An autonomous underwater vehicle path planning method, system and electronic device

Abstract

The application relates to the technical field of vehicle path planning, and discloses an autonomous underwater vehicle path planning method, system and electronic equipment, which comprises the following steps: obtaining and processing initial feature field data to obtain high-resolution feature field data; using the high-resolution feature field data to assign target observation areas to multiple vehicles; extracting local feature field data corresponding to each vehicle, and initializing a local Gaussian process model corresponding to the vehicle; dynamically updating the local Gaussian process model using real-time feature field data collected by the vehicle, and predicting local feature field prediction data around each vehicle using the local Gaussian process model; planning a vehicle path for each vehicle based on the local feature field prediction data and using an improved ant colony algorithm; and monitoring the task execution state of each vehicle, and reassigning tasks or ending tasks. The application can efficiently assign task areas to multiple vehicles and plan paths for the vehicles, so that adaptive and efficient observation of dynamic ocean features is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of underwater vehicle path planning, and in particular to an autonomous underwater vehicle path planning method, system and electronic device. BACKGROUND

[0002] An autonomous underwater vehicle (AUV) plays an important role in the field of ocean observation, and its adaptive sampling path planning method is one of the key technologies. Existing AUV underwater adaptive sampling path methods mainly fall into two categories. One is a planning method based on preset rules. The AUV dynamically adjusts the sampling strategy according to real-time observed environmental information or prior knowledge during task execution. This type of method, such as threshold method, trigger mechanism method and region division method, is simple to implement and has low computational cost, but it relies on artificially designed rules and is difficult to adapt to complex and dynamically changing environments. Moreover, since the sampling strategy is adjusted based on local information, it cannot optimize the sampling path from a global perspective, which can easily lead to resource waste or important areas not being covered. The other is a planning method based on model-predicted information of environmental characteristics. This method relies on mathematical models and optimization algorithms to accurately describe environmental characteristics and predict trends. The purpose is to optimize sampling behavior and plan paths systematically to enable the AUV to obtain as much useful data as possible within limited resources, thereby improving task execution efficiency and data quality. However, the optimization algorithm in the traditional model-based method usually has a large amount of calculation, and it is difficult to meet the performance requirements in tasks with high real-time requirements. Moreover, it requires sufficient historical data and real-time observation data to support model construction and updating, and when the data is insufficient, the path optimization effect will be greatly reduced. Therefore, based on the difficulties of complex and dynamically changing environments, how to efficiently allocate task areas and improve the adaptability and accuracy of path planning to meet the needs of feature field observation, thereby achieving efficient observation of the marine environment, is a problem that needs to be solved. SUMMARY

[0003] The main purpose of the present application is to overcome the shortcomings and deficiencies of the prior art, and to provide an autonomous underwater vehicle path planning method, system and electronic device. The local feature field is predicted based on Gaussian process regression, breaking through the limitation of traditional adaptive path planning that cannot cope with complex and dynamically changing environments. The problem of complex and large amount of calculation of global model prediction is solved. The task area can be efficiently allocated and the paths of each AUV can be planned to meet the needs of feature field observation, thereby achieving efficient adaptive observation of the marine environment.

[0004] In order to achieve the above purpose, the following technical solutions are adopted in the present application:

[0005] In a first aspect, the present application provides an autonomous underwater vehicle path planning method, comprising:

[0006] obtaining initial feature field data, preprocessing the initial feature field data to obtain high-resolution feature field data;

[0007] allocating target observation areas for the multiple vehicles to perform observation tasks using the high-resolution feature field data;

[0008] extracting local feature field data around a current position of each vehicle based on the high-resolution feature field data, and initializing a local Gaussian process model corresponding to the observation task performed by each vehicle based on the local feature field data;

[0009] dynamically updating the local Gaussian process model using real-time feature field data collected by each vehicle, and predicting local feature field prediction data around each vehicle based on the dynamically updated local Gaussian process model;

[0010] planning a navigation path for each vehicle using an improved ant colony algorithm based on the local feature field prediction data of each vehicle;

[0011] monitoring a task execution state of each vehicle based on the navigation path, and reassigning tasks or ending tasks based on the task execution states of all vehicles.

[0012] As a preferred technical solution, obtaining initial feature field data, preprocessing the initial feature field data to obtain high-resolution feature field data, comprises:

[0013] obtaining initial feature field data of a to-be-observed area, and performing interpolation processing on the initial feature field data using a preset interpolation method to obtain high-resolution feature field data, the resolution of the high-resolution feature field data being higher than that of the initial feature field data.

[0014] As a preferred technical solution, allocating target observation areas for the multiple vehicles to perform observation tasks using the high-resolution feature field data, comprises:

[0015] dividing multiple to-be-observed areas with different feature information based on the high-resolution feature field data using a clustering algorithm, and calculating a feature recognition criterion value of the to-be-observed areas;

[0016] determining a feature value of interest from the feature recognition criterion value based on a preset feature recognition criterion threshold, and marking a region corresponding to the clustering of the feature value of interest as a region of interest;

[0017] performing secondary classification on the region of interest to determine multiple sub-regions of interest, and dividing the multiple sub-regions of interest into a plurality of target observation areas to be observed according to a preset evaluation index and the number of vehicles;

[0018] According to the evaluation result of the target observation area, different numbers of vehicles are dispatched to the divided target observation areas for observation.

[0019] As a preferred technical solution, the real-time feature field data collected by each vehicle is used to dynamically update a local Gaussian process model, and the local feature field prediction data around each vehicle is predicted based on the dynamically updated local Gaussian process model, including:

[0020] An input feature matrix of the local Gaussian process model is defined;

[0021] Based on the real-time feature field data collected by each vehicle on the flight path in the target observation area, a target output vector is obtained;

[0022] The input feature matrix and the target output vector are used to update the hyperparameters through an optimization function to update the local Gaussian process model;

[0023] The input feature matrix is input into the updated local Gaussian process model to generate the local feature field prediction data of each vehicle.

[0024] As a preferred technical solution, based on the local feature field prediction data of each vehicle, an improved ant colony algorithm is used to plan the flight path of each vehicle, including:

[0025] An information matrix of the ant colony algorithm is defined to obtain the information concentration of the target observation area;

[0026] Based on the local feature field prediction data of the current path point of the vehicle, the feature recognition criterion value of the feature values of the current path point and the adjacent path point changing with the spatial position is calculated;

[0027] Based on a preset state transition function, the transition probability from the current path point to the adjacent path point is calculated in combination with the travel direction of the vehicle from the current path point to the adjacent path point, the feature recognition criterion value, and the information concentration of the target observation area, and the next path point is determined based on the transition probability;

[0028] Real-time feature field data from the current path point to the next path point is obtained, a feature field data change value is determined, the information matrix is updated based on the feature field data change value, and a single ant completes one path planning;

[0029] When all ants complete path planning, the best flight path is determined as the flight path of each vehicle through a preset path evaluation function.

[0030] As a preferred technical solution, based on the transition probability to determine the next path point, it further includes:

[0031] Check whether the aircraft is out of the target observation area, if out of the target observation area, start the re-discovery behavior, detect the adjacent path point in the preset search mode, and detect the adjacent path point meeting the preset condition as the next path point.

[0032] As a preferred technical solution, based on the navigation path, the task execution state of each aircraft is monitored, and the task is re-assigned or ended based on the task execution state of all aircrafts, comprising:

[0033] Based on the distribution of the navigation path, the coverage rate of the navigation path of the aircraft to the target observation area is calculated in real time, and the task execution state of each aircraft is determined based on the coverage rate;

[0034] When it is monitored that the task execution state of the aircraft meets the preset task completion condition, a global Gaussian process model is constructed based on the real-time feature field data collected by all aircrafts;

[0035] The global uncertainty information of the feature field is generated by using the global Gaussian process model, and a new target observation area is assigned to the aircraft completing the task based on the global uncertainty information.

[0036] As a preferred technical solution, it also comprises:

[0037] When the global uncertainty information decreases to a preset uncertainty threshold, the observation task of all aircrafts is ended.

[0038] Secondly, the application provides an autonomous underwater vehicle path planning system for executing the autonomous underwater vehicle path planning method of the first aspect, comprising:

[0039] A data acquisition module is configured to acquire initial feature field data, and pre-process the initial feature field data to obtain high-resolution feature field data;

[0040] A task allocation module is configured to allocate target observation areas for multiple aircrafts to execute observation tasks by using the high-resolution feature field data;

[0041] A local model initialization module is configured to extract local feature field data around the current position of each aircraft based on the high-resolution feature field data, and initialize a local Gaussian process model corresponding to the observation task of each aircraft based on the local feature field data;

[0042] A local feature field prediction module is configured to dynamically update the local Gaussian process model by using real-time feature field data collected by each aircraft, and predict local feature field prediction data around each aircraft based on the dynamically updated local Gaussian process model;

[0043] a path planning module, configured to plan a navigation path of each vehicle based on the local feature field prediction data of each vehicle by using an improved ant colony algorithm;

[0044] a task monitoring module, configured to monitor a task execution state of each vehicle based on the navigation path, and to perform task re-allocation or end the task based on the task execution states of all vehicles.

[0045] In a third aspect, the present application provides an electronic device, which comprises:

[0046] at least one processor; and a memory connected to the at least one processor in communication;

[0047] The memory stores computer program instructions executable by the at least one processor, and the computer program instructions are executed by the at least one processor to enable the at least one processor to perform the autonomous underwater vehicle path planning method.

[0048] In summary, compared with the prior art, the technical scheme provided by the present application has at least the following effective effects:

[0049] The present application provides an autonomous underwater vehicle path planning method, which obtains initial feature field data, pre-processes the initial feature field data to obtain high-resolution feature field data, allocates target observation areas for multiple vehicles to perform observation tasks by using the high-resolution feature field data, extracts local feature field data around a current position of each vehicle based on the high-resolution feature field data, initializes a local Gaussian process model corresponding to the observation task performed by each vehicle based on the local feature field data, dynamically updates the local Gaussian process model by using real-time feature field data collected by each vehicle, predicts local feature field prediction data around each vehicle based on the dynamically updated local Gaussian process model, plans a navigation path of each vehicle by using an improved ant colony algorithm based on the local feature field prediction data of each vehicle, monitors a task execution state of each vehicle based on the navigation path, and performs task re-allocation or ends the task based on the task execution states of all vehicles. The local environment information can be more accurately obtained, the local prediction is performed by using the local Gaussian process model, and the limitations of the traditional adaptive path planning in dealing with complex and dynamically changing environments are broken through. Secondly, the improved ant colony adaptive path planning algorithm is combined with the local feature field prediction information obtained by using the local Gaussian process model to optimize the navigation paths of multiple autonomous underwater vehicles, so that the autonomous underwater vehicles can better adapt to the complex and dynamically changing marine environment, a more reasonable navigation path is planned, the adaptability and accuracy of the path planning are improved, and the marine environment is efficiently observed.

[0050] In addition, the application calculates, in real time, coverage of a flight path of the aerial vehicle on the target observation area based on distribution of the flight path, determines a task execution state of each aerial vehicle based on the coverage, constructs a global Gaussian process model based on real-time feature field data collected by all the aerial vehicles when it is monitored that the task execution state of the aerial vehicle meets a preset task completion condition, generates global uncertainty information of the feature field by using the global Gaussian process model, and allocates a new target observation area for the aerial vehicle completing the task based on the global uncertainty information. When the global uncertainty information is reduced to a preset uncertainty threshold, the overall observation task is completed, and the uncertainty of global information of the feature field is reduced. Based on the global uncertainty information of the feature field, the multiple autonomous underwater vehicles are dynamically re-allocated to different areas for observation, the observation area range of the multiple autonomous underwater vehicles for cooperative observation is wider, the observation efficiency is higher, and more useful sampling data are obtained, and the problem of waste of resources or important areas not covered in the traditional method is effectively solved. BRIEF DESCRIPTION OF DRAWINGS

[0051] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0052] Figure 1 A flowchart of an autonomous underwater vehicle path planning method provided by an embodiment of the present application is shown in the figure.

[0053] Figure 2 A flowchart of an autonomous underwater vehicle path planning method provided by another embodiment of the present application is shown in the figure.

[0054] Figure 3 A flowchart of a multiple autonomous underwater vehicle task area allocation method provided by an embodiment of the present application is shown in the figure.

[0055] Figure 4 A flowchart of a Gaussian process model generating local feature field prediction information or uncertainty information provided by an embodiment of the present application is shown in the figure.

[0056] Figure 5 A flowchart of adaptive path planning by using an improved ant colony algorithm provided by an embodiment of the present application is shown in the figure.

[0057] Figure 6 A flowchart of adaptive path planning by using an improved ant colony algorithm provided by another embodiment of the present application is shown in the figure.

[0058] Figure 7A structural schematic diagram of an autonomous underwater vehicle path planning system provided in an embodiment of the present application;

[0059] Explanation of reference signs:

[0060] The data acquisition module 101, the task allocation module 102, the local model initialization module 103, the local feature field prediction module 104, the path planning module 105, and the task monitoring module 106. DETAILED DESCRIPTION

[0061] In order for those skilled in the art to better understand the scheme of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0062] In the present application, the phrase "embodiment" means that the specific features, structures or characteristics described in conjunction with the embodiment can be included in at least one embodiment of the present application. The appearance of this phrase at various places in the specification does not necessarily mean the same embodiment, nor is it an independent or alternative embodiment to other embodiments. Those skilled in the art explicitly and implicitly understand that the embodiments described in the present application can be combined with other embodiments.

[0063] Please refer to Figure 1 and Figure 2 In an embodiment of the present application, an autonomous underwater vehicle path planning method is provided, comprising the following steps:

[0064] S1: obtaining initial feature field data, pre-processing the initial feature field data to obtain high-resolution feature field data.

[0065] First, the initial feature field data of the area to be monitored is obtained by satellite remote sensing technology. The initial feature field data is coarse-resolution feature field data, i.e. feature field data with low spatial resolution. Low spatial resolution means that the sampling interval in the spatial dimension is greater than a preset threshold (the sampling interval in the spatial dimension is large), resulting in relatively sparse distribution of data points and less data. The feature field data can be temperature data, salinity data, flow field data, etc. Temperature data can identify temperature fronts, salinity data can identify salinity gradients, and flow field data can analyze eddies or coastal currents, etc.

[0066] S2: using the high-resolution feature field data to allocate target observation areas for multiple vehicles to execute observation tasks.

[0067] Specifically, the high-resolution feature field data is used to divide the to-be-observed region into multiple target observation regions through a clustering algorithm, and the target observation regions are assigned to the respective vehicles.

[0068] S3: Based on the high-resolution feature field data, local feature field data around the current position of each vehicle is extracted, and a local Gaussian process model corresponding to the observation task performed by each vehicle is initialized based on the local feature field data.

[0069] Specifically, the local Gaussian process model is initialized through the local feature field information. The expression of the initially constructed local Gaussian process model is the expression of a normal Gaussian process, which delivers an initial hyperparameter. This process provides a local feature information framework for the local region and reveals the main trend and change pattern of the local region.

[0070] S4: The local Gaussian process model is dynamically updated using real-time feature field data collected by each vehicle, and local feature field prediction data around each vehicle is predicted based on the dynamically updated local Gaussian process model.

[0071] S5: Based on the local feature field prediction data of each vehicle, an improved ant colony algorithm is used to plan a navigation path for each vehicle.

[0072] S6: Based on the navigation path, the task execution state of each vehicle is monitored, and the task is re-assigned or ended based on the task execution state of all vehicles.

[0073] In step S1, the initial feature field data is obtained, and the initial feature field data is preprocessed to obtain high-resolution feature field data, including:

[0074] The initial feature field data of the to-be-observed region is obtained, and the initial feature field data is processed by a preset interpolation method to obtain high-resolution feature field data. The resolution of the high-resolution feature field data is higher than that of the initial feature field data. The high-resolution high-resolution feature field data refers to a sampling interval in the spatial dimension that is less than a preset threshold.

[0075] In this embodiment, the initial feature field data is processed by the Kriging interpolation method to obtain high-resolution feature field data.

[0076] Please refer to Figure 3 , step S2 specifically includes:

[0077] S21: Based on the high-resolution feature field data, multiple to-be-observed regions with different feature information are divided through a clustering algorithm, and a feature recognition criterion value of the to-be-observed region is calculated. The feature recognition criterion value refers to a key quantitative indicator of the to-be-observed region.

[0078] The embodiment adopts a K_MEANS clustering method to cluster high-resolution feature field data, and obtains a to-be-observed region with different feature information.

[0079] S22: determining a feature value of interest from the feature recognition criterion value based on a preset feature recognition criterion threshold, and marking a region corresponding to the clustering of the feature value of interest as a region of interest;

[0080] Specifically, the feature recognition criterion values of the plurality of to-be-observed regions with different feature information are compared with the preset feature recognition criterion threshold, and the feature value corresponding to the feature recognition criterion value greater than the preset feature recognition criterion threshold is determined as the feature value of interest.

[0081] S23: performing secondary classification on the region of interest to determine a plurality of sub-regions of interest, and dividing the plurality of sub-regions of interest into a plurality of target observation regions to be observed according to a preset evaluation index and the number of vehicles.

[0082] In actual implementation, the region of interest is classified again according to a density-based clustering method (DBSCAN) method with noise. The area and the feature recognition criterion value of the region after the secondary classification are used as an evaluation function, and the classified regions are divided into high-quality to-be-observed regions and low-quality to-be-observed regions according to the evaluation index from high to low.

[0083] S24: dispatching different numbers of vehicles to the divided target observation regions for observation according to the evaluation results of the target observation regions.

[0084] Specifically, different numbers of vehicles are dispatched to the high-quality to-be-observed regions and the low-quality to-be-observed regions for observation according to the quality of the target observation regions. More autonomous underwater vehicles are dispatched to the high-quality regions.

[0085] Further, please refer to Figure 4 , the specific steps of step S4 include:

[0086] S41: defining an input feature matrix of a local Gaussian process model;

[0087] S42: obtaining a target output vector based on real-time feature field data collected by each vehicle on a target observation region flight path;

[0088] S43: updating the local Gaussian process model by updating hyperparameters through an optimization function using the input feature matrix and the target output vector;

[0089] S44: inputting the input feature matrix into the updated local Gaussian process model to generate local feature field prediction data of each vehicle.

[0090] Wherein, by using the input feature matrix and the target output vector, the hyperparameters are updated by optimizing the function to update the local Gaussian process model, and specifically further comprising:

[0091] Calculate the covariance matrix K: calculate the covariance matrix K according to the input two-dimensional feature matrix X and the covariance function k(x,x'), for example, the covariance function adopts the isotropic Rational Quadratic (RQ) kernel function:

[0092]

[0093] In the formula, σ 2 is the signal variance, α is the shape parameter, and l is the length scale. It should be noted that other kernel functions can also be selected according to requirements.

[0094] The log marginal likelihood estimation method is used to optimize the hyperparameters in the embodiments of the application, and the Laplace approximation method, the Leave-One-Out approximation method, and the sparse likelihood approximation method can also be used for optimization.

[0095] Taking the log marginal likelihood estimation method as an example for description:

[0096] Calculate the log marginal likelihood function: according to the likelihood function Calculate the log marginal likelihood function:

[0097]

[0098] Wherein, y is the output target vector, X is the input two-dimensional feature matrix, K is the covariance matrix, is the noise variance, I is the unit matrix, and n is the number of data points.

[0099] Optimize the hyperparameters: optimize the hyperparameters by maximizing the log marginal likelihood function; during the optimization process, the hyperparameters are constantly adjusted until the log marginal likelihood function reaches the maximum value or meets the convergence condition.

[0100] Finally, the updated hyperparameter value is obtained through the above optimization process. The updated hyperparameter value will replace the initial hyperparameters and be used to update the local Gaussian process regression model. The updated hyperparameters can better describe the distribution characteristics of the data and improve the prediction accuracy of the model.

[0101] Use the updated local Gaussian process regression model for prediction, and the prediction distribution is:

[0102] f * |X,y,X * ~N(m * ,∑ * );

[0103] wherein m * is the predicted mean, and * is the predicted covariance matrix; through the predicted distribution, the local feature prediction field data can be obtained, which is used for the next path point selection of the autonomous underwater vehicle.

[0104] In the present application, the accuracy of the local environmental information is gradually improved through dynamic updating of the parameters in the local Gaussian process model. With the passage of time and the increase in data volume, the model is gradually corrected and optimized, so that the accuracy of local observation is continuously improved. Through the updated Gaussian process regression model, local feature field prediction information is continuously obtained for the next step selection of the autonomous underwater vehicle.

[0105] With reference to Figure 5 , in step S5, based on the local feature field prediction data of each vehicle, an improved ant colony algorithm is used to plan the navigation path of each vehicle, specifically including:

[0106] S51: defining the pheromone matrix of the ant colony algorithm, and obtaining the pheromone concentration of the target observation area;

[0107] S52: based on the local feature field prediction data of the current path point of the vehicle, calculating the feature recognition criterion value of the feature value of the current path point and the adjacent path point changing with the spatial position;

[0108] S53: based on the preset state transition function, combining the travel direction of the vehicle from the current path point to the adjacent path point, the feature recognition criterion value, and the pheromone concentration of the target observation area, calculating the transition probability from the current path point to the adjacent path point, and determining the next path point based on the transition probability;

[0109] S54: obtaining the real-time feature field data from the current path point to the next path point, determining the feature field data change value, updating the pheromone matrix based on the feature field data change value, and completing the path planning of a single ant;

[0110] S55: when all ants complete the path planning, determining the best navigation path as the navigation path of each vehicle through the preset path evaluation function.

[0111] Further, after determining the next path point based on the transition probability, it further includes:

[0112] checking whether the vehicle has left the target observation area, if the vehicle has left the target observation area, starting a re-discovery behavior to detect the adjacent path points in a preset search mode, and detecting the adjacent path point meeting the preset condition as the next path point.

[0113] For example, please refer to Figure 6, initialize the state of each vehicle and the parameters of the improved ant colony algorithm, wherein the parameters of the improved ant colony algorithm include basic parameters of the ant colony algorithm, such as the number of ants, the initial concentration of pheromone, the pheromone evaporation factor, the heuristic factor, the maximum number of iterations, etc.; in each iteration, each ant obtains local feature field prediction data of each vehicle centered on the current position by using a local Gaussian process model according to the current position; the local feature field prediction data includes temperature characteristic values, salinity characteristic values or flow field characteristic values; based on the local feature field prediction data, the travel direction of the vehicle from the current point to the adjacent path point, the feature recognition criterion value of the feature difference between the current point and the next point changing with the spatial position, and the pheromone concentration of the target observation area, the transition probability of the next point is calculated;

[0114] The feature difference ΔT is expressed as:

[0115] ΔT = |T(x curren ) - T(x next ) |,

[0116] wherein T(x current ) is the characteristic value of the current position, and T(x next ) is the characteristic value of the candidate position.

[0117] Check whether the current point is in the taboo table, if it is in the taboo table, skip the point and select the next point; the taboo table is used to record the points that have been visited by the vehicle and the points within the preset range of the visited points, to ensure that the recently visited nodes are not selected repeatedly, so as to avoid a circular path.

[0118] Use the roulette wheel algorithm to select the next point according to the calculated transition probability;

[0119] By selecting the next point according to the transition probability combined with the roulette wheel algorithm, the diversity of the path can be increased and the local optimum can be avoided.

[0120] Determine the position of the next point, and add the current point and the adjacent points within the preset range of the current point to the taboo table;

[0121] Based on the position of the next point, the vehicle goes to the next point and obtains the characteristic value data on the path as real-time update data;

[0122] After the vehicle completes each step of movement, update the pheromone matrix based on the characteristic value data on the path;

[0123] After each step of movement is completed, update the pheromone concentration on the path according to the collected feature difference. For example, increase the pheromone concentration in the path segment where the feature changes significantly, to guide the subsequent autonomous underwater vehicle to preferentially select these paths.

[0124] Based on the updated pheromone matrix, the best path is updated through a path evaluation function, and a path with a feature difference between different features greater than a preset value is selected as a new best path.

[0125] Cumulative feature value change update: the feature difference after selecting the next point each time is accumulated and used as a basis for path evaluation, and the cumulative feature value change AT all The formula is:

[0126] ΔT all = AT all + ΔT.

[0127] Path evaluation update: compare the cumulative feature difference of each path, and select the path with higher cumulative feature difference as the new best path. The formula for path evaluation update is:

[0128] ΔT all_best = max(ΔT all_best , ΔT all,k ),

[0129] Where k represents the kth path in the current iteration.

[0130] Detect whether all ants complete a path selection, if not, repeat the above steps to complete an iteration;

[0131] Detect whether the current iteration number reaches the maximum iteration number, if so, output the optimal path.

[0132] The present application improves the state transition function and pheromone update method of the traditional ant colony algorithm based on the feature value, and realizes the optimization of the target feature.

[0133] Specifically, during the execution of step S6, based on the distribution of the navigation path, the coverage rate of the navigation path of the vehicle on the target observation area is calculated in real time, and the task execution state of each vehicle is determined based on the coverage rate;

[0134] When it is detected that the task execution state of the vehicle meets the preset task completion condition, a global Gaussian process model is constructed based on the real-time feature field data collected by all vehicles;

[0135] The global uncertainty information of the feature field is generated using the global Gaussian process model, and a new target observation area is assigned to the vehicle that has completed the task based on the global uncertainty information. The vehicle that has completely observed the task is assigned a target observation area with high uncertainty for further observation, thereby reducing the global uncertainty.

[0136] Further, when the global uncertainty information is reduced to a preset uncertainty threshold, the observation task of all vehicles is ended.

[0137] In the embodiment of the application, after the initial allocation is completed, when it is judged that a certain area is observed in the process of the autonomous underwater vehicle traveling and observing, in order to optimize the observation, the second task allocation is performed, the autonomous underwater vehicle that completes the observation task is sent to the area with a greater uncertainty than the preset uncertainty (i.e., high uncertainty) according to the uncertainty information, so that the observation area range of the collaborative observation of the multiple autonomous underwater vehicles is more extensive.

[0138] The task allocation of the multiple autonomous underwater vehicles maximizes the observation of the unknown feature field based on the prediction information, dynamically allocates the observation task according to the online feature field prediction information, and uses the uncertainty information of the Gaussian process for task re-allocation, so that the observation efficiency is higher and more useful sampling data is obtained.

[0139] The embodiment of the application uses simple K_MEANS clustering and DBSCAN clustering to cluster the feature values, realizes the allocation of the area, and realizes the second allocation in combination with the uncertainty after the global Gaussian process, and improves the observation efficiency of the multiple autonomous underwater vehicles. According to the real-time prediction information of the feature field, the task allocation of the multiple autonomous underwater vehicles is performed, and when the multiple autonomous underwater vehicles detect that a certain area is observed, the global prediction is performed by using the global Gaussian process regression to determine the uncertainty of the global prediction information. The task allocation is performed again according to the uncertainty, and then the observation is performed again, so as to maximize the observation of the feature area and be more close to the observation of the complex and dynamic feature field by the multiple autonomous underwater vehicles.

[0140] In the application, the local prediction is performed by using the local Gaussian process regression, the limitation that the traditional adaptive path planning cannot cope with the complex and dynamically changing environment is broken through, the problem that the global model prediction calculation is complex and the calculation amount is large in the current planning method based on the model of the dynamic prediction information of the environmental features is improved, the purpose of quickly predicting by using simple information to optimize the observation is achieved, and the reliability of the model prediction is increased according to the real-time data updating the hyperparameters of the model.

[0141] Based on the same idea as the path planning method of the autonomous underwater vehicle in the above embodiment, the application also provides an autonomous underwater vehicle path planning system, which can be used to execute the path planning method of the autonomous underwater vehicle. In order to facilitate the description, only the part related to the embodiment of the application is shown in the structural schematic diagram of the autonomous underwater vehicle path planning system embodiment, and those skilled in the art can understand that the illustrated structure does not constitute a limitation on the system, and the system can include more or fewer components than the illustrated components, or some components can be combined, or different components can be arranged.

[0142] Please refer to Figure 7In another embodiment of the present application, an autonomous underwater vehicle path planning system is provided, which comprises: a data acquisition module 101, a task allocation module 102, a local model initialization module 103, a local feature field prediction module 104, a path planning module 105, and a task monitoring module 106;

[0143] The data acquisition module 101 is configured to acquire initial feature field data, pre-process the initial feature field data, and obtain high-resolution feature field data.

[0144] The task allocation module 102 is configured to allocate target observation areas for multiple vehicles to perform observation tasks using the high-resolution feature field data.

[0145] The local model initialization module 103 is configured to extract local feature field data around the current position of each vehicle based on the high-resolution feature field data, and initialize a local Gaussian process model corresponding to the observation task performed by each vehicle based on the local feature field data.

[0146] The local feature field prediction module 104 is configured to dynamically update the local Gaussian process model using real-time feature field data collected by each vehicle, and predict local feature field prediction data around each vehicle based on the dynamically updated local Gaussian process model.

[0147] The path planning module 105 is configured to plan a navigation path for each vehicle using an improved ant colony algorithm based on the local feature field prediction data of each vehicle.

[0148] The task monitoring module 106 is configured to monitor the task execution state of each vehicle based on the navigation path, and perform task re-allocation or end the task based on the task execution state of all vehicles.

[0149] In some preferred embodiments, the data acquisition module 101 is specifically configured to perform interpolation processing on the initial feature field data using an interpolation method to obtain high-resolution feature field data, and the resolution of the high-resolution feature field data is higher than that of the initial feature field data.

[0150] In some preferred embodiments, the task allocation module 102 is specifically configured to:

[0151] divide the high-resolution feature field data into multiple to-be-observed areas with different feature information using a clustering algorithm, and calculate a feature recognition criterion value of the to-be-observed area;

[0152] determine a feature value of interest from the feature recognition criterion value based on a preset feature recognition criterion threshold, and mark a region corresponding to the clustering of the feature value of interest as a region of interest;

[0153] The region of interest is classified twice to determine a plurality of sub-regions of interest, and the plurality of sub-regions of interest are divided into a plurality of target observation regions to be observed according to a preset evaluation index and the number of vehicles.

[0154] According to the preset evaluation index of the target observation region, different numbers of vehicles are dispatched to the divided target observation region for observation.

[0155] It should be noted that the autonomous underwater vehicle path planning system of the present application corresponds to the autonomous underwater vehicle path planning method of the present application. The technical features and advantages described in the embodiments of the autonomous underwater vehicle path planning method are applicable to the embodiments of the autonomous underwater vehicle path planning system. For specific content, please refer to the description in the method embodiments. Here, no further description is given, and this is hereby declared.

[0156] In addition, in the implementation of the autonomous underwater vehicle path planning system of the above-mentioned embodiments, the logical division of each program module is only an example. In actual application, the above-mentioned function allocation can be completed by different program modules according to the needs, for example, considering the configuration requirements of the corresponding hardware or the convenience of software implementation. That is, the internal structure of the autonomous underwater vehicle path planning system is divided into different program modules to complete all or part of the functions described above.

[0157] In another embodiment, an electronic device for implementing the autonomous underwater vehicle path planning method is provided, which includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor; when the processor executes the computer program, the autonomous underwater vehicle path planning method of any one embodiment of the present application is implemented.

[0158] For example, in this embodiment, the computer program can be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present application. The one or more modules can be a series of computer program instruction segments that can complete a specific function, which are used to describe the execution process of the computer program in the device.

[0159] The device can be a desktop computer, a notebook, a palm computer, and a cloud server, etc. The device can include, but is not limited to, a processor and a memory.

[0160] The processor can be a central processing unit (CPU), and can also be other general-purpose processors, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor or the like, which is the control center of the device, and connects all parts of the device through various interfaces and lines. The memory can be used to store the computer programs and / or modules, and the processor realizes various functions of the device by running or executing the computer programs and / or modules stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application program required by a function, etc.; in addition, the memory can include a high-speed random access memory, and can also include a nonvolatile memory, for example, a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state memory device.

[0161] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by a computer program instructing relevant hardware. The program can be stored in a non-volatile computer readable storage medium, and when executed, can include the processes of the above-mentioned embodiment methods. Any reference to memory, storage, database or other medium used in the embodiments provided by the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0162] Any combination of the technical features of the above embodiments can be made. In order to make the description simple, all possible combinations of the technical features in the above embodiments are not described, however, as long as the combination of the technical features does not exist, it should be considered as the scope of the present application.

[0163] The above embodiments are the preferred embodiments of the present application, but the embodiments of the present application are not limited to the above embodiments, and any changes, modifications, substitutions, combinations and simplifications made without departing from the spirit and principle of the present application shall be equivalent replacement methods and shall be included in the protection scope of the present application.

Claims

1. A path planning method for an autonomous underwater vehicle, characterized in that, include: Acquire initial feature field data, and preprocess the initial feature field data to obtain high-resolution feature field data; The high-resolution feature field data is used to assign target observation areas to multiple vehicles to perform observation tasks. Based on the high-resolution feature field data, local feature field data around the current position of each vehicle is extracted, and a local Gaussian process model corresponding to the observation task performed by each vehicle is initialized based on the local feature field data. The local Gaussian process model is dynamically updated using real-time feature field data collected by each vehicle, and the local feature field prediction data around each vehicle is predicted based on the dynamically updated local Gaussian process model. Based on the local feature field prediction data of each vehicle, the improved ant colony algorithm is used to plan the flight path of each vehicle. Based on the flight path, the mission execution status of each vehicle is monitored, and the mission is reassigned or terminated based on the mission execution status of all vehicles.

2. The autonomous underwater vehicle path planning method according to claim 1, characterized in that, Acquire initial feature field data, and preprocess the initial feature field data to obtain high-resolution feature field data, including: Initial feature field data of the region to be observed is obtained, and the initial feature field data is interpolated using a preset interpolation method to obtain high-resolution feature field data, wherein the resolution of the high-resolution feature field data is higher than that of the initial feature field data.

3. The path planning method for autonomous underwater vehicles according to claim 1, characterized in that, The high-resolution feature field data is used to assign target observation areas for multiple vehicles to perform observation tasks, including: Based on high-resolution feature field data, multiple observation regions with different feature information are divided by clustering algorithm, and feature identification criterion values ​​of the observation regions are calculated. Based on a preset feature identification criterion threshold, the feature values ​​of interest are determined from the feature identification criterion values, and the regions corresponding to the clusters of the feature values ​​of interest are marked as regions of interest. The region of interest is further classified to determine multiple sub-regions of interest. These multiple sub-regions of interest are then divided into several target observation areas to be observed according to preset evaluation indicators and the number of vehicles. Based on the evaluation results of the target observation area, different numbers of vehicles are dispatched to the designated target observation area for observation.

4. The path planning method for autonomous underwater vehicles according to claim 1, characterized in that, The local Gaussian process model is dynamically updated using real-time feature field data collected by each vehicle. Based on the dynamically updated local Gaussian process model, the predicted local feature field data around each vehicle is calculated, including: Define the input feature matrix of the local Gaussian process model; Based on the real-time feature field data collected by each vehicle along its flight path in the target observation area, the target output vector is obtained; The local Gaussian process model is updated by using the input feature matrix and the target output vector and updating the hyperparameters through an optimization function. The input feature matrix is ​​fed into the updated local Gaussian process model to generate local feature field prediction data for each vehicle.

5. The autonomous underwater vehicle path planning method according to claim 1, characterized in that, Based on the local feature field prediction data of each vehicle, an improved ant colony algorithm is used to plan the flight path of each vehicle, including: Define the pheromone matrix for the ant colony algorithm and obtain the pheromone concentration in the target observation area; Based on the local feature field prediction data of the current path point of the vehicle, the feature identification criterion value of the feature value of the current path point and the neighboring path points as the spatial position changes is calculated. Based on a preset state transition function, combined with the vehicle's direction of travel from the current path point to a neighboring path point, the feature identification criterion value, and the pheromone concentration in the target observation area, the transition probability from the current path point to a neighboring path point is calculated, and the next path point is determined based on the transition probability. The ant acquires real-time feature field data from the current path point to the next path point, determines the change value of the feature field data, updates the pheromone matrix based on the change value of the feature field data, and completes one path planning for a single ant. Once all ants have completed their path planning, the optimal navigation path is determined using a pre-defined path evaluation function and used as the navigation path for each ant.

6. The path planning method for autonomous underwater vehicles according to claim 5, characterized in that, After determining the next path point based on the aforementioned transition probability, the process includes: Check if the vehicle has left the target observation area. If it has, initiate a rediscovery process to detect nearby path points using a preset search method. Detect a nearby path point that meets the preset conditions as the next path point.

7. The autonomous underwater vehicle path planning method according to claim 1, characterized in that, Based on the flight path, monitor the mission execution status of each vehicle, and reassign or terminate missions based on the mission execution status of all vehicles, including: Based on the distribution of the navigation path, the coverage of the target observation area by the navigation path of the vehicle is calculated in real time, and the mission execution status of each vehicle is determined based on the coverage. When the mission execution status of the aircraft is detected to meet the preset mission completion conditions, a global Gaussian process model is constructed based on the real-time feature field data collected by all aircraft. The global uncertainty information of the feature field is generated using the global Gaussian process model, and a new target observation area is assigned to the spacecraft that has completed the mission based on the global uncertainty information.

8. The path planning method for autonomous underwater vehicles according to claim 7, characterized in that, Also includes: When the global uncertainty information is reduced to a preset uncertainty threshold, the observation mission of all vehicles ends.

9. A path planning system for an autonomous underwater vehicle, characterized in that, For performing the method as described in any one of claims 1 to 8, comprising: The data acquisition module is used to acquire initial feature field data and preprocess the initial feature field data to obtain high-resolution feature field data. The task allocation module is used to allocate target observation areas for multiple vehicles to perform observation tasks using the high-resolution feature field data. The local model initialization module is used to extract local feature field data around the current position of each vehicle based on the high-resolution feature field data, and to initialize the local Gaussian process model corresponding to the observation task performed by each vehicle based on the local feature field data. The local feature field prediction module is used to dynamically update the local Gaussian process model using real-time feature field data collected by each vehicle, and to predict the local feature field prediction data around each vehicle based on the dynamically updated local Gaussian process model. The path planning module is used to plan the flight path of each vehicle based on the prediction data of the local feature field of each vehicle and using an improved ant colony algorithm. The mission monitoring module is used to monitor the mission execution status of each vehicle based on the navigation path, and to reassign or terminate missions based on the mission execution status of all vehicles.

10. An electronic device, characterized in that, The electronic device includes: At least one processor; and a memory communicatively connected to said at least one processor; The memory stores computer program instructions that can be executed by the at least one processor, which are executed by the at least one processor to enable the at least one processor to perform the autonomous underwater vehicle path planning method as described in any one of claims 1 to 8.

Images