY plane, and their magnitudes are restricted to 7 values, namely {0.3, 0.35, …, 0.6}, and their angles are restricted to 360 values, namely {0°, 1°, …, 359°}; Draw two line segments with equal spacing and perpendicular to the X-axis. Divide the first line segment l 1 into six equal parts to obtain 7 nodes, representing the magnitude of the propulsion speed; divide the second line segment l 2 into 359 equal parts to obtain 360 nodes, representing the direction of the propulsion speed; let the symbol K(x i , y j ) represent a node, where x i is the abscissa of the line segment l i , and y j is the ordinate of the line segment l i . K(x 0 , y 0 ) represents the origin, i.e., the starting point. Then K(x 1 , y 3 ) represents a propulsion speed of 0.4 m / s; Suppose an ant crawls from any x i to x i+1 in equal time, regardless of the distance between nodes. All ants start from the origin K(x 0 , y 0 ) simultaneously, then they will reach the line segment l 1 simultaneously and reach the end point on the line segment l 2 simultaneously, completing one cycle; Let τ ij denote the amount of information remaining at node K(x i , y j ). At the initial moment, the amount of information at each node is the same. Let denote the probability that the k-th ant crawls from any point x i-1 to node K(x i , y j ), which is expressed as where λ is the information heuristic factor, representing the role of pheromone in the movement of ants; μ is the expected heuristic factor, representing the role of heuristic information in the movement of ants; η ij represents the heuristic information of node K(x i ,y j ), which is defined here as In the formula, The value is taken as follows: In the first loop, In each subsequent loop, is the ordinate value corresponding to the optimal path (optimal performance index) generated in the previous loop; The amount of information τ i at time t + 1 on the path node K(x j , y ij )(t + 1) can be adjusted according to the following rules: τ ij (t + 1) = (1 - ρ)·τ ij (t) + Δτ ij (t) In the formula, ρ represents the pheromone evaporation coefficient, then 1 - ρ represents the pheromone residue factor. To prevent the infinite accumulation of information, the value range of ρ is: Δτ ij (t) represents the information increment on the path node K(x i ,y j ) in this iteration. At the initial moment, Δτ ij (0) = 0; represents the amount of pheromone left by the k-th ant on the path node K(x i ,y j ) in this iteration and can be expressed as Among them, C represents the pheromone intensity at (i, j), and E k represents the objective function value of the k-th ant in the current iteration. After all ants complete one iteration, the pheromones on all paths are updated. When all ants have converged to obtain an optimal path, the algorithm ends, and the optimal path of the ant colony algorithm, its corresponding optimal propulsion speed magnitude and direction are output; Calculate the magnitude and direction of the optimal propulsion speed corresponding to the optional paths in the upper-layer algorithm path selection list in turn, and transmit this optimal propulsion speed magnitude and direction and the corresponding energy consumption list back to the upper-layer algorithm; The specific steps of step 3.3 are as follows: Similarly, taking the path energy consumption as its objective function, at this time is the escape path length, expressed as the sum of the lengths of each path segment; it is set that when searching for the escape path, the covered area can move from the current grid to the surrounding 8 grids. When moving in the four directions of {2, 4, 6, 8}, the distance is 1 unit, and when moving in the four directions of {1, 3, 5, 7}, the distance is 1.4; The defined transition probability of ant k on the grid map is as follows: Among them, T k represents the path points that ant k can choose next step, and η ij represents the distance between two adjacent grids: After one cycle ends, the amount of information of the ant movement path can be adjusted according to the following rules: τ ij (t + 1) = (1 - ρ)·τ ij (t) + Δτ ij (t) Among them, E k represents the objective function value of the k-th ant in this iteration. After an ant completes an iteration, the pheromone on all paths is updated. When all ants have converged to obtain an optimal path, the algorithm ends, and the optimal path of the ant colony algorithm, its corresponding escape path, and the list of propulsion speeds are output; Transmit this escape path and the list of propulsion speeds of the corresponding path back to the upper-layer algorithm.

5. According to the method for optimal energy consumption underwater area coverage based on a two-layer planning framework under the influence of ocean currents as described in claim 1, characterized in that, In step 4, the upper-layer algorithm updates the planned path based on the result of the lower-layer algorithm, and iterates in this way until the task area is completely covered, and outputs the optimal coverage path. The specific steps are as follows: Step 4.1 The lower-layer algorithm transmits the fitness value list. The specific steps are as follows: The upper-layer bio-inspired neural network algorithm calculates the neuron activity value based on the optimal propulsion speed magnitude and direction transmitted by the lower-layer ant colony algorithm and the corresponding energy consumption list, determines the next position, updates the planned path, adds this path to the coverage path list, and uses this position as the starting point for the next iteration; Step 4.2 The lower-layer algorithm indicates that the underwater robot is trapped in a dead zone, and the specific steps are as follows: If the underwater robot is trapped in a dead zone, the lower-layer algorithm does not have a corresponding fitness value list, and what is transmitted to the upper layer is the escape path and the propulsion speed list of the corresponding path. At this time, the upper-layer algorithm adds this escape path to the coverage path list and uses the end point of the escape path as the starting point of the upper-layer algorithm for the next loop iteration; During iteration, the upper-layer bio-inspired neural network algorithm does not need to calculate the neuron activity value. It directly transmits the list of selectable paths at the current point to the lower-layer algorithm and updates the coverage path list in the manner of Step 4.1; The planned path is updated through the orderly combination of Step 4.1 and Step 4.2, and this loop iteration is performed until the algorithm termination condition is met, and the optimal coverage path is output.

Images