AUV (Autonomous Underwater Vehicle) comprehensive path planning method for communication efficiency and obstacle avoidance
By combining improved ant colony algorithms and 3D DWA algorithms, the problems of underwater communication efficiency and obstacle avoidance of AUVs were solved, enabling AUVs to navigate efficiently and safely in complex marine environments.
Patent Information
- Application Number
- CN202411046448.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-01
- Publication Date
- 2024-11-08
AI Technical Summary
AUVs face challenges such as poor communication environment, low acoustic communication rate, and dynamic obstacle avoidance when operating underwater, which are difficult to solve effectively with existing technologies.
An improved ant colony algorithm combined with a three-dimensional dynamic window method is adopted. By introducing Pareto sorting and non-dominated sorting into global path planning and combining signal-to-noise ratio calculation, the path length and communication loss are optimized. The three-dimensional DWA algorithm is used to avoid dynamic obstacles in real time.
It enables safe navigation and efficient communication of AUVs in complex marine environments, significantly improves the efficiency and safety of path planning, and optimizes path length and communication efficiency.
Smart Images

Figure CN118915762A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of AUV path planning, and in particular relates to an AUV comprehensive path planning method oriented to communication efficiency and obstacle avoidance. Background Art
[0002] Autonomous Underwater Vehicle (AUV) is an autonomous underwater vehicle that has a wide range of applications in the fields of ocean exploration, seabed exploration and resource development. AUV has autonomous navigation, obstacle avoidance, data collection and communication capabilities, and can complete various tasks under human control. They usually use advanced sensors and control systems to autonomously perform pre-set tasks in underwater environments, such as topographic mapping, seabed exploration and marine life monitoring.
[0003] In deep-sea exploration, exploration of seabed resources and archaeological excavation, AUV can significantly improve the success rate of mission completion, reduce costs and safety risks. However, AUV faces difficulties such as poor communication environment, low acoustic communication rate and dynamic obstacle avoidance when operating underwater. To solve these problems, a multi-objective path planning method combining an improved ant colony optimization (ACO) and a three-dimensional dynamic window approach (DWA) was proposed. Summary of the invention
[0004] The purpose of the embodiments of the present invention is to provide an AUV comprehensive path planning method for communication efficiency and obstacle avoidance, aiming to solve the problems raised in the above background technology.
[0005] The embodiment of the present invention is implemented as follows: an AUV comprehensive path planning method for communication efficiency and obstacle avoidance includes the following steps:
[0006] Step 1: Construct AUV communication application scenario and underwater acoustic communication environment model:
[0007] Step 2: Improvement and fusion of 3D ACO and DWA algorithms:
[0008] The AUV navigation part will use the improved ant colony algorithm as the global path planning algorithm to avoid static obstacles, and the three-dimensional DWA algorithm will be used as the local path planning algorithm to avoid dynamic obstacles; in the global path planning, the Pareto sorting process will be added to the ant colony algorithm, and the traditional ant colony algorithm will be modified from strengthening the optimal path to strengthening a group of non-dominated paths to achieve the purpose of multi-objective optimization, namely the shortest path and the minimum communication loss.
[0009] An embodiment of the present invention provides an AUV comprehensive path planning method for communication efficiency and obstacle avoidance, which aims to optimize the global path planning of the AUV, complete obstacle avoidance, and minimize the path length and communication propagation loss. By introducing non-dominated sorting, improved heuristic functions and pheromone update rules, the improved three-dimensional ant colony algorithm can effectively solve the multi-objective optimization problem and achieve a balance between path length and propagation loss. In addition, combined with the improved three-dimensional dynamic window method, according to the global route given by the three-dimensional ant colony algorithm, obstacles are avoided in real time in a dynamic ocean environment to ensure the safe navigation and low communication loss of the AUV. By assigning different weights to the pitch angular velocity sampling area in the dynamic window method. This method enables the AUV to respond to dynamic obstacles in the environment more flexibly and further improve the efficiency and safety of task execution. Experimental results show that compared with traditional methods, this method shows significant advantages in both path length and communication efficiency, showing its application potential in complex marine environments. BRIEF DESCRIPTION OF THE DRAWINGS
[0010] Figure 1 This is a diagram of underwater communication application scenarios;
[0011] Figure 2 is the seawater absorption coefficient in the operating frequency band of the underwater acoustic application system;
[0012] Figure 3 Schematic diagram of the ant's visual search space (where a: forward movement distance is L x,max , the lateral and vertical movement distances are L y,max , L z,max ; b: Forward, lateral and vertical movement distances are equal L x,max =L y,max =L z,max =1; c: forward, lateral and vertical movement distances are equal L x,max =L y,max =L z,max =2);
[0013] Figure 4 Update pseudocode for global pheromone;
[0014] Figure 5 It is the velocity vector space of the three-dimensional DWA algorithm;
[0015] Figure 6 It is a schematic diagram of the velocity vector space with different weights of the three-dimensional DWA algorithm;
[0016] Figure 7 Combine global and local path planning for collision avoidance;
[0017] Figure 8 Provide underwater simulation environment for AUV;
[0018] Fig. 9 The paths searched by the traditional ant colony algorithm at different viewing angles, with the communication node at (0,0,0) (a: observation angle Az 45°Ei 45°; b: observation angle Az 135°Ei 45°; c: observation angle Az 0°Ei 90°);
[0019] Fig.10 The paths searched by the traditional ant colony algorithm at different viewing angles, with the communication node at (200, 200, 200) (where a: observation angle Az 45°Ei 45°; b: observation angle Az-45°Ei 45°; c: observation angle Az 0°Ei 90°);
[0020] Fig.11 To improve the path searched by the ant colony algorithm under different viewing angles, its communication node is at (0,0,0) (where a: observation angle Az 45°Ei 45°; b: observation angle Az 135°Ei 45°; c: observation angle Az 0°Ei 90°);
[0021] Fig.12 To improve the path searched by the ant colony algorithm under different viewing angles, its communication node is at (200, 200, 200) (where a: observation angle Az 45°Ei 45°; b: observation angle Az-45°Ei 45°; c: observation angle Az 0°Ei 90°);
[0022] Fig.13 It is the non-dominated solution searched by the traditional ant colony algorithm and the improved ant colony algorithm, and its communication node is at (0,0,0);
[0023] Fig.14 It is the non-dominated solution searched by the traditional ant colony algorithm and the improved ant colony algorithm, and its communication node is at (200,200,200);
[0024] Fig.15 Schematic diagram of the DWA algorithm without weights avoiding obstacles along the reference path;
[0025] Fig.16 Schematic diagram of the weighted DWA algorithm avoiding obstacles along the reference path. DETAILED DESCRIPTION
[0026] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.
[0027] The specific implementation of the present invention is described in detail below in conjunction with specific embodiments.
[0028] An embodiment of the present invention provides an AUV comprehensive path planning method for communication efficiency and obstacle avoidance, comprising the following steps:
[0029] Step 1: Construct AUV communication application scenario and underwater acoustic communication environment model:
[0030] Step 1.1: Build AUV communication application scenario:
[0031] The main participants of AUV underwater communication include AUV, surface mother ship, underwater network nodes, buoys, unmanned surface vessels (USV) and unmanned aerial vehicles (UAV). The communication content includes control instructions, location information and information obtained by AUV, such as seabed topography and biological information. Figure 1 shown.
[0032] At present, subject to the influence of transmission distance, transmission rate and specific transmission technology, the main transmission content of underwater communication includes underwater data collection and surface data transmission. In terms of underwater data collection: the communication content between AUV and AUV is mainly to confirm the distance and position between each other in order to avoid or work together. The communication between AUV and the mother ship is transmitted by the mother ship with low data rate control instructions, and then the AUV uses high data rate to transmit the images, data and other information obtained underwater with the assistance of USV. AUV can complete the information exchange with the land information center through surface floats or underwater communication nodes. The underwater communication nodes and floats are laid in advance by the ground staff. The placement of floats in unfamiliar waters is mainly completed by drones. In terms of surface data transmission: the land information center uses radio communication to send information to the mother ship, drone or float, and then it uses sonar communication or hydroacoustic communication to send it to AUV. Since radio communication is used on the surface, the information rate transmitted between the land information center and the mother ship, drone and float is higher, and the transmission loss is smaller.
[0033] For AUV, due to the slow underwater data transmission rate and large propagation loss, the data needs to be transmitted to the surface USV or float first, and then forwarded by it. The main research scenario here is that the AUV maintains a small data transmission loss with the float or USV during the task (operating from the starting point to the target point).
[0034] Step 1.2: Construct the AUV underwater acoustic communication environment model:
[0035] In underwater environments, since electromagnetic waves (especially high-frequency electromagnetic waves, such as radio waves and microwaves) propagate very poorly in water, sound waves are mainly used for communication. During the propagation of sound waves, sound energy is converted into other forms of energy and absorbed by the medium. Its propagation loss TL in the medium is mainly composed of expansion loss and attenuation loss. The propagation loss in seawater is a function of distance l and signal frequency f, which can be approximately expressed as:
[0036] A(l,f)=l k [α abs (f)] l (1)
[0037] Where k is the expansion factor, α abs (f) is the seawater absorption loss coefficient. The above formula is expressed in decibel (dB), and the sound transmission loss TL is:
[0038] TL=10k·log(1000l)+10l·logα abs (f) (2)
[0039] Where k is the distance of sound wave propagation, in km. The first term represents the expansion loss, and the second term represents the absorption loss. The expansion factor k is related to the propagation mode and propagation path of the sound wave. Depending on the propagation conditions, k takes different values, and here it is 1.5. Seawater absorption coefficient α abs (f) can be expressed in dB form using the Thorp formula:
[0040]
[0041] Seawater absorption coefficient α abs (f) The relationship with the operating frequency is as follows Figure 2 shown.
[0042] In addition to frequency, the seawater absorption coefficient is also affected by water temperature, salinity, pressure, etc. Generally, as the temperature, salinity and pressure increase, the seawater absorption coefficient will also increase. Here, the operating frequency selected is 20kHz, the system bandwidth is 100kHz, and the seawater temperature is 10℃. The seawater absorption coefficient α is obtained. abs (f) is 3.30.
[0043] Noise in the ocean is one of the main sources of interference for reliable transmission of underwater acoustic signals. In the frequency range of 200Hz to 50kHz, the main source of environmental noise is the wind level on the ocean surface. When the wind level doubles, the environmental noise increases by about 5dB. The peak of environmental noise is about 500Hz, and then it decreases by about 6dB per frequency doubling. For example, at 10kHz, the spectral density of environmental noise is between 28 and 50dB / Hz. Ocean environmental noise can be described by the Wenz model of the cylinder, that is:
[0044]
[0045] Where N t 、N sh 、N w and N th They represent turbulent noise, shipping noise, sea surface noise and thermal noise respectively. f is the frequency in kHz; sh is the shipping density in the water area, with a value of (0-1); w is the wind speed on the ocean surface in m / s. Turbulent noise N t Main range: f<10Hz; shipping noise N sh The main impact is 10~100Hz; the sea surface noise caused by wind mainly affects the frequency range of 100Hz~100kHz; thermal noise mainly affects f>100kHz. The total noise spectrum density synthesized by the above types of environmental noise can be expressed as:
[0046] N(f)=N t (f)+N sh (f)+N w (f)+N th (f) (5)
[0047] In a specific frequency band, the noise spectral density decreases linearly with the logarithmic scale and can be approximately expressed as:
[0048] 10logN(f)≈N1-ηlogf(6)
[0049] N1 is the logarithmic value of the noise level at a certain reference frequency and can be regarded as a reference value of the noise level. η describes the slope or attenuation rate of the noise level as it changes with frequency. When the transmitted signal frequency f and power P are given, the propagation loss A(l,f) obtained by equation (1) and the noise spectral density level N(f) obtained by equation (6) can be used to obtain the signal-to-noise ratio at distance l:
[0050]
[0051] Where S(l,f) is the signal power spectrum density at distance l and frequency f, and the system bandwidth is determined as:
[0052] B(l)=[f min (l),f max (l)] (8)
[0053] Where B(l) represents the working bandwidth of the system at distance l, f min (l) and f max (l) represent the lowest and highest operating frequencies at distance l respectively, and the received signal-to-noise ratio is shown in formula (9):
[0054]
[0055] By integrating the signal-to-noise ratio at each distance point, we can obtain the average signal-to-noise ratio on the entire path:
[0056]
[0057] L1 and L2 define the range of distance, and the average signal-to-noise ratio from L1 to L2 is obtained. By comparing this average value, the path with better communication quality can be evaluated and selected.
[0058] Step 2: Improvement and fusion of 3D ACO and DWA algorithms:
[0059] The AUV navigation part will use the improved ant colony algorithm as the global path planning algorithm to avoid static obstacles, and the three-dimensional DWA algorithm will be used as the local path planning algorithm to avoid dynamic obstacles. In the global path planning, the Pareto sorting process is added to the ant colony algorithm, and the traditional ant colony algorithm is modified from strengthening the optimal path to strengthening a group of non-dominated paths, so as to achieve the purpose of multi-objective optimization, namely the shortest path and the minimum communication loss. Specifically:
[0060] Step 2.1: Improve the 3D ACO mathematical model:
[0061] Since the application background is an underwater three-dimensional environment, and the problem to be solved is a multi-objective optimization problem (taking into account the shortest path and communication efficiency), the increase in the number of grids will bring a great degree of computational burden. Therefore, the visual search space method is used to reduce the difficulty of calculation, so that ants can only access some nearby nodes, thereby reducing the amount of calculation. The signal-to-noise ratio calculation part is added to the heuristic function, so that ants can consider the impact of path length and communication efficiency when searching for paths. The non-dominated sorting part is introduced when updating the global pheromone. The pheromone on the shortest path of the traditional ant colony algorithm is modified to strengthen the pheromone on a group of non-dominated solutions on the Pareto frontier to solve the multi-objective optimization problem. Specifically include:
[0062] Visual Search Space:
[0063] In three-dimensional space, as the number of grids increases, a large degree of computational burden will be caused. In order to reduce the complexity of the search, the movement of the AUV is assumed to be forward movement, lateral movement and vertical movement. The maximum forward movement distance of the ant is L x,max , the maximum lateral movement distance is L y,max , the maximum vertical movement distance is L z,max When an ant moves from the current node to the next node, the searchable area is limited to the ant's visible search area, which simplifies the search space and improves the search efficiency of the ant colony algorithm. Figure 3 As shown in a, Figure 3 b and 3c are the movement distance L x,y,z,max When 1 and 2 are respectively, the visible search space is 26 and 124 grids.
[0064] State transition probability:
[0065] In the process of searching the path by the ant colony algorithm, the state transition probability is calculated mainly based on the pheromone and the heuristic information of each node. At time t, when the mth ant (m=1,2,...,M) moves from node i to node j, the state transition probability of each optional node around node i is first calculated. As shown in formula (11), node j is then determined by the roulette method.
[0066]
[0067] Among them, τ ij (t) is the pheromone concentration from the current node i to the node j at time t; H ij (t) is the path heuristic information from the current node i to node j at time t; allowed m is the set of optional subsequent grids around the current i-th grid for the k-th ant; μ and ρ are important factors of pheromone and heuristic information, respectively, where H ij (t) is calculated as follows:
[0068] H ij (t)=Safe(x,y,z)×(Distance(x,y,z)+SNR(x,y,z)) (12)
[0069] Where Safe(x,y,z) indicates whether the point is safe and feasible. If the point is an obstacle, the value is 0, otherwise it is 1, as shown in formula (13):
[0070]
[0071] Distance(x,y,z) is the distance between point (x,y,z) and target point (x d ,y d ,z d ), the Euclidean distance between them is calculated as follows:
[0072]
[0073] Among them, the molecule K d is a constant, indicating the importance of the path length. SNR(x,y,z) indicates the signal-to-noise ratio of the point (x,y,z). It can be obtained from formula (9). The heuristic function design is shown in formula (15):
[0074]
[0075] Where S0 is the signal power spectrum density, which represents the signal power level within a specific frequency or frequency band. l (l,f) is the propagation loss, which can be calculated by formula (1-2). This value is related to the distance between the current point and the target point. B is the bandwidth, which is used to describe the maximum data rate that can be transmitted by the signal or data transmission channel. N(f) is the noise power spectrum density at a specific frequency f, which can be obtained by formula (16) and calculated as follows:
[0076]
[0077] Pheromone update rules:
[0078] Pheromone update includes local update and global update. Local update means that when an ant passes through any node, the pheromone at that point will decrease, increasing the probability of the ant exploring a new path and achieving the purpose of global search, as shown in formula (17):
[0079] τ ij (t) = (1-ζ)·τ ij (t-1)+ζ·τ0 (17)
[0080] Where ζ is the attenuation coefficient of the pheromone.
[0081] Global update means that after all ants in each generation have finished searching, all paths are evaluated and the best path is selected to strengthen the pheromone on the path. The Pareto sorting process is introduced here to improve the global pheromone update mechanism in the ant colony algorithm. Specifically, in each iteration of the algorithm, not only a single optimal solution is focused on, but a set of non-dominated solutions, namely the Pareto frontier, is identified through Pareto sorting. This process starts from the first generation, and the non-dominated solution set is selected by Pareto sorting the multi-objective fitness values of the paths found by the ants, and the global Pareto solution set and its corresponding path set are initialized accordingly.
[0082] Subsequently, in each iteration of the algorithm, the newly discovered paths are integrated with the existing global Pareto solution set and Pareto sorted again. This step ensures the continuous updating and optimization of the global Pareto solution set. At the same time, in order to guide the ants to explore better solutions, the pheromones on each path in the global Pareto solution set are strengthened. It is worth noting that the degree of strengthening of pheromones on each path is closely related to the solution congestion of the path. Solutions with lower congestion will receive more pheromone enhancements to encourage the algorithm to search for more diverse excellent solutions.
[0083] The entire search process aims to find an excellent and diverse solution set by continuously updating and strengthening the global Pareto solution set to better approximate the real Pareto frontier. In this way, the algorithm can increase the diversity of solutions while ensuring the quality of the solution, thereby effectively solving multi-objective optimization problems. The pseudo code for global pheromone update is as follows: Figure 4 shown.
[0084] Pheromone Update τ ij (t+1) and the pheromone increment Δτ ij (t) As shown in formula (18-19):
[0085] τ ij (t+1)=(1-ρ)·τ ij (t)+ρ·Δτ ij (t) (18)
[0086]
[0087] Among them, C(sol) is the crowding degree of the solution sol in the global Pareto frontier GlobalParetoSet. Under the premise of strengthening the pheromone on the path corresponding to the non-dominated solution, the smaller the crowding degree, the larger the pheromone increment Δτ, and the larger the crowding degree, the smaller the pheromone increment Δτ, thereby ensuring the diversity of solutions under the premise of excellent solutions; ρ is the pheromone update coefficient; Q is a constant.
[0088] Fitness function design:
[0089] The two optimization goals to be solved are path length and communication efficiency. The communication efficiency is characterized by the signal-to-noise ratio. After all ants in each generation have finished searching, the paths of all ants are evaluated using the fitness function. The fitness function is designed as shown in formula (20-21):
[0090] fitness 1,m =judge·k1·L m (20)
[0091]
[0092] Among them, fitness 1,m is the first fitness function value of the mth ant searching for the path, fitness 2,m is the second fitness function value of the mth ant searching for the path, L m is the length of the path searched by the mth ant, SNR avg,mis the average signal-to-noise ratio of the path searched by the mth ant, k1 and k2 are constants. (fitness1, m, fitness2, m) is used to represent the solution of the multi-objective optimization problem. The smaller the fitness function value, the better the solution and the more ideal the path.
[0093] Step 2.2: Construct a 3D DWA algorithm:
[0094] There are many uncertainties in the underwater environment. In addition to static obstacles, AUVs are also threatened by dynamic obstacles such as fish or other AUVs. The DWA algorithm is used as a local path planning algorithm, but the traditional DWA algorithm is mostly used in two-dimensional planes. Therefore, it is modified to a three-dimensional DWA algorithm, and different priorities are given to the velocity sampling space, so that it can maintain good communication efficiency while avoiding obstacles.
[0095] The DWA algorithm samples multiple groups of velocities in the velocity space (ν, ω), simulates the motion trajectory of the velocity within the prediction period, and evaluates the trajectory through the evaluation function, and finally selects the velocity group (ν, ω) corresponding to the optimal trajectory to drive the robot motion. Specifically:
[0096] AUV kinematic model:
[0097] Assume that the underwater AUV motion model is a non-omnidirectional mobile robot. The AUV realizes forward and backward motion through a fixed thruster, and adjusts the heading through a limited angle steering jet system. The state vector x and control input u of the AUV are as follows:
[0098] x=[x,y,z,ψ,θ,ν,r,q] T (twenty two)
[0099] u=[v,r,q] (23)
[0100] Among them, x, y, z are the position coordinates of the robot in three-dimensional space, ψ, θ are the yaw angle and pitch angle, which respectively represent the inclination of the AUV in the horizontal plane and the vertical direction, v is the forward speed of the AUV, r, q are the angular velocities of the yaw and pitch rotation around the z-axis and y-axis respectively. The relationship between the position, orientation and speed of the robot at the next moment t+Δt and the current moment t is as follows:
[0101]
[0102] AUV velocity sampling space:
[0103] Due to the limitations of AUV hardware and structure, the velocity sampling space V s (ν, r, q) is limited to a certain range. According to the hardware conditions of the AUV, there is a speed boundary limit. At this time, the sampled speed space Vm for:
[0104] V m ={(ν,r,q)|v∈[v min ,v max ],r∈[r min ,r max ],q∈[q min ,q max ]} (25)
[0105] Due to the influence of motor performance, the acceleration has a limited range. There is a dynamic window in the forward simulation cycle of the AUV moving trajectory. The speed within this window is the achievable speed of the AUV:
[0106]
[0107] where v c ,r c ,q c is the current speed of the AUV, is the maximum deceleration of the AUV, is the maximum acceleration of the AUV. For safety reasons, the AUV can decelerate in time when encountering obstacles, and there is a maximum deceleration limit:
[0108]
[0109] By comprehensively considering the above three constraints, the range of the velocity space (ν, r, q) can be determined. As the acceleration of the motor changes, the velocity space will change dynamically. By sampling the velocity group (ν, r, q) that meets the constraints, the corresponding trajectory can be obtained. The three-dimensional dynamic window is shown as follows: Figure 5 shown.
[0110] Figure 5 The x-axis is the linear velocity of the AUV along the forward direction, ranging from [0,v max ], the y-axis is the yaw angular velocity, the range is [r min ,r max ], the z axis is the pitch angular velocity, the range is [q min ,q max ], the black area is the obstacle area in the velocity vector space, and the gray area is the velocity space V that the AUV can sample m , the yellow area is the dynamic window area V affected by the motor performance d , the yellow area is removed from the part that overlaps with the black area, and the remaining part is the actual feasible speed area V after considering the above three constraints. m ∩V d ∩V aAfter obtaining the speed space, different speed combinations (ν, r, q) can be obtained and scored to select the optimal speed group, as shown in the following formula:
[0111] G(v,r,q)=α×heading(v,r,q)+β×distance(v,r,q)+γ×velocity(v,r,q)(28)
[0112] Where heading(v,r,q) is the azimuth scoring function. The smaller the angle between the AUV heading and the target point, the larger the heading(v,r,q) function value is, driving the AUV towards the target point. distance(v,r,q) is the obstacle distance evaluation function. If the distance between the AUV predicted trajectory and the obstacle is greater than the AUV's safe distance, there is no risk of collision. Otherwise, the risk of collision is very high and the predicted trajectory is discarded. velocity(v,r,q) is the speed evaluation function. The faster the speed is while ensuring that there is no obstacle, the larger the velocity(v,r,q) function value is.
[0113] When AUV communicates underwater, especially in open waters, the main communication method is to communicate with floats, USVs, etc., which then transmit the communication information to drones, mother ships or shore base stations. The deeper the water, the more difficult it is for AUV to communicate with the surface, and the greater the communication loss. Therefore, when avoiding obstacles, AUVs often use the floating method to avoid obstacles while ensuring safety. This can reduce the impact of communication losses to a certain extent. By giving the pitch angular velocity [q min ,0] and [0,q min ] area, so that the AUV can consider more about floating operation to avoid obstacles under the same speed group (ν, r, q), so as to obtain less communication loss, higher communication efficiency and signal-to-noise ratio as shown in the following formula:
[0114]
[0115] Where k1 ranges from [0, 0.5). When the AUV is diving (q < 0), the communication loss is large, γ = k1, and a lower score is obtained; when the AUV is surfacing (q < 0), the communication loss is small, γ = 1-k1, and a higher score is obtained. The velocity sampling vector diagram is shown in Figure 6 The dark red area in the figure is the AUV floating speed sampling area (q>0), which has a higher weight than the light red area. When v and r are the same, the speed group with q>0 is preferred to enable the AUV to avoid obstacles by floating and reduce communication losses.
[0116] Step 2.3: Improve the fusion of ACO algorithm and DWA algorithm:
[0117] The ACO algorithm is based on group random search and has good global optimization effects. ACO does not require centralized control and has good scalability and distributed computing capabilities. By combining Pareto sorting, it can handle multi-objective optimization problems and find solutions that balance multiple objectives.
[0118] The DWA algorithm focuses on the current local environment of the AUV, can quickly respond to changes in the surrounding environment, and naturally considers the kinematic and dynamic constraints of the AUV, making it suitable for processing the local path planning and navigation of the AUV.
[0119] The ACO algorithm has a strong global path planning capability, and the DWA algorithm as a local path planning algorithm can handle local optimal problems and dynamic obstacles and adapt to complex and changing environments. Use the improved ACO algorithm to plan a global path, and then use the local planning algorithm to navigate along the reference route obtained by the global path planning, avoiding dynamic obstacles and emergencies in real time. The combination of the two can make full use of global and local information, improve the quality and efficiency of path planning, and is suitable for AUVs that need to respond quickly and navigate effectively in complex environments. Improved ACO and DWA algorithms combined with navigation, such as Figure 7 shown.
[0120] As a preferred embodiment of the present invention, in order to verify the effectiveness and feasibility of the improved ACO algorithm, the improved ACO algorithm was compared with the traditional ACO algorithm, and the measurement indicators were the coverage (Coverage) and distribution uniformity (Spacing or Distribution Uniformity) of the solutions obtained by the two algorithms. Two simulation experiments were designed: first, the improved ACO algorithm was compared with the traditional ACO algorithm, and the effectiveness of the algorithm was verified through 20 groups of comparative experiments. Then the improved DWA algorithm was combined with the improved ACO algorithm to verify the collision avoidance ability of the two algorithms. The algorithm was implemented and run on the following configured computer (Windows 64bit with twelve cores (Intel (R) Core (TM) i7-12700F @ 2.10GHz) and 16GB of RAM) using MATLAB R2022a.
[0121] (1) Comparative experiment between improved ACO algorithm and traditional ACO algorithm:
[0122] Since the improved ACO algorithm introduces the Pareto sorting part, the fitness function is different from the traditional ACO algorithm. The fitness function of the ACO algorithm is shown as follows:
[0123]
[0124] Where L mis the path length searched by the mth ant, SNR m is the signal-to-noise ratio of the path, k1 and k2 are weight coefficients, and Fitness m It is the linear weighting of the above two optimization objectives. The smaller the value, the better the solution. In the simulation experiment, the algorithms all use the same parameters.
[0125] A grid map is used to simulate the underwater depth range of 0 to 200m. The map size is 200×200×200. The AUV is required to navigate safely from the starting point to the target point. The optimization goal is the shortest path and the highest communication signal-to-noise ratio. The simulation map environment is as follows: Figure 8 shown.
[0126] Figure 8 In the figure, the black dot (196,49,-170) and the red dot (30,178,-90) are the starting point and the target point respectively, and the blue area is the seabed terrain. In the experiment, the communication nodes (green hexagonal stars) are set at two different positions and directions on the water surface to compare the search results of the two algorithms.
[0127] The communication nodes are set at the upper left water surface position (0,0,0) and the upper right water surface position (200,200,0) of the line between the starting point and the target point for simulation experiments, and a set of paths searched by the two algorithms are obtained, as shown in Figure 9-12 shown.
[0128] from Fig. 9 and 10 From the observations in , we can see that when the communication nodes are set at different water surfaces, in order to achieve a better balance between path length and communication efficiency, the path as a whole tends to tilt towards the communication node. This trend may be due to the fact that the location of the communication node affects the priority direction of path selection to a certain extent. However, the paths searched by the traditional ant colony algorithm are relatively concentrated and simple in form, and the differences and characteristics between each path are not obvious. This may be because the ant colony algorithm only strengthens the optimal path after each generation of search.
[0129] from Fig.11 and 12 It can be seen that the paths searched by the improved ant colony algorithm are more inclined toward the communication nodes as a whole, and the diversity and characteristics of the path distribution are more obvious. This is because after each generation of search, the ant colony algorithm strengthens the pheromones of a group of paths on the non-dominated frontier, and determines the pheromone increment based on the searched non-dominated solution congestion distance. The better the solution, the greater the pheromone increment.
[0130] The distribution of non-dominated solutions searched by the two algorithms at different locations of communication nodes is as follows: Fig.13 and 14As shown in the figure. FitnessFunction 1 is the fitness function related to the path length, and Fitness Function2 is the fitness function related to the signal-to-noise ratio. The blue points in the figure are the non-dominated solutions searched by the traditional ant colony algorithm, and the red points are the non-dominated solutions searched by the improved ant colony algorithm. It can be seen from the figure that the solutions searched by the improved ant colony algorithm are better and more diverse than the previous solutions. In terms of path length, the search capabilities of the two algorithms are not much different, but in terms of communication efficiency, the improved algorithm is significantly better than the traditional ant colony algorithm. The solution sets obtained by the traditional algorithm and the improved algorithm are respectively represented by X before , X after Indicates that, by calculating the coverage range C and distribution uniformity S of the two to each other, we can get that when the communication node is at (0,0,)0: C(X aft,er X b ) ef = or 0 e .428,6C(X before ,X after )=0.1111,S before =1.0687, S after =0.5364, when the communication node is at (200,200,0): C(X before ,X after )=0.2222,S before =0.6879, S after =0.5989. The improved searched solutions are better than those searched by the traditional algorithm in terms of coverage and distribution uniformity. The results are shown in the following table:
[0131] Table 1 Related experimental results
[0132]
[0133] (2) Improve the simulation experiment combining ACO algorithm and DWA algorithm;
[0134] After the improved ACO searches for the global path, it uses it as a reference path and uses the DWA algorithm to search along the global path to avoid dynamic obstacles encountered in the process. This paper assumes that there are two dynamic obstacles appearing in front of the AUV from different angles and observes the collision avoidance method of the AUV. Without assigning weights, the DWA algorithm searches along the reference path as follows: Fig.15As shown in the figure. The yellow dots represent dynamic obstacles such as marine life or other AUVs, and the green dots represent the actual route of the AUV. As can be seen from the figure, when no weight is assigned, the DWA algorithm avoids obstacles randomly. When encountering the first dynamic obstacle, it avoids the obstacle by diving, but this method will also cause an increase in communication loss as the depth increases. When passing the second dynamic obstacle, it passes from the side. Although it can pass safely, passing from the top will reduce a certain amount of communication loss.
[0135] The DWA algorithm after weighting is as follows: Fig.16 As shown in the figure, it can be seen that after combining the improved ACO algorithm with the DWA algorithm, both static obstacles and dynamic obstacles can be avoided. By assigning different weights to the pitch angular velocity area in the DWA algorithm, the AUV can deal with dynamic obstacles in the environment more flexibly. The AUV can choose to float up to avoid obstacles, which not only ensures safe avoidance but also reduces the diving depth. Doing so helps to improve the communication signal-to-noise ratio because shallower waters usually have better signal transmission characteristics. Therefore, by dynamically adjusting the pitch angular velocity weights, the AUV can more effectively cope with complex underwater environments and improve the efficiency and safety of mission execution.
[0136] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims
1. A comprehensive AUV path planning method for communication efficiency and obstacle avoidance, characterized in that: The following steps are involved: Step 1: Construct AUV communication application scenario and underwater acoustic communication environment model: Step 2: Improvement and fusion of 3D ACO and DWA algorithms: The AUV navigation part will use the improved ant colony algorithm as the global path planning algorithm to avoid static obstacles, and the three-dimensional DWA algorithm will be used as the local path planning algorithm to avoid dynamic obstacles; in the global path planning, the Pareto sorting process will be added to the ant colony algorithm, and the traditional ant colony algorithm will be modified from strengthening the optimal path to strengthening a group of non-dominated paths to achieve the purpose of multi-objective optimization, namely the shortest path and the minimum communication loss.
2. The AUV comprehensive path planning method for communication efficiency and obstacle avoidance according to claim 1 is characterized in that: The step 1 comprises the following specific steps: Step 1.1: Build AUV communication application scenario: AUV underwater communication participants include AUV, surface mother ship, underwater network nodes, buoys, USV and UAV. The communication content includes control instructions, location information and information obtained by AUV; Step 1.2: Construct the AUV underwater acoustic communication environment model: In an underwater environment, the propagation loss TL of sound waves during propagation includes two parts: expansion loss and attenuation loss. The propagation loss in seawater is a function of the distance l and the signal frequency f, which can be approximately expressed as: A(l,f)=l k [a abs (f)] l (1) Where k is the expansion factor, α abs (f) is the seawater absorption loss coefficient; the above formula is expressed in dB form, and the sound transmission loss TL is: TL=10k·log(1000l)+10l·logα abs (f) (2) Where l is the distance of sound wave propagation in km; the first term represents the expansion loss, and the second term represents the absorption loss; the expansion factor k is related to the propagation mode and propagation path of the sound wave. Depending on the propagation conditions, k takes different values, and here it is 1.5; the seawater absorption coefficient α abs (f) Expressed in dB form using Thorp's formula: The selected operating frequency is 20kHz, the system bandwidth is 100kHz, the seawater temperature T is 10℃, and the seawater absorption coefficient α is obtained. abs (f) is 3.30; The ocean ambient noise is described by the cylindrical Wenz model, namely: Where N t 、N sh 、N w and N th represent turbulence noise, shipping noise, sea surface noise and thermal noise respectively; f is the frequency, in kHz; sh is the shipping density in the water area, ranging from 0 to 1; w is the wind speed on the ocean surface, in m / s; turbulent noise N t Impact range: f<10Hz; shipping noise N sh The frequency range is 10~100Hz; the frequency range of sea surface noise caused by wind is 100Hz~100kHz; the thermal noise influence range f>100kHz; the total noise spectrum density synthesized by the above environmental noises is expressed as: N(f)=N t (f)+N sh (f)+N w (f)+N th (f) (5) In a specific frequency band, the noise spectral density decreases linearly with the logarithmic scale, which can be approximately expressed as: 10logN(f)≈N1-ηlogf(6) N1 is the logarithmic value of the noise level at a certain reference frequency, which is regarded as a reference value of the noise level. η describes the slope or attenuation rate of the noise level changing with frequency. When the transmitted signal frequency f and power P are given, the propagation loss A(l,f) obtained by formula (1) and the noise spectral density level N(f) obtained by formula (6) can be used to obtain the signal-to-noise ratio at distance l: Where S(l,f) is the signal power spectrum density at distance l and frequency f, and the system bandwidth is determined as: B(l)=[f min (l),f max (1)] (8) Where B(l) represents the working bandwidth of the system at distance l, f min (l) and f max (l) represent the lowest and highest operating frequencies at distance l respectively, and the received signal-to-noise ratio is shown in formula (9): Integrate the signal-to-noise ratio at each distance point to obtain the average signal-to-noise ratio on the entire path: Among them, L1 and L2 define the range of distance, and the average signal-to-noise ratio from L1 to L2 is obtained; by comparing this average value, the path with better communication quality can be evaluated and selected.
3. The AUV comprehensive path planning method for communication efficiency and obstacle avoidance according to claim 2 is characterized in that: The step 2 comprises the following specific steps: Step 2.1: Improve the mathematical model of ACO algorithm: The visual search space method is used to reduce the difficulty of calculation, so that ants can only visit some nearby nodes, thereby reducing the amount of calculation. The signal-to-noise ratio calculation part is added to the heuristic function, so that ants can consider the impact of path length and communication efficiency when searching for paths. The non-dominated sorting part is introduced when updating global pheromones. The traditional ant colony algorithm is modified from strengthening the pheromones on the shortest path to strengthening the pheromones corresponding to a group of non-dominated solutions on the Pareto frontier, so as to solve the multi-objective optimization problem. Step 2.2: Construct a 3D DWA algorithm: The three-dimensional DWA algorithm is used, and different priorities are assigned to the speed sampling space. The DWA algorithm simulates the motion trajectory of the speed within the prediction period by sampling multiple groups of speeds in the speed space (ν, ω), and evaluates the trajectory through the evaluation function. Finally, the speed group (ν, ω) corresponding to the optimal trajectory is selected to drive the robot movement. Step 2.3: Improve the fusion of ACO algorithm and DWA algorithm: The improved ACO algorithm is used to plan a global path, and then the local planning algorithm is used to navigate along the reference route obtained by the global path planning, avoiding dynamic obstacles and emergencies in real time.
4. The AUV comprehensive path planning method for communication efficiency and obstacle avoidance according to claim 3 is characterized in that: The step 2.1 specifically includes: Visual Search Space: The movement of the AUV is assumed to be forward movement, lateral movement and vertical movement; the maximum forward movement distance of the ant is L x,max , the maximum lateral movement distance is L y,max , the maximum vertical movement distance is L z,max ; When an ant moves from the current node to the next node, the searchable area is limited to the ant's visible search area; State transition probability: At time t, when the mth ant moves from node i to node j, the state transition probability of each optional node around node i is first calculated. As shown in formula (11), node j is then determined by the roulette method; Among them, τ ij (t) is the pheromone concentration from the current node i to the node j at time t; H ij (t) is the path heuristic information from the current node i to node j at time t; allowed m is the set of optional subsequent grids around the current i-th grid for the m-th ant; μ and ρ are the important factors of pheromone and heuristic information respectively, where H ij (t) is calculated as follows: H ij (t)=Safe(x,y,z)×(Distance(x,y,z)+SNR(x,y,z)) (12) Where Safe(x,y,z) indicates whether the point is safe and feasible. If the point is an obstacle, the value is 0, otherwise it is 1, as shown in formula (13): Distance(x,y,z) is the distance between point (x,y,z) and target point (x d ,y d ,z d ), the Euclidean distance between them is calculated as follows: Among them, the molecule K d is a constant, indicating the importance of the path length; SNR(x,y,z) indicates the signal-to-noise ratio of the point (x,y,z), which is obtained from formula (9). The heuristic function design is shown in formula (15): Where S0 is the signal power spectrum density, which represents the signal power level at a specific frequency or frequency band; A l (l,f) is the propagation loss calculated by equations (1) and (2), which is related to the distance between the current point and the target point. B is the bandwidth, which is used to describe the maximum data rate that can be transmitted by the signal or data transmission channel. N(f) is the noise power spectral density at a specific frequency f, which is obtained by equation (16) and is calculated as follows; Pheromone update rules: Pheromone update includes two parts: local update and global update. The local update is shown in formula (17): t ij (t)=(1-ζ)·τ ij (t-1)+ζ·τ0 (17) Where ζ is the attenuation coefficient of pheromone; The Pareto sorting process is introduced during global update to improve the global pheromone update mechanism in the ant colony algorithm. Specifically, in each iteration of the algorithm, a set of non-dominated solutions, namely the Pareto frontier, is identified through Pareto sorting. This process starts from the first generation, and the multi-objective fitness values of the paths found by the ants are Pareto sorted to select the non-dominated solution set, and the global Pareto solution set and its corresponding path set are initialized accordingly. Subsequently, in each iteration of the algorithm, the newly discovered paths are integrated with the existing global Pareto solution set and Pareto sorted again; at the same time, the pheromones on each path in the global Pareto solution set are strengthened; The entire search process continuously updates and strengthens the global Pareto solution set to find a solution set that meets the requirements to better approximate the real Pareto frontier; Pheromone Update τ ij (t+1) and the pheromone increment Δτ ij (t) is as shown in equations (18)-(19): t ij (t+1)=(1-ρ)·τ ij (t)+ρ·Δτ ij (t) (18) Among them, C(sol) is the crowding degree of the solution sol in the global Pareto frontier GlobalParetoSet. Under the premise of strengthening the pheromone on the path corresponding to the non-dominated solution, the smaller the crowding degree, the larger the pheromone increment Δτ, and the larger the crowding degree, the smaller the pheromone increment Δτ, so as to ensure the diversity of solutions under the premise of excellent solutions; ρ is the pheromone update coefficient; Q is a constant; Fitness function design: The two optimization goals to be solved are path length and communication efficiency. The communication efficiency is characterized by the signal-to-noise ratio. After all ants in each generation have finished searching, the paths of all ants are evaluated using the fitness function. The fitness function is designed as shown in equations (20) and (21): fitness 1,m =judge·k1·L m (20) Among them, fitness 1,m is the first fitness function value of the mth ant searching for the path, fitness 2,m is the second fitness function value of the mth ant searching for the path, L m is the length of the path searched by the mth ant, SNR avg,m is the average signal-to-noise ratio of the path searched by the mth ant, k1 and k2 are constants; use (fitness 1,m ,fitness 2,m ) represents the solution to the multi-objective optimization problem. The smaller the value of the fitness function, the better the solution obtained and the more ideal the path obtained.
5. The AUV integrated path planning method for communication efficiency and obstacle avoidance according to claim 4 is characterized in that: The step 2.2 specifically includes: AUV kinematic model: Assume that the underwater AUV motion model is a non-omnidirectional mobile robot. The AUV realizes forward and backward motion through a fixed thruster, and adjusts the heading through a limited angle steering jet system. The state vector x and control input u of the AUV are as follows: x=[x,y,z,ψ,θ,ν,r,q] T (22) u=[v,r,q] (23) Where x, y, z are the position coordinates of the robot in three-dimensional space, ψ, θ are the pitch angle and pitch angle, respectively representing the inclination of the AUV in the horizontal plane and vertical direction, v is the forward speed of the AUV, r, q are the angular velocities around the z-axis and y-axis respectively, and the relationship between the position, orientation and speed of the robot at the next moment t+Δt and the current moment t is as follows: AUV velocity sampling space: Due to the limitations of AUV hardware and structure, the velocity sampling space V s (v, r, q) is limited to a certain range; according to the hardware conditions of the AUV, there is first a speed boundary limit, and the sampled speed space V m for: V m ={(ν,r,q)|v∈[v min ,v max ],r∈[r min ,r max ],q∈[q min ,q max ]} (25) Due to the influence of motor performance, the acceleration has a limited range. There is a dynamic window in the forward simulation cycle of the AUV moving trajectory. The speed within this window is the achievable speed of the AUV: where v c ,r c ,q c is the current speed of the AUV, is the maximum deceleration of the AUV, is the maximum acceleration of the AUV. Based on safety considerations, the AUV can decelerate in time when encountering an obstacle, and there is a maximum deceleration limit: The actual feasible speed range after considering the above three constraints is V m ∩V d ∩V a ; After obtaining the speed space, different speed combinations (ν, r, q) can be obtained, and they can be scored to select the optimal speed group, as shown in the following formula: G(v,r,q)=α×heading(v,r,q)+β×distance(v,r,q)+γ×velocity(v,r,q)(28) Where heading(v,r,q) is the azimuth scoring function. The smaller the angle between the AUV heading and the target point, the larger the heading(v,r,q) function value is, driving the AUV towards the target point. distance(v,r,q) is the obstacle distance evaluation function. If the distance between the AUV predicted trajectory and the obstacle is greater than the safe distance of the AUV, there is no risk of collision. Otherwise, the risk of collision is very high and the predicted trajectory is discarded. velocity(v,r,q) is the speed evaluation function. The faster the speed is while ensuring that there is no obstacle, the larger the velocity(v,r,q) function value is. When the AUV communicates underwater, it gives the pitch angular velocity [q min ,0] and [0,q max ] area, so that the AUV can consider more about floating operation to avoid obstacles under the same speed group (ν, r, q), so as to obtain less communication loss, higher communication efficiency and signal-to-noise ratio, as shown in the following formula: Wherein, the value range of k1 is [0, 0.5). When the AUV is diving (q < 0), the communication loss is large, γ = k1, and a lower score is obtained; when the AUV is surfacing (q < 0), the communication loss is small, γ = 1-k1, and a higher score is obtained.
Citation Information
Cited By
Path planning method and system for automatic driving trolley
CN119472685A
Path Planning Method and System for an Autonomous Vehicle
CN119472685B
Sea-air integrated cross-domain heterogeneous collaborative planning method
CN119717868A
Multi-target ship communication path planning method based on offshore waveguide channel map
CN120074715A
Cooperative path planning method, system and equipment for ocean cross-domain unmanned system
CN120506959A
