A method for constructing an energy representation model and an unmanned sea and air system

Abstract

The present application provides a method for constructing an energy characterization model and a sea and air unmanned system, which is applied to a sea and air unmanned system, and the sea and air unmanned system includes at least one unmanned device, and the method includes: obtaining an initial map of the target area where the target task is located and a motion model of the sea and air unmanned system; when each unmanned device performs the target task in the target area, obtaining the internal working condition data, environmental data and external working condition data of each unmanned device; updating the initial map according to the internal working condition data, environmental data, external working condition data and motion model of each unmanned device to obtain an environmental energy map; constructing a Gaussian distribution energy field according to the external working condition data; fusing the environmental energy map and the Gaussian distribution energy field to obtain an energy characterization model of the sea and air unmanned system. In this way, the energy characterization model is constructed in combination with the environmental energy map and the Gaussian distribution energy field, and the accurate quantification of the energy consumption of the sea and air unmanned system is achieved.

Description

Technical Field

[0001] The present application relates to the technical field of unmanned systems, and in particular to an energy characterization model construction method and a sea and air unmanned system. Background Art

[0002] With the rapid development of unmanned technology, the application of unmanned air and sea systems (including drones, unmanned boats, etc.) in regional observation, environmental monitoring, emergency rescue and other fields is becoming more and more extensive. In order to better perform tasks, it is very important to control the energy consumption during the execution of tasks. However, since unmanned air and sea systems use multiple power systems, and the working environment and mission requirements are complex and changeable, how to accurately quantify their energy consumption in actual applications has become an urgent problem to be solved. Summary of the invention

[0003] The embodiments of the present application provide an energy characterization model construction method and a sea and air unmanned system.

[0004] According to the first aspect of the present application, there is provided a method for constructing an energy characterization model, which is applied to a sea and air unmanned system, wherein the sea and air unmanned system includes at least one unmanned device, and the method includes: obtaining an initial map of a target area where a target task is located and a motion model of the sea and air unmanned system; when each unmanned device performs the target task in the target area, obtaining internal operating data, environmental data, and external operating data of each unmanned device; updating the initial map according to the internal operating data, environmental data, external operating data, and motion model of each unmanned device to obtain an environmental energy map; constructing a Gaussian distribution energy field according to the external operating data; and fusing the environmental energy map and the Gaussian distribution energy field to obtain an energy characterization model of the sea and air unmanned system.

[0005] According to one embodiment of the present application, the initial map is updated according to the internal working condition data, environmental data, external working condition data and motion model of each unmanned device to obtain an environmental energy map, including: extracting features from the internal working condition data, environmental data, external working condition data and motion data output by the motion model of each unmanned device based on a neural network to obtain a feature vector set; classifying the feature vector set according to the Monte Carlo algorithm and fuzzy reasoning to obtain a first feature vector set and a second feature vector set; performing particle filtering on the first feature vector set based on the Monte Carlo algorithm and the motion model to obtain a first state estimation result of each unmanned device; performing fuzzy reasoning on the second feature vector set based on a set fuzzy rule base to obtain a second state estimation result of each unmanned device; constructing a local environmental energy map of each unmanned device according to the feature vector set, the first state estimation result and the second state estimation result of each unmanned device; updating the initial map according to the local environmental energy map of each unmanned device to obtain an environmental energy map.

[0006] According to one embodiment of the present application, the particle filtering processing is performed on the first feature vector set based on the Monte Carlo algorithm and the motion model to obtain the first state estimation result of each unmanned device, including: predicting the state of particles of each unmanned device according to the motion model; calculating the observation model according to the first feature vector set; calculating particle weights according to the observation model and actual observation data; updating the long-term change mean and short-term change mean of the particles according to preset long-term change rate and short-term change rate; updating the particles according to the long-term change mean and short-term change mean; and weighted averaging the states of the updated particles to obtain the first state estimation result of each unmanned device.

[0007] According to one embodiment of the present application, fuzzy reasoning is performed on the second feature vector set based on a set fuzzy rule base to obtain a second state estimation result for each unmanned device, including: fuzzifying the second feature vector set of each unmanned device according to a set input fuzzy set and a set membership function to obtain a fuzzy vector set, wherein the fuzzy vector set includes the fuzzy value and membership of each feature vector in the second feature vector set; fuzzy reasoning is performed on the fuzzy vector set according to a set fuzzy rule base to obtain a fuzzy result; and the fuzzy result is refined according to a maximum membership method to obtain a second state estimation result for each unmanned device.

[0008] According to one embodiment of the present application, constructing a Gaussian distribution energy field based on the external operating condition data includes: determining parameters of a Gaussian model based on a set statistical method and the external operating condition data, the parameters including a mean and a standard deviation; constructing a Gaussian distribution energy field based on the mean, standard deviation and a Gaussian distribution equation.

[0009] According to one embodiment of the present application, the method also includes: using the energy characterization model to predict the energy of the set execution time period of the target task to obtain an energy prediction result; obtaining the actual energy data of the set execution time period of the target task; verifying the energy characterization model based on the actual energy data and energy prediction result of the target task; and when the energy characterization model is verified, performing path planning operations based on the set path planning algorithm and ant colony algorithm.

[0010] According to one embodiment of the present application, the method further includes: updating the motion model and the energy characterization model according to the actual energy data and the energy prediction result, the motion model including a kinematic model and a dynamic model.

[0011] According to the second aspect of the present application, a sea and air unmanned system is provided, which includes: at least one unmanned device; a perception and positioning module, connected to each unmanned device, used to obtain an initial map of a target area where a target task is located and a motion model of the sea and air unmanned system, and when each unmanned device performs the target task in the target area, obtain the internal working condition data, environmental data and external working condition data of each unmanned device; an environmental characterization module, connected to the perception and positioning module, used to update the initial map according to the internal working condition data, environmental data, external working condition data and motion model of each unmanned device to obtain an environmental energy map; a state mapping module, connected to the environmental characterization module, used to construct a Gaussian distribution energy field according to the external working condition data; the environmental energy map and the Gaussian distribution energy field are merged to obtain an energy characterization model of the sea and air unmanned system.

[0012] According to one embodiment of the present application, the perception and positioning module includes a point cloud lidar sensor, an inertial measurement unit, an obstacle avoidance binocular camera and a high-precision wheel odometer.

[0013] According to one embodiment of the present application, the environmental characterization module includes a remote control module and a sea-air cross-domain hybrid perception module, and the sea-air cross-domain hybrid perception module includes a micro radar array, an electro-optical sensor, a sonar, a satellite communication unit and a multi-channel communication interface.

[0014] The method of the embodiment of the present application obtains an initial map of the target area where the target task is located and a motion model of the sea and air unmanned system; when each unmanned device performs the target task in the target area, obtains the internal working condition data, environmental data, and external working condition data of each unmanned device; updates the initial map according to the internal working condition data, environmental data, external working condition data, and motion model of each unmanned device to obtain an environmental energy map; constructs a Gaussian distribution energy field according to the external working condition data; and fuses the environmental energy map and the Gaussian distribution energy field to obtain an energy characterization model of the sea and air unmanned system. In this way, the embodiment of the present application constructs an energy characterization model based on the combination of the environmental energy map and the Gaussian distribution energy field, and realizes the accurate quantification of the energy consumption of the execution of the sea and air unmanned system tasks based on the energy characterization model.

[0015] It should be understood that the teachings of the present application are not required to achieve all of the beneficial effects described above, but specific technical solutions can achieve specific technical effects, and other embodiments of the present application can also achieve beneficial effects not mentioned above. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] By reading the detailed description below with reference to the accompanying drawings, the above and other purposes, features and advantages of the exemplary embodiments of the present application will become readily understood. In the accompanying drawings, several embodiments of the present application are shown in an exemplary and non-limiting manner, wherein:

[0017] In the drawings, the same or corresponding reference numerals represent the same or corresponding parts.

[0018] Figure 1 A schematic diagram of the implementation process of the energy characterization model construction method provided in an embodiment of the present application is shown;

[0019] Figure 2 A schematic diagram showing a motion model of the energy characterization model construction method provided in an embodiment of the present application;

[0020] Figure 3 A schematic diagram of the implementation process of the map update operation of the energy characterization model construction method provided in the embodiment of the present application is shown;

[0021] Figure 4 A schematic diagram showing the principle of an energy characterization model of a specific application example of the energy characterization model construction method provided in an embodiment of the present application;

[0022] Figure 5 A schematic diagram of the implementation process of a specific application example of the energy characterization model construction method provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0023] In order to make the purpose, features, and advantages of the present application more obvious and easy to understand, the technical solutions in the embodiments of the present application will be clearly and completely described 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, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0024] Figure 1 A schematic diagram of the implementation flow of the energy characterization model construction method provided in an embodiment of the present application is shown.

[0025] refer to Figure 1 , the embodiment of the present application provides a method for constructing an energy characterization model, the method comprising:

[0026] Operation 101 is to obtain an initial map of a target area where a target mission is located and a motion model of a sea and air unmanned system.

[0027] Sea and air unmanned systems usually consist of at least one unmanned system. An unmanned device can be regarded as an unmanned system. Unmanned devices usually include drones or unmanned boats. When performing tasks such as regional observation, environmental monitoring, and emergency rescue, it is usually necessary to coordinate multiple drones and / or unmanned boats. During the execution of the task, it is very important to control energy consumption.

[0028] Energy consumption is usually related to geographical factors and kinematic factors of unmanned equipment during the mission. Therefore, the initial map and motion model of the target area where the sea and air unmanned system is located for this target mission are first obtained. The target area refers to the area that needs to be involved in this target mission.

[0029] In one embodiment of the present application, the initial map is obtained based on the environmental information collected by each unmanned device.

[0030] In one embodiment of the present application, the motion model of the sea and air unmanned system is a model with spatiotemporal motion constraints, which can be constructed based on the task starting point (starting point) and the task end point (target point) shown in the target task. The final presentation of the motion model can refer to Figure 2 , Figure 2 A schematic diagram of a motion model of the energy characterization model construction method provided in an embodiment of the present application is shown, wherein: Indicates that the rigid body is at position i, Indicates that the rigid body moves to position i+1, , Respectively represent the horizontal and vertical axes of the rigid body at position i, Represents the line connecting the two points before and after the displacement, The line connecting the two points before and after the displacement is The angle of the axis, Represents the arc of the rigid body motion trajectory, , Respectively represent the horizontal and vertical axes of the rigid body at position i+1, The line connecting the two points before and after the displacement is The angle of the axis, Represents the arc of the motion trajectory The corresponding center angle of the circle.

[0031] Operation 102 , when each unmanned device performs a target task in a target area, obtain internal operating data, environmental data, and external operating data of each unmanned device.

[0032] In the process of each unmanned device performing the target task in the target area, the environmental data of the location of the unmanned device and the internal and external working condition data of the unmanned device can be obtained in real time. Environmental data can usually include environmental and hydrological information such as wind field, wave and flow field. Internal working condition data usually include various motion parameters and task parameters of unmanned equipment, such as displacement, speed, acceleration, angle, angular acceleration, position, distance, direction of movement, etc. The data type of external working condition data is the same as that of internal working condition data. External working condition data usually refers to data of other external systems outside the sea and air unmanned system, such as data of surrounding or opponent systems.

[0033] Operation 103 : updating the initial map according to the internal working condition data, environmental data, external working condition data and motion model of each unmanned equipment to obtain an environmental energy map.

[0034] Internal working condition data, environmental data, external working condition data and motion data output by the motion model are all data that are strongly related to energy consumption. By fusing the internal working condition data, environmental data, external working condition data and motion data output by the motion model and introducing the fused data into the initial map, an environmental energy map including energy can be obtained.

[0035] Operation 104 : constructing a Gaussian distribution energy field according to external operating condition data.

[0036] In the process of executing the target task, the influence of the external system is also very important. For example, when chasing the opponent's system, the location of the opponent's system's unmanned equipment has a certain degree of influence on energy consumption. The closer to the opponent's system, the greater the impact on energy consumption, and the farther away, the smaller the impact. Therefore, when constructing an energy representation model, it is necessary to fully consider the impact of the external system on energy consumption, and the energy consumption impact of the external system can be represented based on its energy field.

[0037] In this embodiment of the present application, the energy field of the external system is a mixed Gaussian probability energy field constrained in time and space, that is, a Gaussian distribution energy field. The Gaussian distribution energy field can be constructed based on the location of the external system and the Gaussian distribution formula, which will not be repeated here.

[0038] Operation 105 , fusing the environmental energy map and the Gaussian distribution energy field to obtain an energy characterization model of the sea and air unmanned system.

[0039] The environmental energy map and the Gaussian distribution energy field are fused to obtain an energy characterization model. Specifically, the environmental energy map and the Gaussian distribution energy field can be converted into a unified format, and the two can be fused by energy superposition to obtain the energy characterization model.

[0040] Therefore, the embodiment of the present application constructs an energy characterization model based on the combination of the environmental energy map and the Gaussian distribution energy field, thereby achieving accurate quantification of the energy consumption of the sea and air unmanned system mission execution based on the energy characterization model.

[0041] Figure 3 A schematic diagram of the implementation flow of the map update operation of the energy characterization model construction method provided in an embodiment of the present application is shown.

[0042] refer to Figure 3 In one embodiment of the present application, the above operation 103 updates the initial map according to the internal working condition data, environmental data, external working condition data and motion model of each unmanned device to obtain an environmental energy map, including:

[0043] Operation 301, extracting features from internal working condition data, environmental data, external working condition data and motion data output by a motion model of each unmanned device based on a neural network to obtain a feature vector set;

[0044] Operation 302, classifying the feature vector set according to the Monte Carlo algorithm and fuzzy reasoning to obtain a first feature vector set and a second feature vector set;

[0045] Operation 303, performing particle filtering on the first feature vector set based on a Monte Carlo algorithm and a motion model to obtain a first state estimation result of each unmanned device;

[0046] Operation 304, performing fuzzy reasoning on the second feature vector set based on a set fuzzy rule base to obtain a second state estimation result of each unmanned device;

[0047] Operation 305, constructing a local environment energy map of each unmanned device according to the feature vector set of each unmanned device, the first state estimation result and the second state estimation result;

[0048] Operation 306 , updating the initial map according to the local environmental energy map of each unmanned device to obtain an environmental energy map.

[0049] After extracting features from internal working condition data, environmental data, external working condition data, and motion data output by the motion model based on a neural network, a feature vector set is generated. The Monte Carlo algorithm and fuzzy reasoning process data of different data types respectively. For example, the data type processed by the Monte Carlo algorithm is usually a data type with a specific value, while the data type processed by fuzzy reasoning is usually fuzzy. Therefore, the feature vector set can be divided into a first feature vector set and a second feature vector set according to the data type processed by the Monte Carlo algorithm and the fuzzy reasoning.

[0050] In one embodiment of the present application, a data type classification model can be trained based on common data of Monte Carlo algorithm and fuzzy reasoning combined with a neural network, and then the data classification model is used to distinguish the feature vector set into a first feature vector set and a second feature vector set.

[0051] After obtaining the first feature vector set and the second feature vector set, the Monte Carlo algorithm and fuzzy reasoning are used to perform state estimation on the first feature vector set and the second feature vector set, respectively, to obtain a first state estimation result and a second state estimation result. Among them, using the Monte Carlo algorithm to perform state estimation on the first feature vector set can be regarded as using the Monte Carlo algorithm and the motion model to perform particle filtering on the first feature vector set, and using fuzzy reasoning to perform state estimation on the second feature vector set can be regarded as using set fuzzy rules to perform fuzzy reasoning on the second feature vector set.

[0052] After obtaining the first state estimation result and the second state estimation result of each unmanned device, the feature vector set, the first state estimation result and the second state result are used to construct a local environmental energy map of each unmanned device according to the location, and then the local environmental energy map of each unmanned device is placed into the initial map to update the initial map to obtain the final environmental energy map.

[0053] In one embodiment of the present application, the above operation 303, performing particle filtering on the first feature vector set based on the Monte Carlo algorithm and the motion model to obtain a first state estimation result of each unmanned device, includes:

[0054] Predict the state of each unmanned device particle based on the motion model;

[0055] calculating an observation model based on the first set of eigenvectors;

[0056] Calculate particle weights based on the observation model and actual observation data;

[0057] Update the particle's long-term change mean and short-term change mean according to the preset long-term change rate and short-term change rate;

[0058] Update particles according to the long-term change mean and the short-term change mean;

[0059] The updated particle states are weighted averaged to obtain the first state estimation result of each unmanned device.

[0060] By default, there is a known motion model for the sea and air unmanned system. Based on the motion equations in the motion model, the state of each particle of each unmanned device in the sea and air unmanned system can be predicted. Then, based on the information in the first eigenvector set, an observation model for describing the state of the unmanned device is constructed, and the prediction results of the observation model are compared with the actual observation data. A weight can be assigned to each particle based on the comparison result to obtain the particle weight of each particle, where the higher the particle weight, the more consistent the particle is with the actual observation data.

[0061] In order to ensure the accuracy of state estimation, the selection of particles is very important. Therefore, long-term and short-term change rates are preset to update the long-term and short-term change means of particles, so as to better identify particle distribution and remove particles with large errors. The update of long-term and short-term mean values ​​of particles can be carried out using the following formula:

[0062]

[0063] in, is the long-term mean change, is the short-term change mean, is the preset long-term rate of change, is the preset short-term rate of change, is the target value.

[0064] After obtaining the long-term change mean and the short-term change mean, the probability of updating particles can be calculated, and the particles can be updated according to the probability. Updating particles can include adding random particles and retaining important particles. The probability of updating particles can be calculated by the following formula:

[0065]

[0066] in, Represents probability.

[0067] exist When the probability threshold is exceeded, the particle is updated. If the probability threshold is not exceeded, the particle update is not performed. The probability threshold can be set according to the actual needs of the sea and air unmanned system, and this application does not make specific limitations.

[0068] After the particles are updated, the states of all particles are weighted averaged according to the particle weight of each particle to obtain the first state estimation result of each unmanned device.

[0069] In one embodiment of the present application, the above operation 304, performing fuzzy reasoning on the second feature vector set based on the set fuzzy rule base to obtain the second state estimation result of each unmanned device, includes:

[0070] Fuzzifying the second feature vector set of each unmanned device according to the set input fuzzy set and the set membership function to obtain a fuzzy vector set, wherein the fuzzy vector set includes the fuzzy value and the membership degree of each feature vector in the second feature vector set;

[0071] Perform fuzzy reasoning on the fuzzy vector set according to the set fuzzy rule base to obtain the fuzzy result;

[0072] The fuzzy results are refined according to the maximum membership method to obtain the second state estimation result of each unmanned device.

[0073] Based on the actual needs of the sea and air unmanned system, the input variables, output variables and input fuzzy sets of the input variables are defined in advance, and the membership function for the sea and air unmanned system is designed based on the Gaussian membership function and the operating environment of the sea and air unmanned system. The operating environment can be analyzed based on the initial map. In addition, in order to support the fuzzy reasoning ability of the sea and air unmanned system, fuzzy rules are extracted based on the professional knowledge and experience in the field of sea and air unmanned systems, and fuzzy rules are automatically generated using open source data and machine learning methods and expressed in the form of If-Then, where multiple fuzzy rules are combined according to the relationship between the input variables and the output variables, and multiple fuzzy rules constitute a set fuzzy rule base.

[0074] In this embodiment of the present application, the input variables are multiple feature vectors in the second feature vector set, and the output variable is the state of the sea and air unmanned system. When the second feature vector set is obtained, the second feature vector is fuzzified according to the set input fuzzy set to obtain a fuzzy feature set, which includes the fuzzy value and membership degree of each feature vector in the second feature vector set. Afterwards, based on the set fuzzy rule base, the fuzzy inference Mamdani method is used to perform fuzzy inference on the fuzzy vector set, and the fuzzy vector set is matched with the fuzzy rules in the set fuzzy rule base to obtain a fuzzy result, that is, a fuzzy result of the state. Finally, the maximum membership method is used to refine the fuzzy result to obtain the second state estimation result of each unmanned device.

[0075] In one embodiment of the present application, with the continuous application of sea and air unmanned systems, the fuzzy rules in the set fuzzy rule base are optimized by applying data using reinforcement learning, and the membership function and the reasoning parameters of the fuzzy reasoning method are optimized using optimization algorithms such as genetic algorithms and particle swarm algorithms.

[0076] In one embodiment of the present application, the operation 104, constructing a Gaussian distribution energy field according to external working condition data, includes:

[0077] Determine the parameters of the Gaussian model based on the set statistical method and external working condition data, the parameters include mean and standard deviation;

[0078] According to the mean, standard deviation and Gaussian distribution equation, a Gaussian distribution energy field is constructed.

[0079] The Gaussian distribution energy field is mainly the Gaussian energy field of the external system of the sea and air unmanned system. According to the external working condition data, the Gaussian model parameters of the external system can be analyzed based on the set statistical method, and then the Gaussian distribution energy field is constructed based on the parameters of the Gaussian model and the Gaussian distribution equation. Among them, the parameters of the Gaussian model include the mean and standard deviation, and the Gaussian distribution energy field can be constructed by the following formula:

[0080] Among them, P G is a Gaussian distributed energy field, Represents the number of distribution areas in the Gaussian distribution energy field, is the area in the Gaussian distribution energy field, and are the lengths of the region on the y-axis and x-axis respectively, x and y are the coordinates of the x-axis and y-axis respectively, is the standard deviation, and are the x-axis coordinate and y-axis coordinate of the reference point of the Gaussian distribution energy field. The reference point may be the location of the external system.

[0081] In one embodiment of the present application, path planning is also performed, and the path planning mainly includes the following operations:

[0082] The energy characterization model is used to predict the energy of the set execution period of the target task to obtain the energy prediction result;

[0083] Obtaining actual energy data for a set execution period of a target task;

[0084] The energy characterization model is verified based on the actual energy data and energy prediction results of the target task;

[0085] When the energy characterization model is verified, the path planning algorithm and the ant colony algorithm are set based on the path planning algorithm to perform path planning operations.

[0086] Each unmanned device will also compare the actual energy data of the set execution period with the energy prediction results obtained by the energy characterization model in the set execution period to verify the energy characterization model. If the verification is passed, it means that the energy characterization model is accurately constructed, and then the path planning operation is performed according to the path planning algorithm and the ant colony algorithm. Among them, the set execution period can be the entire period for executing the target task, or a period configured according to the actual needs of the task.

[0087] In one embodiment of the present application, when the error between the actual energy data and the energy prediction result is less than the set error, it is determined that the verification of the energy characterization model has passed. The set error can be configured according to the actual situation, for example, 1%, and this application does not make any specific limitation.

[0088] In one embodiment of the present application, the path planning operation mainly includes path planning and obstacle avoidance, and the path planning algorithm is an improved heuristic algorithm. In the process of path planning, the improved The algorithm performs global path planning and uses the ant colony algorithm to perform local obstacle avoidance adjustments in local small-scale sea area operations to avoid dynamic obstacles in real time.

[0089] In one embodiment of the present application, the motion model and the energy characterization model are also updated based on the actual energy data and the energy prediction results, and the motion model includes a kinematic model and a dynamic model.

[0090] Based on the above energy characterization model construction method, the embodiment of the present application further provides a sea and air unmanned system for executing the above energy characterization model construction method, the system comprising:

[0091] At least one unmanned device;

[0092] A perception and positioning module is connected to each unmanned device and is used to obtain an initial map of the target area where the target task is located and a motion model of the sea and air unmanned system, and to obtain internal working condition data, environmental data, and external working condition data of each unmanned device when each unmanned device performs the target task in the target area;

[0093] The environment characterization module is connected to the perception and positioning module and is used to update the initial map according to the internal working condition data, environmental data, external working condition data and motion model of each unmanned device to obtain an environmental energy map;

[0094] The state mapping module is connected to the environment characterization module and is used to construct a Gaussian distribution energy field according to external working condition data; the environment energy map and the Gaussian distribution energy field are integrated to obtain an energy characterization model of the sea and air unmanned system.

[0095] In one embodiment of the present application, the perception and positioning module includes a point cloud lidar sensor, an inertial measurement unit, an obstacle avoidance binocular camera and a high-precision wheel odometer.

[0096] The sea and air unmanned system point cloud lidar sensor, inertial measurement unit IMU (Inertial Measurement Unit), obstacle avoidance binocular camera and high-precision wheel odometer constitute the perception and positioning module. Among them, the point cloud lidar sensor and obstacle avoidance binocular camera are used to accurately measure the undulation information of the sea surface, detect and track static and dynamic obstacles, and are used to obtain environmental data and make initial maps; the inertial measurement unit IMU and high-precision wheel odometer are used to measure the acceleration and angular velocity of objects through accelerometers and gyroscopes, and provide high-precision internal and external working condition data of the sea and air unmanned system, mainly including position, attitude, motion parameters and mission parameters.

[0097] In one embodiment of the present application, a perception and positioning module is mounted on each unmanned device, wherein the laser radar of the point cloud laser radar sensor can be a single-line laser radar, or a 16-line / 32-line or higher-thread laser radar, the inertial measurement unit IMU can be a nine-axis sensor, the obstacle avoidance binocular camera can be a Zed binocular camera, and the high-precision wheel odometer can be a wheel odometer.

[0098] In one embodiment of the present application, the environmental characterization module includes a remote control module and a sea-air cross-domain hybrid perception module, and the sea-air cross-domain hybrid perception module includes a micro radar array, an electro-optical sensor, a sonar, a satellite communication unit and a multi-channel communication interface.

[0099] The remote control module is used for remote control and communication. Each unmanned device is equipped with a sea-air cross-domain hybrid perception module, and collects environmental data through multiple channels through the multi-channel communication interface of the sea-air cross-domain hybrid perception module.

[0100] In one embodiment of the present application, the sea and air unmanned system also includes a distributed dynamic topology network for communication, which realizes cross-domain communication between the ocean and the land based on a self-organizing network, and integrates the fifth generation mobile communication (5G) technology and satellite communication technology to realize high-speed data transmission and low-latency communication across the sea and air.

[0101] In one embodiment of the present application, the sea and air unmanned system also includes a computing system for realizing various calculations in the process of building the energy model, and is connected to each module and each unmanned device of the sea and air unmanned system. The computing system mainly includes an edge flight control system and a motion control system, an edge flight platform, a cloud server and a terminal.

[0102] It should be noted that the description of the unmanned sea and air system in the embodiment of the present application is similar to the description of the above method embodiment, and has similar beneficial effects as the method embodiment, so it will not be repeated. Figures 1 to 3 The present invention can be understood by referring to the description of any one of the accompanying drawings.

[0103] To further illustrate the technical solution of the present application, a specific application example is provided below.

[0104] Figure 4 A schematic diagram showing the principles of an energy characterization model of a specific application example of the energy characterization model construction method provided in an embodiment of the present application is shown.

[0105] refer to Figure 4 The energy characterization model of this specific application example of the embodiment of the present application is mainly used in sea and air unmanned systems including drones and unmanned boats. The energy characterization model supports kinematic modeling and dynamic modeling. The modeling mainly relies on the external environment (environmental data), the internal working conditions of drones and unmanned boats (internal working condition data), and the external working conditions of other unmanned equipment outside the sea and air unmanned system (external working condition data). The modeling operation includes initial construction and modeling update. The energy characterization model also presets a multimodal model based on the historical data of the sea and air unmanned system. The multimodal model provides the fuzzy reasoning capability of the energy characterization model, which mainly includes three parts: fuzzy mathematical representation, fuzzy knowledge base, and fuzzy logic reasoning.

[0106] When the sea and air unmanned systems are continuously performing tasks, the energy characterization model estimates the parameters of the UAV and the unmanned boat based on the environmental data, external working condition data, and internal working condition data. The parameters are state parameters. In the parameter estimation process, the data in the parameter estimation process is used to update the knowledge of the multimodal model, and the parameter error is determined through sensitivity analysis after the parameters are determined. When the error does not exceed the set error, the state parameter verification is determined to be passed and the state estimation is completed. In order to ensure the accuracy of the energy characterization model, the kinematic model and dynamic model obtained by kinematic modeling and dynamic modeling are continuously corrected according to the error of parameter estimation.

[0107] During the kinematic modeling and dynamic modeling process, multiple adaptive kinematic models and dynamic models can be constructed according to the various states of the UAV and unmanned boat. After parameter estimation, the corresponding kinematic model and dynamic model are switched in real time according to the determined state parameters combined with the influence of random fluctuations so that the energy characterization model can adapt to the state estimation of the sea and air unmanned systems under different states.

[0108] After the state estimation is performed, the state estimation result is quantified into energy, that is, the energy conversion is completed.

[0109] Figure 5A schematic diagram of the implementation process of a specific application example of the energy characterization model construction method provided in an embodiment of the present application is shown.

[0110] refer to Figure 5 , this specific application example of the present application, a method for constructing an energy characterization model is also called a method for constructing an energy characterization model of a sea-air cross-domain collaborative UAV and unmanned boat, the sea-air unmanned system is a sea-air cross-domain platform, the sea-air cross-domain platform includes a perception and positioning module 1, an environmental characterization module 2, a state mapping module 3 and environmental information and hydrological information 4, the perception and positioning module 1 includes a point cloud lidar sensor 1-1, an inertial measurement unit 1-2, an obstacle avoidance binocular camera 1-3 and a high-precision wheel odometer 1-4, the environmental characterization module 2 includes a remote control module 2-1 and a sea-air cross-domain hybrid perception module 2-2, the environmental information and hydrological information 4 includes a wind field 4-1, a wave 4-2 and a flow field 4-3, this specific application example of the present application specifically includes:

[0111] S1, initialize feature map information, update the motion model of the sea and air unmanned system, and build a dynamic window.

[0112] After the sea and air unmanned system starts to perform the mission, the feature map information used to construct the initial map is initialized first, and the motion model of the sea and air unmanned system is updated to construct a dynamic window.

[0113] S2, global map vector construction.

[0114] In order to perform this task in the target area, it is necessary to construct a global map vector, that is, the initial map.

[0115] Based on the Monte Carlo algorithm of neural network, the perception and positioning module 1 of each unmanned equipment collects environmental data, internal working condition data and external working condition data in combination with environmental information and hydrological information 4, and establishes an environmental model through fuzzy reasoning in combination with the state mapping module 3. The environmental model is the initial map.

[0116] In the process of constructing the environment model, the state mapping module 3 also improves the fuzzy control rule base 3-2 according to the environment model.

[0117] S3, realize cross-domain collaborative perception of sea and air.

[0118] The remote control module 2-1 remotely controls the sea and air unmanned systems to achieve collaborative operations.

[0119] S4, build an energy representation model based on environment representation and state mapping.

[0120] The sea and air cross-domain hybrid perception module 2-2 of each unmanned device collects environmental data in the ocean and air environment through an adaptive Monte Carlo algorithm, performs state estimation based on fuzzy reasoning of the state mapping module 3, and quantifies the state estimation into energy to construct a local dynamic map of each unmanned device, and fuses the local dynamic maps to obtain an energy characterization model.

[0121] S5, indicator update optimal feedback system.

[0122] During the mission execution of the sea and air unmanned system, the energy characterization model is continuously constructed through the above method, and the model parameter indicators of the energy characterization model are fed back to the system for optimization.

[0123] After re-estimating the state through the state mapping module 3, the established environment model 3-1 is evaluated using reinforcement learning and bionic optimization 3-3 and the local dynamic map to optimize the parameters of the environment model.

[0124] S6, energy consumption estimation, if not achieved, repeat S4, if achieved, proceed to S7.

[0125] During the mission execution of the sea and air unmanned system, the energy characterization model is used to estimate the energy consumption in real time, and the estimated energy consumption is compared with the actual energy consumption. If the error between the estimated energy consumption and the actual energy consumption meets the error requirement, S7 is performed, and if the error requirement is not met, S4 is performed. Among them, the error meeting the error requirement can be regarded as the error being less than the set error.

[0126] S7, output energy characterization parameters and execute path planning.

[0127] Output the model parameters of the current energy characterization model, namely the energy characterization parameters, and perform path planning.

[0128] It should be noted that this specific application example of the present application is similar to the description of the above method embodiment and has similar beneficial effects as the method embodiment, so it will not be described in detail. Figures 1 to 3 The present invention can be understood by referring to the description of any one of the accompanying drawings.

[0129] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this application can be executed in parallel, sequentially or in different orders, as long as the expected results of the technical solution disclosed in this application can be achieved, and this document is not limited here.

[0130] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for constructing an energy characterization model, characterized in that: Applied to a sea and air unmanned system, the sea and air unmanned system includes at least one unmanned device, and the method includes: Obtaining an initial map of the target area where the target mission is located and a motion model of the sea and air unmanned system; When each unmanned device performs the target task in the target area, obtaining internal working condition data, environmental data and external working condition data of each unmanned device; The initial map is updated according to the internal working condition data, environmental data, external working condition data and motion model of each unmanned device to obtain an environmental energy map; Constructing a Gaussian distribution energy field according to the external working condition data; The environmental energy map and the Gaussian distribution energy field are merged to obtain an energy characterization model of the sea and air unmanned system; The updating of the initial map according to the internal working condition data, environmental data, external working condition data and motion model of each unmanned equipment to obtain the environmental energy map includes: Based on the neural network, feature extraction is performed on the internal working condition data, environmental data, external working condition data of each unmanned device and the motion data output by the motion model to obtain a feature vector set of each unmanned device; Classifying the feature vector set according to the Monte Carlo algorithm and fuzzy reasoning to obtain a first feature vector set and a second feature vector set; Performing particle filtering on the first feature vector set based on the Monte Carlo algorithm and the motion model to obtain a first state estimation result of each unmanned device; Performing fuzzy reasoning on the second feature vector set based on a set fuzzy rule base to obtain a second state estimation result of each unmanned device; Constructing a local environment energy map of each unmanned device according to the feature vector set of each unmanned device, the first state estimation result, and the second state estimation result; The initial map is updated according to the local environmental energy map of each unmanned device to obtain an environmental energy map.

2. The method according to claim 1, characterized in that: The performing particle filtering on the first feature vector set based on the Monte Carlo algorithm and the motion model to obtain a first state estimation result of each unmanned device includes: predicting the state of each particle of the unmanned device according to the motion model; Calculating an observation model based on the first feature vector set; Calculate particle weights according to the observation model and actual observation data; Update the particle's long-term change mean and short-term change mean according to the preset long-term change rate and short-term change rate; Update particles according to the long-term change mean and the short-term change mean; The updated particle states are weighted averaged to obtain the first state estimation result of each unmanned device.

3. The method according to claim 1, characterized in that Performing fuzzy reasoning on the second feature vector set based on a set fuzzy rule base to obtain a second state estimation result of each unmanned device includes: Fuzzifying the second feature vector set of each unmanned device according to the set input fuzzy set and the set membership function to obtain a fuzzy vector set, wherein the fuzzy vector set includes the fuzzy value and the membership degree of each feature vector in the second feature vector set; Performing fuzzy reasoning on the fuzzy vector set according to a set fuzzy rule base to obtain a fuzzy result; The fuzzy result is refined according to the maximum membership method to obtain the second state estimation result of each unmanned device.

4. The method according to claim 1, characterized in that The step of constructing a Gaussian distribution energy field according to the external working condition data includes: Determining parameters of the Gaussian model based on a set statistical method and the external operating condition data, the parameters including a mean and a standard deviation; A Gaussian distribution energy field is constructed according to the mean, standard deviation and Gaussian distribution equation.

5. The method according to claim 1, characterized in that The method further comprises: The energy characterization model is used to predict the energy of the set execution period of the target task to obtain an energy prediction result; Acquiring actual energy data of a set execution period of the target task; Verifying the energy characterization model according to actual energy data and energy prediction results of the target task; When the energy characterization model is verified to be successful, a path planning operation is performed based on a set path planning algorithm and an ant colony algorithm.

6. The method according to claim 5, characterized in that The method further comprises: The motion model and the energy representation model are updated according to the actual energy data and the energy prediction result, and the motion model includes a kinematic model and a dynamic model.

7. A sea and air unmanned system, characterized in that: The sea and air unmanned system includes: At least one unmanned device; A perception and positioning module connected to each unmanned device, used to obtain an initial map of a target area where a target task is located and a motion model of the sea and air unmanned system, and to obtain internal working condition data, environmental data, and external working condition data of each unmanned device when each unmanned device performs the target task in the target area; The environment characterization module is connected to the perception and positioning module, and is used to update the initial map according to the internal working condition data, environmental data, external working condition data and motion model of each unmanned device to obtain an environmental energy map; wherein, the updating of the initial map according to the internal working condition data, environmental data, external working condition data and motion model of each unmanned device to obtain an environmental energy map includes: extracting features from the internal working condition data, environmental data, external working condition data and motion data output by the motion model of each unmanned device based on a neural network to obtain a feature vector set of each unmanned device; classifying the feature vector set according to the Monte Carlo algorithm and fuzzy reasoning to obtain a first feature vector set and a second feature vector set; performing particle filtering processing on the first feature vector set based on the Monte Carlo algorithm and the motion model to obtain a first state estimation result of each unmanned device; performing fuzzy reasoning on the second feature vector set based on a set fuzzy rule base to obtain a second state estimation result of each unmanned device; constructing a local environmental energy map of each unmanned device according to the feature vector set, the first state estimation result and the second state estimation result of each unmanned device; updating the initial map according to the local environmental energy map of each unmanned device to obtain an environmental energy map; The state mapping module is connected to the environment characterization module and is used to construct a Gaussian distribution energy field according to the external working condition data; the environment energy map and the Gaussian distribution energy field are merged to obtain an energy characterization model of the sea and air unmanned system.

8. The system according to claim 7, characterized in that The perception and positioning module includes a point cloud lidar sensor, an inertial measurement unit, an obstacle avoidance binocular camera and a high-precision wheel odometer.

9. The system according to claim 7, characterized in that The environmental characterization module includes a remote control module and a sea-air cross-domain hybrid perception module, and the sea-air cross-domain hybrid perception module includes a micro radar array, an electro-optical sensor, a sonar, a satellite communication unit and a multi-channel communication interface.

Images