An energy-saving path planning method for AUV inspection in water tunnels

Abstract

The present invention proposes an energy-saving path planning method for an AUV for water tunnel inspection. First, the tunnel inspection AUV is optimized and planned in a constructed environmental model. Then, a multi-attribute heuristic factor is added to the ant colony algorithm to plan the AUV path. Then, the speed of the AUV is planned to obtain a global path planning for the tunnel inspection AUV. Local dynamic planning is designed for the tunnel inspection AUV, and a method based on a two-layer planning framework is used to switch between global planning and local planning. The planned path and speed are obtained based on the global planning module and the local planning module, and the tunnel inspection AUV is subjected to fixed-depth energy-saving control. Energy-saving and control effects are taken into account in speed control and depth control, thereby completing the energy-saving path planning of the tunnel inspection AUV.

Description

Technical Field

[0001] The present invention belongs to the field of energy-saving technology for tunnel inspection AUVs, and in particular, relates to an energy-saving path planning method for water conveyance tunnel inspection AUVs. Background Art

[0002] Water transfer tunnels are typically located underground or within mountains, characterized by complex structures, harsh environments, and high operational risks. Regular inspections of water transfer tunnels are designed to promptly identify and assess internal defects, eliminate potential hazards that could affect the tunnel's stable operation, and ultimately extend its service life and ensure its normal operation.

[0003] For water supply tunnel inspection, the diver inspection method is not only time-consuming and labor-intensive, but also has a complex and changeable underwater environment with great risks. Although the existing remote-controlled cable robot inspection can obtain corresponding inspection images and data in real time, the umbilical cable is a hindrance and is prone to friction and scraping with the tunnel wall, causing entanglement and breakage. Due to reasons such as power supply voltage drop, the umbilical cable cannot be too long, and the inspection distance is limited.

[0004] The purpose of energy-saving operation planning is to ensure that the tunnel inspection AUV reaches the inspection point safely and achieve energy saving; the traditional ant colony algorithm can only optimize the shortest path, while tunnel inspection not only considers distance but also energy consumption and safety issues. Summary of the Invention

[0005] To address the above issues, the present invention proposes an energy-saving path planning method for AUVs used for water tunnel inspection. It uses energy sources such as electron accelerators, ultraviolet sources, and X-ray sources to charge lunar dust in order to study the charged characteristics of lunar dust. Based on this, the method further studies the impact of charged lunar dust on various devices.

[0006] The present invention is achieved through the following technical solutions:

[0007] An energy-saving path planning method for an AUV for water tunnel inspection: the method specifically comprises the following steps:

[0008] Step 1: Perform optimal planning for the tunnel inspection AUV in the constructed environment model;

[0009] Step 2: Based on the optimization planning in step 1, a multi-attribute heuristic factor is added to the ant colony algorithm to plan the AUV path;

[0010] Step 3: Based on the path planned in step 2, the speed of the AUV is planned to obtain the global path planning of the tunnel inspection AUV.

[0011] Step 4: Design local dynamic planning for the tunnel inspection AUV, and use a method based on a two-layer planning framework to switch between global planning and local planning;

[0012] Step 5: Based on the planned path and speed obtained in Step 4, perform depth and energy-saving control on the tunnel inspection AUV, taking into account both energy conservation and control effects in speed control and depth control, and complete the energy-saving path planning of the tunnel inspection AUV.

[0013] Further, in Step 1,

[0014] Suppose there are M AUVs. The current position of the m-th AUV is i, and its next available target position is j. Each AUV needs a state transition probability to select the position it will visit next. The state transition probability is a function of the current position i and the candidate position j. Specifically, at the t-th iteration, the probability that the m-th AUV transfers from position i to position j is given by the following formula:

[0015]

[0016] where τ ij represents the pheromone concentration between positions i and j at the t-th iteration; η ij represents the heuristic function value, which is usually the reciprocal of factors such as the distance or time between two positions. α is the pheromone concentration. The larger α is, the more AUVs have passed through this section of the path, thus releasing more pheromones, indicating that this path is better. β is the importance of the heuristic function value relative to the state transition probability. The larger β is, the higher the heuristic function value, indicating that this path better meets the preset goals or constraints.

[0017] Introduce a random parameter q to increase the exploration of the algorithm. q is a random number uniformly distributed in the interval [0, 1], and q0 is a constant (0 < q0 < 1). When q < q0, the m-th AUV selects the next target position according to the principle of the maximum state transition probability. That is, when q > q0, the roulette method is used to select the next target position, and the formula is as follows:

[0018]

[0019] where allowed m ={C - tabu m} represents the set of all next positions j that can be reached at the current position i at the current moment; [η ij (t)] = 1 / d jg , d jg represents the distance from the next position j to the final target point g. Planning from the distance magnitude, the shorter this distance is, the more likely this point will be selected;

[0020] After the individual completes a complete optimization, it is necessary to control the influence of pheromones and heuristic functions on the next cycle. The pheromones need to be controlled. This process corresponds to the volatilization of pheromones in nature. After each algorithm iteration, the pheromone control formula is as follows:

[0021]

[0022]

[0023] Where ρ is the pheromone constraint, Q is the starting parameter of pheromone, L m is the path length obtained by the mth AUV after completing the search, V t m is the coordinate set of the path that the m-th AUV passes through in the t+1 optimization.

[0024] Furthermore, in step 2,

[0025] The multi-attribute heuristic factors include distance factor, energy saving factor and safety factor;

[0026] The basic ant colony algorithm was improved and applied to the energy-saving operation planning of tunnel inspection AUVs in tunnel environments. The traditional distance factor heuristic function was improved to enable it to converge to the optimal path faster and reduce the computational complexity of invalid searches at the initial moment. The energy factor and tunnel safety factor were added to the heuristic function so that the searched detection path can meet the requirements of tunnel inspection. The multi-attribute heuristic function was constructed as follows:

[0027] η ij =η l ij ·η e ij ·η c ij (5)

[0028] Among them, η ij is the distance factor, η e ij is the energy consumption factor, η c ij Safety factor

[0029] The path selection in the initial stage mainly depends on the heuristic function. Since no AUV has passed by, no path has pheromones. The guidance of setting the target direction for the initial search is as follows:

[0030]

[0031] Where: L(i) is the distance traveled by the AUV from the current position i; D jdis the length from the next target point j to the final target point g; d jz The sum of the distances between the next target point j and the line connecting the starting point and the final target point; d ij is the length from the current position i to the next target point j;

[0032] Substitute the energy consumption model into the energy saving factor:

[0033]

[0034] When the tunnel inspection AUV does not delay the execution of the task, it is safest to navigate in the center; establish a model from the center point of the tunnel cross section, set the set of center points, that is, the axis of the tunnel as a function of the center position of the safe area, and design a model with decreasing influence outward. as follows:

[0035]

[0036] Among them, l r is the tunnel radius, l ja Indicates the distance from the center of navigation at position j.

[0037] Furthermore, in step 3, speed planning refers to determining the optimal speed of the tunnel inspection AUV at each path point based on the tunnel inspection AUV's dynamic model, environmental parameters, and mission requirements;

[0038] The energy-saving speed planning problem is transformed into a discrete optimization problem, that is, the optimal speed at each path point is selected from a given set of candidate speeds, and the optimization algorithm is used to search for the best combination;

[0039] Use the following steps to perform speed planning based on energy saving:

[0040] Step 3.1: construct a discrete search space based on the given global path and candidate velocity set, and initialize relevant parameters;

[0041] Step 3.2, generate a certain number of virtual AUVs and randomly assign them to different starting positions;

[0042] In step 3.3, each virtual AUV selects the speed at each path point in the candidate speed set according to the probability transfer rule and calculates the corresponding total energy consumption;

[0043] In step 3.4, each virtual AUV updates the pheromone concentration according to its total energy consumption and records its historical best solution;

[0044] Step 3.5, determine whether the termination condition is met, that is, whether the expected position is reached. If so, output the global optimal solution; otherwise, return to step 3.3.

[0045] Furthermore, in step 4,

[0046] A method based on a two-layer planning framework is used to switch between a global static environment and a local dynamic environment; the method based on the two-layer planning framework is divided into a global planning layer and a local planning layer;

[0047] The global planning layer is the global path generated according to steps 1 to 3;

[0048] The local planning layer refers to the local path generated by the feedback control algorithm based on the obstacle avoidance sonar data obtained in real time;

[0049] The following steps are used to adaptively switch the tunnel inspection AUV between the global static environment and the local dynamic environment:

[0050] Step 4.1: Based on the given starting and ending points, as well as underwater environment information, an optimization algorithm is used to generate a global path that meets the constraints and divide it into several segments.

[0051] Step 4.2: The tunnel inspection AUV navigates according to the global path and obtains obstacle avoidance sonar data in real time;

[0052] Step 4.3: If the obstacle avoidance sonar data indicates that there is an urgent obstacle in the current segment, the local planning layer is triggered, otherwise the global planning layer is maintained;

[0053] In step 4.4, the local planning layer uses the dynamic window algorithm (DWA) to generate a temporary path that meets safety requirements and is as close as possible to the original segment direction and speed distribution characteristics based on constraints such as collision avoidance, and sends it to the actuator;

[0054] Step 4.5: The tunnel inspection AUV navigates along the temporary path and continues to acquire obstacle avoidance sonar data.

[0055] In step 4.6, if the sonar data shows that there are no emergency obstacles in the current segment, the route is restored to the original segment; otherwise, the temporary path is maintained.

[0056] Furthermore, the dynamic window consists of three parts: motion model, sampling space and predicted trajectory;

[0057] The motion model is specifically:

[0058] The dynamic window method uses a mass point model, which is modeled as follows:

[0059]

[0060] In the entire sampling space, the speed and angular velocity of the unmanned system are assumed to be constant;

[0061] The sampling space is specifically:

[0062] Constraint 1: The maximum speed limit and maximum angular velocity limit are set according to the parameters of the tunnel inspection AUV, which determines the upper and lower limits of the dynamic window sampling;

[0063] V s ={v∈[v min ,v max ],ω∈[ω min ,ω max ]} (10)

[0064] Constraint 2: Set the linear acceleration and angular acceleration constraints based on the parameters of the tunnel inspection AUV;

[0065]

[0066] Constraint 3: Safe obstacle avoidance distance constraint, which ensures that the tunnel inspection AUV performs maximum acceleration braking at the current speed to avoid collision with the nearest obstacle;

[0067]

[0068] The trajectory prediction

[0069] G(v,ω)=α·zhidao(v,ω)+β·juli(v,ω)+γ·sudu(v,ω) (13)

[0070] Among them, zhidao is the guidance evaluation function, which represents the deviation between the current heading angle and the desired heading angle; α is the guidance function weight; juli represents the distance between the tunnel detection AUV and the nearest obstacle. When the distance between the tunnel detection AUV and the obstacle exceeds a certain value, the function value is set to a constant; β is the safe obstacle avoidance distance weight; sudu is the speed evaluation function, which represents the navigation speed of the tunnel detection AUV. When the constraints are met, the larger the value, the better; γ is the navigation speed weight.

[0071] Furthermore, in step 5,

[0072] An event-triggered mechanism is introduced into the depth control of the underwater vehicle. A trigger function and sliding mode control law are designed. The AUV is only required to update the control rate when an event is triggered, thus achieving energy-saving control of the AUV's depth.

[0073] The synovial membrane control of the event triggering mechanism is specifically as follows:

[0074] Assuming the expected heave speed w d , expected pitch velocity q d , expected pitch angle θd and the expected depth z d are all constants, i.e. Definition: Heave velocity error w e =ww d , pitch angular velocity error q e =qq d , pitch angle error θ e =θ-θ d , depth error z e =zz d ;

[0075]

[0076] The traditional sliding variable is defined by formula (14):

[0077] s(t)=Cx(t) (15)

[0078] The control law u(t) is designed to reach the sliding mode s(t) = 0 in a finite time while always keeping the trajectory of s(t) unchanged;

[0079] Differentiating formula (15), we can obtain:

[0080]

[0081] The control law is designed as follows:

[0082] u(t)=-(CB) -1 {CAx(t)+Ksign[s(t)]} (17)

[0083] The control law adopts the sampling state, and subsequent updates are triggered by events; from formula (15), we can get:

[0084]

[0085] s(t)=Cx(t i ) represents t i Sliding variable at t i+1 At this moment, the control law is updated and remains constant until the next trigger is reached.

[0086] An energy-saving path planning system for AUVs used in water tunnel inspections:

[0087] The system includes a global planning module, a local planning module and a depth control module:

[0088] The global planning module first optimizes the tunnel inspection AUV in the constructed environment model; then adds a multi-attribute heuristic factor to the ant colony algorithm to plan the AUV path; and then plans the AUV's speed to obtain the global path planning of the tunnel inspection AUV.

[0089] The local planning module designs local dynamic planning for the tunnel inspection AUV and switches between global planning and local planning using a method based on a two-layer planning framework;

[0090] The depth control module obtains the planned path and speed based on the global planning module and the local planning module, performs fixed-depth energy-saving control on the tunnel inspection AUV, takes into account both energy saving and control effects in speed control and depth control, and completes the energy-saving path planning of the tunnel inspection AUV.

[0091] An electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.

[0092] A computer-readable storage medium is used to store computer instructions, which implement the steps of the above method when executed by a processor.

[0093] Beneficial effects of the present invention

[0094] This invention improves upon the computationally intensive ant colony algorithm. While traditional ant colony algorithms can only optimize the shortest path, tunnel detection considers not only distance but also energy consumption and safety. To improve the convergence rate of the planning algorithm while navigating the tunnel detection AUV in the most energy-efficient and safest manner, a multi-attribute search algorithm based on a multi-attribute heuristic function was designed. The sum of the distances from the next target point to the starting and end points is added to the distance function, providing guidance on the target direction for the initial optimization search. This addresses the issues of disordered early searches and waste of computing resources, improves convergence rate, and thus achieves energy savings.

[0095] After determining the optimal energy-saving path and speed, the tunnel inspection AUV can track this path by controlling its longitudinal velocity and vertical depth. An event-triggered algorithm reduces computational effort to achieve energy savings, while sliding mode control ensures control accuracy. An event-triggered sliding mode control was designed. First, a control loop was designed to verify its stability and the absence of the Zeno phenomenon. An observer was designed to obtain real-time data. Finally, field experiments with fixed depths were conducted to verify the algorithm's feasibility.

[0096] To improve the endurance of AUVs used in long-distance water tunnel inspections, an event-triggered sliding mode controller was designed. The controller only needs to send instructions to the next module (thrust distribution module) when certain discrete events are triggered. This effectively reduces the controller's energy consumption, achieving energy conservation for the AUV. This controller effectively suppresses disturbances, achieves state control error convergence, and avoids the Zeno phenomenon in the entire system. Compared to simulations using PID and event-triggered control, this controller offers the following advantages:

[0097] (1) Faster. It can reach the desired depth in a short time (10s), while other methods can reach it in 15-35s.

[0098] (2) More accurate. When the desired depth is reached, no steady-state error will occur.

[0099] (3) More stable. When the desired depth is reached, there will be no overshoot.

[0100] (4) Saving computing resources. By transmitting only when certain discrete events are triggered, computing resources are saved. Simulations show a 5.07% energy saving effect, and field tests show a 2.72% energy saving effect.

[0101] The use of tunnel inspection AUV can participate in the inspection of long water transmission tunnels, saving manpower, material resources and energy, and creating direct economic value. BRIEF DESCRIPTION OF THE DRAWINGS

[0102] Figure 1 This is an example diagram of selecting an optional path in a grid map according to the present invention;

[0103] Figure 2 It is a dynamic programming schematic diagram of the present invention;

[0104] Figure 3 Schematic diagram of the tunnel inspection AUV of the present invention;

[0105] Figure 4 Schematic diagram of the tunnel model of the present invention;

[0106] FIG5 (a) is a schematic diagram of the energy-saving operation plan of the present invention; (b) is a top view of the energy-saving operation plan; (c) is a side view of the energy-saving operation plan;

[0107] FIG6 (a) is a schematic diagram of a tunnel inspection AUV safe navigation simulation of the present invention; (b) is a schematic diagram of a tunnel inspection AUV avoiding obstacles and reaching the end point;

[0108] Figure 7 An event-triggered control loop for the present invention;

[0109] Figure 8 is the heave velocity and its observer; Figure 9Update the graph for the algorithm;

[0110] Figure 10 It is a comparison diagram of the control algorithms;

[0111] Figure 11 (a) is the comparison diagram of the control algorithm enlarged (mid-term), (b) is the comparison diagram of the control algorithm enlarged (late-term);

[0112] Figure 12 (a) is the desired velocity diagram, (b) is the x-direction displacement diagram, (c) is the velocity diagram under different controls, and (d) is the velocity error diagram;

[0113] Figure 13 (a) is the speed diagram controlled by different algorithms, and (b) is the horizontal displacement diagram. DETAILED DESCRIPTION

[0114] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0115] The tunnel inspection AUV in this embodiment has an overall length of 2.91 meters, a maximum diameter of 0.60 meters, a width of 1.10 meters including the wings, and a 15° wing opening angle. As shown in the figure, the vessel adopts a spindle-shaped design and is equipped with twelve thrusters. Four main thrusters are located at the bow and stern for forward and reverse movement, while one vertical thruster and one lateral thruster are located at the bow and stern. The two vertical thrusters are responsible for submersible and buoyant movement, while the two lateral thrusters are responsible for lateral movement. A pair of wing panels are installed in the middle of the vessel. These panels adopt a conformal design, and when closed, the entire vessel is streamlined.

[0116] Because exposed rebar and gravel often occur at the tunnel edges, it's safer to navigate as close to the center as possible without compromising detection. Therefore, a safety function was designed as a function of the tunnel radius. This safety function was then combined with speed planning and DWA-based local obstacle avoidance planning to create an energy-efficient operation plan. Finally, simulations were conducted to verify the effectiveness of the proposed algorithm.

[0117] Relevance, diversity, and integrity are the three key system characteristics of an ant colony. Thanks to a collaborative mechanism of positive feedback, the algorithm evolves toward optimal results. Shortening its runtime relies on a distributed, parallel search structure, and the abundance of individuals provides the advantage of diversity. Throughout the path search process, each ant chooses its direction based on the pheromone concentration in the environment. Furthermore, each ant adjusts the amount of pheromone it releases based on a heuristic function. Furthermore, in real environments, pheromones gradually evaporate and decay along the path. Three key factors require consideration: pheromone release strategy, heuristic function design, and pheromone update strategy. These three factors collectively influence the probability of each ant choosing different paths.

[0118] The present invention regards the AUV as a "super" ant with intelligence and perception capabilities, and performs optimal planning in a constructed environmental model.

[0119] An energy-saving path planning system for AUVs used in water tunnel inspections:

[0120] The system includes a global planning module, a local planning module and a depth control module:

[0121] The global planning module first optimizes the tunnel inspection AUV in the constructed environment model; then adds a multi-attribute heuristic factor to the ant colony algorithm to plan the AUV path; and then plans the AUV's speed to obtain the global path planning of the tunnel inspection AUV.

[0122] The local planning module designs local dynamic planning for the tunnel inspection AUV and switches between global planning and local planning using a method based on a two-layer planning framework;

[0123] The depth control module obtains the planned path and speed based on the global planning module and the local planning module, performs fixed-depth energy-saving control on the tunnel inspection AUV, takes into account both energy saving and control effects in speed control and depth control, and completes the energy-saving path planning of the tunnel inspection AUV.

[0124] A method for energy-saving path planning for an AUV for water tunnel inspection, the method specifically comprising the following steps:

[0125] Step 1: Perform optimal planning for the tunnel inspection AUV in the constructed environment model;

[0126] In step 1,

[0127] Assume there are M AUVs, and the current position of the mth AUV is i. Then its next optional target position j has multiple possibilities depending on the environment it is in. For example, in a grid map, Figure 1As shown, when i is not on the edge of the map or near non-navigable areas (obstacles or tunnel walls), the maximum number of candidate target positions j for its next step can reach 26.

[0128] Each AUV needs a state transition probability to select the position it will visit next. The state transition probability is a function of the current position i and the candidate position j. Specifically, at the t-th iteration, the probability that the m-th AUV transfers from position i to position j is given by the following formula:

[0129]

[0130] where τ ij represents the pheromone concentration between positions i and j at the t-th iteration; η ij represents the heuristic function value, usually the reciprocal of factors such as the distance or time between two positions. α is the pheromone concentration. The larger α is, the more AUVs have passed through this section of the path, thus releasing more pheromones, indicating that this path is better; β is the importance of the heuristic function value relative to the state transition probability. The larger β is, the higher the heuristic function value, indicating that this path better meets the preset goals or constraints. Therefore, both aspects need to be considered comprehensively when selecting the next position.

[0131] In addition, in practical applications, a random parameter q is often introduced to increase the exploration of the algorithm. q is a random number uniformly distributed in the interval [0, 1], and q0 is a constant (0 < q0 < 1). When q < q0, the m-th AUV selects the next target position according to the principle of the maximum state transition probability. That is, when q > q0, the roulette method is used to select the next target position, and the formula is as follows:

[0132]

[0133] This can avoid the algorithm falling into a local optimal solution and increase the possibility of new solutions appearing in the search space.

[0134] where allowed m ={C - tabu m} represents the set of all next positions j that can be reached at the current position i at the next step; [η ij (t)] = 1 / d jg , d jg represents the distance from the next position j to the final target point g. Planning from the distance magnitude, the shorter this distance is, the more likely this point will be selected.

[0135] After the individual completes a complete optimization, it is necessary to control the influence of pheromones and heuristic functions on the next cycle. The pheromones need to be controlled. This process corresponds to the volatilization of pheromones in nature. After each algorithm iteration, the pheromone control formula is as follows:

[0136]

[0137]

[0138] Where ρ is the pheromone constraint, Q is the starting parameter of pheromone, L m is the path length obtained by the mth AUV after completing the search, V t m is the coordinate set of the path that the m-th AUV passes through in the t+1 optimization.

[0139] Step 2: Based on the optimization planning in step 1, a multi-attribute heuristic factor is added to the ant colony algorithm to plan the AUV path;

[0140] In step 2, the multi-attribute heuristic factors include a distance factor, an energy saving factor, and a safety factor;

[0141] Most of the existing improvements to the ant colony algorithm are aimed at single-objective planning problems and are not well adapted to tunnel scenarios. For example, when a tunnel inspection AUV performs path planning in a water tunnel environment, it is necessary to comprehensively consider multiple factors, such as rockfall obstacles in the tunnel, full utilization of water flow, energy consumption limitations, safety requirements, etc. Therefore, the basic ant colony algorithm was improved and applied to the energy-saving operation planning of tunnel inspection AUVs in a tunnel environment. The traditional distance factor heuristic function was improved so that it can converge to the optimal path faster and reduce the amount of calculation for invalid searches at the initial moment. The energy factor and tunnel safety factor were added to form a heuristic function so that the searched detection path can meet the needs of tunnel inspection. This can speed up the search for the global optimal solution and improve the quality of the solution and the convergence speed. Based on this, a multi-attribute heuristic search method suitable for energy-saving planning of tunnel inspection AUVs in a tunnel environment was obtained. The multi-attribute heuristic function is constructed as follows:

[0142]

[0143] Among them, η ij is the distance factor, η e ij is the energy consumption factor, η c ij Safety factor

[0144] The path selection in the initial stage mainly depends on the heuristic function. Since no AUV has passed by and no path has pheromones, we optimize the existing foundation to solve the problem of poor optimization effect in the early stage. We set the target direction for the initial search, effectively reducing the cost of the algorithm in the early stage due to aimless optimization. The improved distance heuristic function is as follows:

[0145]

[0146] Where: L(i) is the distance traveled by the AUV from the current position i; D jd is the length from the next target point j to the final target point g; d jz The sum of the distances between the next target point j and the line connecting the starting point and the final target point; d ij is the length from the current position i to the next target point j.

[0147] The improved distance heuristic function effectively enhances the directionality and accuracy of path search, and can guide the AUV to continuously approach the target point purposefully during the movement process, solving the problems of disordered search and waste of computing resources in the early stage, improving the convergence speed, and thus achieving energy saving.

[0148] Substitute the energy consumption model into the energy saving factor:

[0149]

[0150] The presence of falling rocks, debris, etc. in the tunnel will affect the safe navigation of the tunnel inspection AUV. The areas that hinder the tunnel inspection AUV from performing its operating tasks are judged as dangerous areas, and the areas suitable for performing inspection tasks are judged as safe areas.

[0151] Dangerous areas in tunnel environments are mostly distributed at the edge of the tunnel. The safe area is the center point of the tunnel cross section. When the tunnel inspection AUV does not delay the execution of the task, it is safest to navigate in the center. A model is established from the center point of the tunnel cross section. The set of center points, that is, the axis of the tunnel, is a function that decreases outward from the center of the safe area. A model with decreasing influence outward is designed to construct as follows:

[0152]

[0153] Among them, l r is the tunnel radius, l ja It indicates the distance from the center of navigation at the time of position j. It can be seen that the closer to the center of navigation, the larger the heuristic function This enables the AUV to navigate as centrally as possible while successfully performing the inspection mission.

[0154] Step 3: Based on the path planned in step 2, the speed of the AUV is planned to obtain the global path planning of the tunnel inspection AUV.

[0155] In step 3, the tunnel inspection AUV must not only plan a safe, efficient, and energy-efficient path through the complex underwater environment, but also its speed. Speed planning involves determining the optimal speed for each path point based on factors such as the AUV's dynamic model, environmental parameters, and mission requirements. Energy-efficient speed planning effectively utilizes water flow to reduce the AUV's battery consumption while ensuring mission completion, minimizing energy consumption throughout the entire path.

[0156] The energy-saving speed planning problem is transformed into a discrete optimization problem, that is, selecting the optimal speed at each path point from a given set of candidate speeds, and using an optimization algorithm to search for the best combination.

[0157] Use the following steps to perform speed planning based on energy saving:

[0158] In step 3.1, a discrete search space is constructed based on the given global path and candidate velocity set, and relevant parameters are initialized.

[0159] In step 3.2, a certain number of virtual AUVs are generated and randomly assigned to different starting positions.

[0160] In step 3.3, each virtual AUV selects the speed at each path point in the candidate speed set according to the probability transfer rule and calculates its corresponding total energy consumption.

[0161] In step 3.4, each virtual AUV updates the pheromone concentration according to its total energy consumption and records its historical best solution.

[0162] Step 3.5, determine whether the termination condition is met, that is, whether the expected position is reached. If so, output the global optimal solution; otherwise, return to step 3.3.

[0163] That is, steps 1-3 determine the global path planning of the tunnel inspection AUV based on the starting and ending points of the tunnel inspection AUV and the information of the underwater environment.

[0164] Step 4: Design local dynamic planning for the tunnel inspection AUV, and use a method based on a two-layer planning framework to switch between global planning and local planning;

[0165] A local dynamic planning system was designed, and the tunnel inspection AUV was able to adaptively switch between global static planning and local dynamic planning. This ensured that when the tunnel inspection AUV identified obstacles through the obstacle avoidance sonar, it could avoid them through local planning and complete the task safely, efficiently and energy-efficiently.

[0166] The method based on a two-layer planning framework is used to complete the switching between the global static environment and the local dynamic environment. The method based on the two-layer planning framework is divided into a global planning layer and a local planning layer.

[0167] The global planning layer is a global planning based on a multi-attribute heuristic search algorithm from step 1 to step 3, generating a global path that meets the constraints of safety, efficiency, energy saving, etc.

[0168] The local planning layer refers to generating a local path that meets constraints such as collision avoidance based on the obstacle avoidance sonar data obtained in real time using a feedback control algorithm.

[0169] The following steps are used to adaptively switch the tunnel inspection AUV between the global static environment and the local dynamic environment:

[0170] In step 4.1, based on the given starting and ending points, as well as underwater environmental information (such as topographic maps, water flow maps, etc.), an optimization algorithm is used to generate a global path that meets the constraints of safety, efficiency, and energy saving, and the path is divided into several segments.

[0171] In step 4.2, the tunnel inspection AUV navigates according to the global path and obtains obstacle avoidance sonar data in real time.

[0172] In step 4.3, if the obstacle avoidance sonar data shows that there are urgent obstacles (such as falling rocks, debris, etc.) in the current segment, the local planning layer is triggered, otherwise the global planning layer is maintained.

[0173] In step 4.4, the local planning layer uses the dynamic window algorithm (DWA) to generate a temporary path that meets safety requirements and is as close as possible to the original segment direction and speed distribution characteristics (to reduce the degree of deviation) based on constraints such as collision avoidance, and sends it to the actuator.

[0174] In step 4.5, the tunnel inspection AUV navigates along the temporary path and continues to acquire obstacle avoidance sonar data.

[0175] In step 4.6, if the sonar data shows that there are no emergency obstacles in the current segment, the route is restored to the original segment; otherwise, the temporary path is maintained.

[0176] The dynamic window algorithm (DWA) is a local planning algorithm that can achieve real-time obstacle avoidance. It converts the position constraints of the tunnel inspection AUV into velocity constraints, samples the velocity according to the constraints, and generates trajectories from a series of actions with selected speeds. The trajectory with the highest score is selected in combination with an evaluation function to achieve the optimal planning problem.

[0177] The dynamic window consists of three parts: motion model, sampling space (dynamic window) and predicted trajectory (objective function).

[0178] The motion model is specifically:

[0179] The dynamic window method uses a mass point model, which is modeled as follows:

[0180]

[0181] In the entire sampling space, the speed and angular velocity of the unmanned system are assumed to be constant.

[0182] The sampling space is specifically:

[0183] Constraint 1: The maximum speed limit and maximum angular velocity limit range are set according to the parameters of the tunnel inspection AUV, which determines the upper and lower limits of the dynamic window sampling.

[0184] V s ={v∈[v min ,v max ],ω∈[ω min ,ω max ]} (10)

[0185] Constraint 2: Set the linear acceleration and angular acceleration constraints according to the parameters of the tunnel detection AUV.

[0186]

[0187] Constraint 3: Safe obstacle avoidance distance constraint, which ensures that the tunnel inspection AUV performs maximum acceleration braking at the current speed to avoid collision with the nearest obstacle.

[0188]

[0189] The trajectory prediction and objective function

[0190] G(v,ω)=α·zhidao(v,ω)+β·juli(v,ω)+γ·sudu(v,ω) (13)

[0191] Where zhidao is the guidance evaluation function, representing the deviation between the current heading angle and the desired heading angle. α is the guidance function weight, juli represents the distance between the tunnel detection AUV and the nearest obstacle. When the distance between the tunnel detection AUV and the obstacle exceeds a certain value, the function value is set to a constant. β is the safe obstacle avoidance distance weight. sudu is the speed evaluation function, representing the tunnel detection AUV's navigation speed. A larger value is preferred when the constraints are met. γ is the navigation speed weight.

[0192] The tunnel inspection AUV energy-saving operation planning simulation includes the energy-optimal path planning experiment of the multi-attribute heuristic search algorithm, the energy-optimal speed planning experiment of the multi-attribute heuristic search algorithm, the obstacle avoidance planning experiment based on DWA, and the tunnel inspection AUV energy consumption comparison experiment;

[0193] To achieve energy-optimized path planning, we first constructed a 3D model of the tunnel, including information such as tunnel shape, water flow, and inspection points. We then set the AUV's starting and ending points, as well as the location and number of inspection points. Since most inspection points are spaced approximately 100 meters apart and located on the tunnel walls, we randomly generated several points spaced 100 meters apart.

[0194]

[0195]

[0196] Table 1 Detection point location table

[0197] The inspection points are input into the multi-attribute heuristic search algorithm, resulting in a planned route as shown in Figure 5. Experiments show that, within the constraints of the multi-attribute heuristic search algorithm, the tunnel inspection AUV can navigate as centrally as possible and accurately reach each inspection point, while performing the inspection task in the most energy-efficient manner. Since the steps in each stage are the same, the first two inspection points, i.e., the first inspection stage, are analyzed and the speed plan is output. The two hovering points are (0, -2.850, -0.800) and (100, 2.790, 1.000).

[0198] The flaps are opened during the acceleration and constant speed phases to utilize the thrust of the water flow to achieve energy saving, and closed during the deceleration phase to reduce the impact of the water flow;

[0199] Based on the tunnel inspection AUV's operating environment and the onboard obstacle avoidance sonar, the obstacle avoidance experiment map was set to a 40m long and 7m wide rectangle. The 7m width simulated a 6.960m tunnel diameter, while the 40m length represented the sonar's detectable range. Basic tunnel inspection AUV parameters, including velocity range, acceleration range, angular velocity range, and angular acceleration range, were entered. Several typical tunnel obstacles were added to the sonar's detectable map to simulate the AUV detecting obstacles during operation. Local obstacle avoidance planning was then performed based on the DWA algorithm. The obstacle avoidance simulation is shown in Figure 6. As can be seen from the figure, when the sonar detected an obstacle, the DWA algorithm was able to perform local obstacle avoidance planning, including continuous left avoidance, continuous right avoidance, and left and right avoidance, to avoid the obstacle. Through global planning and local obstacle avoidance planning, the AUV was able to navigate safely and ultimately reach the operation point, achieving safe inspection operations.

[0200] First, to address the need for a distributed, robust, and easily integrated heuristic global optimization algorithm suitable for complex constraints and multi-objective functions, energy-saving task planning requires a distributed, robust, and easily integrated heuristic global optimization algorithm. Based on the ant colony algorithm, a multi-attribute heuristic factor was added: distance factor, energy factor, and safety factor. Secondly, to meet energy conservation requirements, the multi-attribute heuristic search algorithm achieved 5.205% energy savings, 5.636% energy savings through the wing panels, and 7.599% energy savings through the addition of speed planning. The algorithm also improved its convergence rate from 120 to 73 iterations, shortening search time and enhancing the safe navigation capability of the tunnel inspection AUV, ensuring that it navigates as centrally as possible while meeting mission requirements. Finally, the optimal path and speed were output.

[0201] After determining the optimal energy-saving path and speed through energy-saving planning, the tunnel inspection AUV can track this path by controlling its longitudinal speed and vertical depth. The team focused on balancing energy conservation and control effectiveness in both speed and depth control. An event-triggered algorithm was employed to reduce computational effort and achieve energy conservation. Sliding mode control ensured control accuracy, and an event-triggered sliding mode control was designed. The control loop was first designed to verify its stability and the absence of the Zeno phenomenon. An observer was then designed to obtain real-time data. Finally, field experiments at a fixed depth were conducted to verify the algorithm's feasibility.

[0202] Step 5: Based on the planned path and speed obtained in step 4, the tunnel inspection AUV is controlled at a fixed depth and for energy saving;

[0203] Water tunnels are often tens of kilometers long, so energy-saving control is needed to improve the endurance of AUVs and enable continuous inspections. Sliding mode control is robust to model inaccuracies and external disturbances, making it well-suited for underwater robot control.

[0204] Event triggering can reduce energy consumption and communication requirements for underwater robots. This paper introduces an event-triggered mechanism into the depth control of underwater robots. A trigger function and sliding mode control law are designed. This requires only that the AUV update the control rate when an event is triggered, achieving energy-efficient control of the AUV's depth.

[0205] The synovial membrane control of the event triggering mechanism is specifically as follows:

[0206] Assuming the expected heave speed w d , expected pitch velocity q d , expected pitch angle θ d and the expected depth z d are all constants, i.e. Definition: Heave velocity error w e =ww d , pitch angular velocity error q e =qq d , pitch angle error θ e =θ-θ d , depth error z e =zz d .

[0207]

[0208] The traditional sliding variable is defined by formula (14):

[0209] s(t)=Cx(t) (15)

[0210] The control law u(t) is designed to reach the sliding mode s(t) = 0 in a finite time while always keeping the trajectory of s(t) unchanged. Figure 7 shown.

[0211] Differentiating formula (15), we can obtain:

[0212]

[0213] The control law is designed as follows:

[0214] u(t)=-(CB) -1 {CAx(t)+Ksign[s(t)]} (17)

[0215] The control law adopts the sampling state, and subsequent updates are triggered by events. From formula (15), we can get:

[0216]

[0217] s(t)=Cx(t i ) represents t i Sliding variable at ti+1 At this moment, the control law is updated and remains constant until the next trigger is reached.

[0218] There is a measurement error e(t) in the sampling state in the system, which is given by the following formula:

[0219]

[0220] This measurement error plays an important role in achieving the desired accuracy in the stability problem. i , e(t)=x(t i )-x(t)=0 Since the control is updated only at this instant, the sliding mode will occur at this instant. However, for any t∈[t i ,t i+1 ), the deviation e(t)≠0 and s(t) will deviate from the sliding surface.

[0221] In this paper, we need to prove that the sliding mode control can achieve robust stabilization with respect to the measurement error e(t). This is ensured by executing a simple event that stabilizes the system. In this paper, we need to prove that in a finite execution time t∈[t i ,t i+1 ) does not have the Zeno phenomenon, that is, there is no accumulation of execution times between control rooms. Consider the model of formula (14) and provide proofs for the following assumptions.

[0222] Assumption 1: From formula (15) and formula (14), introduce α∈(0,∞), and then the control law formula (17) will reach the frequency band of the sliding mode near s(t) = 0, which is as follows:

[0223] {x(t)∈R n :|s(t i )|≤α||A|| -1} (20)

[0224] The system meets the following conditions at the same time:

[0225] ||CAe(t)||<α (21)

[0226]

[0227] η>0 (23)

[0228] Proof: By considering the Lyapunov function, we can prove that t∈[t i ,t i+1 ]The existence and correctness of the sliding mode at the moment.

[0229]

[0230] Substituting equation (18) into V and differentiating it with respect to time, we obtain the following equation:

[0231]

[0232] Sometimes the control law cannot be updated continuously. In this case, the control law is given by formula (5-5).

[0233]

[0234] Obviously, for t=t i At the moment Lyapunov function V<0. Then, at t∈(t i ,t i+1 ) condition, substitute formula (21) (22) and obtain:

[0235]

[0236] Therefore, the sliding mode will appear near the s(t)=0 band, as given in equation (20).

[0237] The control law is updated when:

[0238] ||C||||A||||e(t)||=σασ∈[0,1] (28)

[0239] Therefore, ||C||||A||||e(t)||≤σα and formula (21) exists permanently.

[0240] The event triggering moment can be defined as:

[0241] t i+1 = inf{t∈[t i ,+∞]:||C||||A||||e(t)||=σα} (29)

[0242] In order to avoid the Zeno phenomenon caused by event triggering, hypothesis 2 is given.

[0243] Assumption 2: Consider the model of formula (14). By executing t>t i The triggering of formula (28) at the moment makes all the increasing time series in the control rate formula (18) system form sliding mode. i+1 is the trigger moment, then the execution interval t i+1 -t i =T i >0 satisfies:

[0244]

[0245] where ρ(||x(t i )||) and β satisfy:

[0246] ρ(||x(t i )||)=||AB(CB) -1 CA||||x(t i )|| (31)

[0247] β=||B(CB) -1 K||+d max (32)

[0248] Proof 2: Set Γ={t:||C||||A||||e(t)||=0}. For t∈[t i ,t i+1 ]\t i (i=1,2,L,n) time, we can get:

[0249]

[0250] x(t)=x(t i )-e(t) is substituted into formula (33), and then the condition e(t i )=x(t i )-x(t i )=0, the solution of formula (33) can be written as:

[0251]

[0252]

[0253] Solve T according to formula (35) i , proving that hypothesis 2 is true.

[0254] To achieve faster and more effective depth control, it's also necessary to monitor the heave velocity, pitch angular velocity, and water flow rate. Therefore, an observer is needed to improve control accuracy. The water flow rate in a water tunnel is regular and fixed, requiring only prior knowledge and simple acquisition to obtain and input into the system. Therefore, a heave velocity observer was developed.

[0255]

[0256] The basic idea of designing the Luenberger observer is to artificially construct a system with the same parameters and inputs as the original system.

[0257]

[0258] The observation error at this time can be defined as:

[0259]

[0260] The error is:

[0261]

[0262] As can be seen from the above formula, without introducing feedback, the dynamics of the observation error is completely determined by the system matrix A. If the characteristic roots of A are all on the left side of the complex plane, then the observation error will approach zero over time. However, when the system itself is unstable, its zeros and poles need to be configured by introducing feedback. Therefore, the design of the Luenberger observer can be expressed as:

[0263]

[0264] At this time, the dynamic error of the observer can be expressed as:

[0265]

[0266] Then, by selecting L, the characteristic root of the above equation is an appropriate value, thereby ensuring that the error can decay to zero and remain, and is independent of the system's control input u(t) and initial state x(0). The Luenberger observer controls the observation measurement by introducing linear feedback.

[0267] The tunnel inspection AUV is used as the model for the simulation test. The expected heave speed w is selected. d =0, expected pitch speed q d =0, expected pitch angle θ d = 0 and the expected depth z d =0m. Assume that the water flow velocity u = 1m / s, the initial state w0 = 0.02m / s, q0 = 0.04rad / s, θ0 = 0.1rad, z0 = 10m.

[0268] After several adjustments, the appropriate parameters are C = diag[0.51 -0.50 -0.73], K = 0.65, α = 0.5, σ = 0.85, and the interference setting is as follows: d = [0.5sin(10t) 0.5sin(10t) 0] T The simulation results are shown in the figure below:

[0269] According to the simulation process Figure 8 The prediction curve shows that the observer works stably under interference and has good robustness. These characteristics mean that the control task of the navigation depth of the tunnel inspection AUV can be completed. The observer coincides with the actual speed in a short time (within 5s) and can feed back the data to the controller in time. The event-triggered algorithm update is as follows Figure 9As shown in the figure, 1 represents updating the algorithm and 0 represents not updating the algorithm. During the adjustment phase (steps 0–300), the algorithm is constantly updating due to large system errors. However, during the stable phase (steps 300–513), unnecessary calculations are stopped, effectively reducing the energy loss caused by changes in the control law.

[0270] Taking the tunnel inspection AUV as the research object, a variety of control methods were simulated and compared to verify the depth control effect of the sliding mode controller based on event triggering. Figure 10 shown.

[0271] In a 20-second simulation, there are 513 steps, of which 26 require no computation, accounting for approximately 5.07% of the total number of steps. This results in a 5.07% energy saving. This saves 5.07% of computing resources, as shown in Table 2.

[0272] Table 2 Algorithm update count table

[0273]

[0274] During the simulation, Figure 10 It can be seen that the PID controller can achieve depth control of the AUV, but there is overshoot. It stabilizes in about 20 seconds, but the steady-state error is about 0.2m.

[0275] During the simulation, we can see that the event-triggered controller can achieve AUV depth control, but it stabilizes in about 41 seconds. Although it can reduce the amount of calculation, it takes a long time to stabilize.

[0276] As can be seen from the comparison chart, the event-triggered sliding mode control can reach the desired depth faster, more accurately, and more stably, and can achieve 5.07% energy saving in the calculation steps.

[0277] Then, the control simulation of the horizontal plane was carried out under the same experimental conditions. The position (x, y), velocity (u, v) and the sliding mode control trajectory based on event triggering were as follows: Figure 12 shown.

[0278] Depend on Figure 12 (b) It can be seen that the detection point is reached accurately, and the initial and final velocities are both 0, so that the tunnel detection AUV can detect important positions. Figure 12 (d) It can be seen that the event-triggered sliding mode control can track the energy-saving speed faster and more accurately.

[0279] Table 3 Event trigger comparison table

[0280]

[0281] Depend on Figure 12(c) It can be seen that in the first stage (0-20s), the sliding mode control based on event triggering responds faster than PID control and event triggering control, with the smallest overshoot and error, and tracks the target speed the earliest; in the second stage (20-104s), the sliding mode control based on event triggering has no oscillation and steady-state error compared with PID control and event triggering control, and keeps tracking the target speed; in the third stage (104-140s), the sliding mode control based on event triggering responds faster than PID control and event triggering control, with the smallest overshoot and error. Small, the target speed is tracked earliest; it can be seen from Table 3 that both event-triggered sliding mode control and event-triggered control have the effect of saving computational complexity, which are 5.38% and 5.24% respectively. The energy-saving effects are similar, but the sliding mode control based on event triggering is much better than the event triggering control. Event triggering can achieve energy saving, but the control effect is not good after reducing the computational complexity. The excellent control method combined with sliding mode control can effectively make up for its shortcomings and achieve precise control with both energy saving and good control effect, helping tunnel inspection AUV to improve endurance and more accurately inspect longer water tunnels.

[0282] Through experiments, we can see that the sliding mode control based on event triggering can track the energy saving speed faster and more accurately. Figure 13 (b) It can be seen that the detection point is reached accurately, and the initial and final velocities are both 0, so that the tunnel inspection AUV can detect important positions.

[0283] Depend on Figure 13 (a) It can be seen that in the first stage (0-10s), the event-triggered sliding mode control responds faster than PID control and event-triggered control, with the smallest overshoot and the earliest tracking of the target speed; in the second stage (10-18s), the relatively stable stage, the event-triggered sliding mode control has no oscillation and no steady-state error compared with PID control and event-triggered control, and always tracks the target speed; in the third stage (18-28s), the sudden change stage, the event-triggered sliding mode control responds faster than PID control and event-triggered control, with the smallest overshoot and the earliest tracking of the target speed; both event-triggered sliding mode control and event-triggered control have the effect of saving computational complexity, and the energy-saving effects are similar. As can be seen from Table 3, the effect of event-triggered sliding mode control is much better than that of event-triggered control. Event triggering can achieve energy saving, but the control effect is not good after reducing the computational complexity. The excellent control method combined with sliding mode control can effectively make up for its shortcomings and achieve precise control with both energy saving and good control effect, helping tunnel inspection AUV to improve endurance and more accurately inspect longer water transfer tunnels.

[0284] An electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.

[0285] A computer-readable storage medium is used to store computer instructions, which implement the steps of the above method when executed by a processor.

[0286] The above is a detailed introduction to the energy-saving path planning method for water tunnel inspection AUV proposed in the present invention, and the principles and implementation methods of the present invention are explained. The description of the above embodiments is only used to help understand the method and core ideas of the present invention; at the same time, for those skilled in the art, according to the ideas of the present invention, there will be changes in the specific implementation methods and application scopes. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. An energy-saving path planning method for a water conveyance tunnel inspection AUV, characterized in that: The method specifically includes the following steps: Step 1, perform optimization planning on the tunnel inspection AUV in the already constructed environmental model; Step 2, based on the optimization planning in Step 1, add a multi-attribute heuristic factor in the ant colony algorithm to plan the AUV path; Step 3, based on the path planned in Step 2, plan the speed of the AUV to obtain the global path planning of the tunnel inspection AUV; The following steps are used for energy-saving speed planning: Step 3.1, construct a discrete search space according to the given global path and candidate speed set, and initialize relevant parameters; Step 3.2, generate a certain number of virtual AUVs and randomly assign them to different starting positions; Step 3.3, each virtual AUV selects the speed at each path point in the candidate speed set according to the probability transfer rule and calculates its corresponding total energy consumption; Step 3.4, each virtual AUV updates the pheromone concentration according to its total energy consumption and records its historical best solution; Step 3.5, judge whether the termination condition is satisfied, that is, whether the expected position is reached. If it is satisfied, output the global best solution; otherwise, return to Step 3.3; Step 4, design local dynamic planning for the tunnel inspection AUV, and use a method based on a two-layer planning framework to switch between global planning and local planning; Step 5, based on the planned path and speed obtained in Step 4, perform depth-fixed energy-saving control on the tunnel inspection AUV, taking into account both energy saving and control effects in speed control and depth control, and complete the energy-saving path planning of the tunnel inspection AUV.

2. The energy-saving path planning method according to claim 1, characterized in that: In Step 1, Suppose there are M AUVs. The current position of the m-th AUV is i, and the next optional target position is j; each AUV needs a state transition probability to select the position it will visit next; the state transition probability is a function of the current position i and the candidate position j; specifically, at the t-th iteration, the probability that the m-th AUV transfers from position i to position j is given by the following formula: Among them, τ ij represents the pheromone concentration between positions i and j at the tth iteration; η ij represents the heuristic function value; α is the pheromone concentration. A larger α means that more AUVs have passed through the path, thus releasing more pheromones, indicating that the path is better. β is the importance of the heuristic function value relative to the state transition probability. A larger β means a higher heuristic function value, indicating that the path is more in line with the preset goal or constraint conditions. Introduce a random parameter q to increase the exploration of the algorithm; q is a random number uniformly distributed in the interval [0, 1], and q0 is a constant (0 < q0 < 1). When q < q0, the m-th AUV selects the next target position according to the principle of the maximum state transition probability, that is, when q > q0, the roulette method is used to select the next target position, and the formula is as follows: Among them, allowed m ={C-tabu m } represents the set of all next positions j that can be reached at the current position i; [η ij (t)]=1 / d jg , d jg Indicates the distance from the next position j to the final target point g. Planning is based on the distance. The shorter the distance, the more likely this point is to be chosen. After an individual completes a complete optimization, it is necessary to restrict the influence of pheromone and heuristic function on the next loop, and it is necessary to restrict the pheromone. This process corresponds to the evaporation of pheromone in nature. After each algorithm iteration, the restriction formula of pheromone is as follows: Where ρ is the pheromone constraint, Q is the starting parameter of pheromone, L m The length of the path obtained by the mth AUV after completing the search, is the coordinate set of the path that the m-th AUV passes through in the t+1 optimization.

3. The energy-saving path planning method according to claim 2, characterized in that: In Step 2, The multi-attribute heuristic factor includes a distance factor, an energy-saving factor, and a safety factor; The basic ant colony algorithm was improved and applied to the energy-saving operation planning of tunnel inspection AUVs in tunnel environments. The traditional distance factor heuristic function was improved to enable it to converge to the optimal path faster and reduce the computational complexity of invalid searches at the initial moment. The energy factor and tunnel safety factor were added to the heuristic function so that the searched detection path can meet the requirements of tunnel inspection. The multi-attribute heuristic function was constructed as follows: or ij =the l ij ·or e ij ·or c ij (5) Among them, η ij is the distance factor, η e ij is the energy consumption factor, η c ij Safety factor The path selection in the initial stage mainly depends on the heuristic function. Since no AUV has passed by, no path has pheromones. The guidance of setting the target direction for the initial search is as follows: Where: L(i) is the distance traveled by the AUV from the current position i; D jd is the length from the next target point j to the final target point g; d jz The sum of the distances between the next target point j and the line connecting the starting point and the final target point; d ij is the length from the current position i to the next target point j; Substitute the energy consumption model into the energy saving factor: When the tunnel inspection AUV does not delay the execution of the task, it is safest to navigate in the center; establish a model from the center point of the tunnel cross section, set the set of center points, that is, the axis of the tunnel as a function of the center position of the safe area, and design a model with decreasing influence outward. as follows: Among them, l r is the tunnel radius, l ja Indicates the distance from the center of navigation at position j.

4. The energy-saving path planning method according to claim 3, characterized in that: In step 3, speed planning refers to determining the optimal speed of the tunnel inspection AUV at each path point based on the tunnel inspection AUV's dynamic model, environmental parameters, and mission requirements; The energy-saving speed planning problem is transformed into a discrete optimization problem, that is, selecting the optimal speed at each path point from a given set of candidate speeds, and using an optimization algorithm to search for the best combination.

5. The energy-saving path planning method according to claim 4, characterized in that: In step 4, A method based on a two-layer planning framework is used to switch between a global static environment and a local dynamic environment; the method based on the two-layer planning framework is divided into a global planning layer and a local planning layer; The global planning layer is the global path generated according to steps 1 to 3; The local planning layer refers to the local path generated by the feedback control algorithm based on the obstacle avoidance sonar data obtained in real time; The following steps are used to adaptively switch the tunnel inspection AUV between the global static environment and the local dynamic environment: Step 4.1: Based on the given starting and ending points, as well as underwater environment information, an optimization algorithm is used to generate a global path that meets the constraints and divide it into several segments. Step 4.2: The tunnel inspection AUV navigates according to the global path and obtains obstacle avoidance sonar data in real time; Step 4.3: If the obstacle avoidance sonar data indicates that there is an urgent obstacle in the current segment, the local planning layer is triggered, otherwise the global planning layer is maintained; In step 4.4, the local planning layer uses the dynamic window method to generate a temporary path that meets safety requirements and is as close as possible to the original segment direction and speed distribution characteristics based on constraints such as collision avoidance, and sends it to the actuator; Step 4.5: The tunnel inspection AUV navigates along the temporary path and continues to acquire obstacle avoidance sonar data. In step 4.6, if the sonar data shows that there are no emergency obstacles in the current segment, the route is restored to the original segment; otherwise, the temporary path is maintained.

6. The energy-saving path planning method according to claim 5, characterized in that: The dynamic window consists of three parts: motion model, sampling space and predicted trajectory; The motion model is specifically: The dynamic window method uses a mass point model, which is modeled as follows: In the entire sampling space, the speed and angular velocity of the unmanned system are assumed to be constant; The sampling space is specifically: Constraint 1: The maximum speed limit and maximum angular velocity limit are set according to the parameters of the tunnel inspection AUV, which determines the upper and lower limits of the dynamic window sampling; V s ={v∈[v min ,v max ],ω∈[ω min ,ω max ]} (10) Constraint 2: Set the linear acceleration and angular acceleration constraints based on the parameters of the tunnel inspection AUV; Constraint 3: Safe obstacle avoidance distance constraint, which ensures that the tunnel inspection AUV performs maximum acceleration braking at the current speed to avoid collision with the nearest obstacle; The trajectory prediction G(v,ω)=α·zhidao(v,ω)+β·juli(v,ω)+γ·sudu(v,ω) (13) Among them, zhidao is the guidance evaluation function, which represents the deviation between the current heading angle and the desired heading angle; α is the guidance function weight; juli represents the distance between the tunnel detection AUV and the nearest obstacle. When the distance between the tunnel detection AUV and the obstacle exceeds a certain value, the function value is set to a constant; β is the safe obstacle avoidance distance weight; sudu is the speed evaluation function, which represents the navigation speed of the tunnel detection AUV. When the constraints are met, the larger the value, the better; γ is the navigation speed weight.

7. The energy-saving path planning method according to claim 6, characterized in that: In step 5, An event-triggered mechanism is introduced into the depth control of the underwater vehicle. A trigger function and sliding mode control law are designed. The AUV is only required to update the control rate when an event is triggered, thus achieving energy-saving control of the AUV's depth. The synovial membrane control of the event triggering mechanism is specifically as follows: Assuming the expected heave speed w d , expected pitch velocity q d , expected pitch angle θ d and the expected depth z d are all constants, i.e. Definition: Heave velocity error w e =ww d , pitch angular velocity error q e =qq d , pitch angle error θ e =θ-θ d , depth error z e =zz d ; The traditional sliding variable is defined by formula (14): s(t)=Cx(t) (15) The control law u(t) is designed to reach the sliding mode s(t) = 0 in a finite time while always keeping the trajectory of s(t) unchanged; Differentiating formula (15), we can obtain: The control law is designed as follows: u(t)=-(CB) -1 {CAx(t)+Ksign[s(t)]} (17) The control law adopts the sampling state, and subsequent updates are triggered by events; from formula (15), we can get: s(t)=Cx(t i ) represents t i Sliding variable at t i+1 At this moment, the control law is updated and remains constant until the next trigger is reached.

8. An energy-saving path planning system for AUVs used in water tunnel inspection, characterized by: The system is used to perform the steps of the method according to any one of claims 1 to 7: The system includes a global planning module, a local planning module and a depth control module: The global planning module first optimizes the tunnel inspection AUV in the constructed environment model; then adds a multi-attribute heuristic factor to the ant colony algorithm to plan the AUV path; and then plans the AUV's speed to obtain the global path planning of the tunnel inspection AUV. The local planning module designs local dynamic planning for the tunnel inspection AUV and switches between global planning and local planning using a method based on a two-layer planning framework; The depth control module obtains the planned path and speed based on the global planning module and the local planning module, performs fixed-depth energy-saving control on the tunnel inspection AUV, takes into account both energy saving and control effects in speed control and depth control, and completes the energy-saving path planning of the tunnel inspection AUV.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium for storing computer instructions, characterized in that: When the computer instructions are executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images