Method and system for navigation of unmanned aerial vehicles based on laser signal navigator

Abstract

A method and system for navigating an unmanned aerial vehicle is provided. The present invention includes the steps of: receiving laser illumination using a laser sensor to obtain an input byte signal; removing non-key features to obtain an input sample; inputting the input sample into a TS navigation model to obtain a direction indicator; and analyzing the relative position between the laser source of the laser illumination and the unmanned aerial vehicle, and guiding the turning of the unmanned aerial vehicle. The unmanned aerial vehicle navigation method and system based on the laser signal navigator provided in the embodiment of the present invention is based on the laser signal navigator structure, which has a simple device structure, is easy to process, has low hardware cost, and has a higher real-time capability, and positively guides training through good prediction output, improves the influence of important features, and hierarchically minimizes prediction error to ensure the accuracy and reliability of the model.

Description

[Technical field]

[0001] The present application relates to the field of smart navigation, and in particular to a navigation method and system for unmanned aerial vehicles based on a laser signal navigator. [Background technology]

[0002] With the wide application of unmanned aerial vehicles, the navigation technology of unmanned aerial vehicles, i.e., after the aircraft completes the task, it is safe to return it to the designated place by specific technical means, has been increasingly attracting attention from the industry. This process involves multiple technical fields, such as navigation, control, communication and machinery. The particularity of unmanned aerial vehicle navigation is that, compared with manned aircraft, unmanned aerial vehicles often perform tasks in areas that cannot be directly operated by humans, such as forest fire prevention and high-altitude patrol inspection. Such special characteristics bring difficulties to the navigation of unmanned aerial vehicles, since traditional navigation methods may not be effectively applied to these areas.

[0003] In this case, the visual sensor navigation device is an important solution for unmanned aerial vehicle navigation, which can obtain the environmental information around the unmanned aerial vehicle by visual sensors, and use image processing and computer vision technology to analyze and recognize the image, thereby realizing accurate positioning of the unmanned aerial vehicle and accurate judgment of the attitude, providing important technical support for the navigation of unmanned aerial vehicles. However, the unmanned aerial vehicle based on the conventional visual sensor navigation device has problems such as a complicated mechanical structure of the system, high hardware cost, and insufficient real-time calculation. Summary of the Invention [Problem to be solved by the invention]

[0004] SUMMARY OF THE DISCLOSURE The present embodiments aim to provide a method and system for navigation of an unmanned aerial vehicle based on a laser signal navigator. [Means for solving the problem]

[0005] In a first aspect, an embodiment of the present application is a navigation method for an unmanned aerial vehicle based on a laser signal navigator, the laser signal navigator includes a circular circuit board and laser sensors numbered 1 to 16 sequentially distributed on a circumference of the circular circuit board with a uniform angle, the method includes: A step of converting the signals obtained by the laser sensors numbered 1 to 16 being irradiated with laser ten times while maintaining the current orientation within a specific time resolution into byte information to obtain ten input byte signals, the time resolution being determined by the current speed of the unmanned aerial vehicle; adopting Z-score method to remove non-key features in the 10 input byte signals to obtain input samples, inputting the input samples into a trained TS navigation model, and obtaining a direction indicator output from the TS navigation model; transmitting the directional indicator to a main control chip of the unmanned aerial vehicle and guiding the turning of the unmanned aerial vehicle by analyzing a relative position of a laser source of the laser illumination and the unmanned aerial vehicle; repeating the process until the unmanned aerial vehicle reaches a target location; Here, a navigation method for an unmanned aerial vehicle based on a laser signal navigator is provided, in which a training sample set for training the TS navigation model is collected by rotating a circular circuit board of the laser signal navigator, obtained by laser illumination signals received by multiple groups of sensors, training tags corresponding to the training sample set for training the TS navigation model are determined by the driving direction of the unmanned aerial vehicle when it is illuminated by laser, and the rules of each layer in the rule layer of the TS navigation model are determined by hierarchically performing a directional prediction error minimization calculation based on the sample set of the TS navigation model.

[0006] Optionally, the TS navigation model includes an input layer, a membership layer, a rule layer having a multi-layer structure, and an output layer, and the training process thereof includes: inputting the training sample set into the TS navigation model through the input layer to obtain input feature values; said membership layer mapping input feature values ​​to membership function output values ​​to obtain membership values ​​corresponding to said training sample set; The rule layer generates rules and calculates corresponding fitness values ​​according to the input feature values ​​and corresponding membership values, and outputs prediction results from the TS navigation model by an output layer; Here, when optimizing a target rule layer among the rule layers, a good quality output is determined from the predicted output of the layer immediately before the target rule layer obtained based on the training sample set and introduced into the original space of the target rule layer; if the target rule layer is the last layer of the rule layers, the target rule layer is fixed, and the training result is calculated.

[0007] Optionally, the quality output is determined based on a difference between the predicted output of the previous layer of the target rule layer and the corresponding training tags.

[0008] Optionally, a target rule layer among the rule layers includes a plurality of rules, each rule being divided into an antecedent portion and a consequent portion, the antecedent portion being calculated based on feature values ​​and membership values ​​corresponding to the training sample set, and the consequent portion being calculated from the output of the target rule layer and the predicted output of the layer immediately preceding the target rule layer.

[0009] Optionally, a training test set for training the TS navigation model is collected by rotating the circular circuit board of the laser signal navigator, the laser illumination signals being obtained by a plurality of groups of sensors received; After the steps of fixing the target rule layer and calculating the training result, the training process of the TS navigation model includes: The method further includes storing the parameters of the target rule layer if the prediction output accuracy rate corresponding to the training test set is higher than the preset value; otherwise, re-collecting the training sample set and the training test set, and training the TS navigation model.

[0010] Optionally, a training test set for training the TS navigation model is collected by rotating a circular circuit board of the laser signal navigator, and the laser illumination signals received by a plurality of groups of sensors are obtained by: rotating the circular circuit board of the laser signal navigator to sequentially point different sensors directly at the laser source for a fixed time interval; The method includes a step of stitching together multiple third sequences generated within the same time resolution, adding a directional marker of the corresponding unmanned aerial vehicle when a laser illumination signal is received to the end of each third sequence, and obtaining the training sample set and the test sample set.

[0011] Optionally, the step of removing non-key features in the 10 input byte signals by adopting a Z-score method to obtain input samples specifically includes: storing ten input byte signals in a first array; converting the ten input byte signals into ten decimal numbers and storing them in a second array; using a Z-score method to recognize and remove non-key values ​​in the second array and set their corresponding positions in the first array to binary 0; calculating an average value for each element in the updated first array and storing these average values ​​in a third array, each corresponding to one independent input sample in turn.

[0012] The signals obtained by the laser sensors numbered 1 to 16 selectively receiving laser irradiation 10 times are converted into byte information by a register provided in the laser signal navigator.

[0013] Optionally, the time resolution is determined by a current speed of the unmanned aerial vehicle, specifically The faster the current speed of the unmanned aerial vehicle, the shorter the time interval for acquiring a laser irradiation signal, which is represented by the time resolution.

[0014] In a second aspect, an embodiment of the present application is a navigation system for an unmanned aerial vehicle based on a laser signal navigator, the laser signal navigator includes a circular circuit board and laser sensors numbered 1 to 16 sequentially distributed on the circumference of the circular circuit board with a uniform angle, the system includes: A collection module for converting signals obtained by the laser sensors numbered 1 to 16 being irradiated with laser ten times while maintaining the current orientation within a specific time resolution into byte information to obtain ten input byte signals, the time resolution being determined by the current speed of the unmanned aerial vehicle; a navigation module for adopting a Z-score method to remove non-key features in the 10 input byte signals to obtain input samples, inputting the input samples into a trained TS navigation model, and obtaining a direction indicator output from the TS navigation model; a steering module for transmitting the directional indicator to a main control chip of the unmanned aerial vehicle and guiding the steering of the unmanned aerial vehicle by analyzing a relative position of a laser source of the laser illumination and the unmanned aerial vehicle; and a cycling module for repeating the process until the unmanned aerial vehicle reaches a target location. Here, a navigation system for an unmanned aerial vehicle based on a laser signal navigator is provided, in which a training sample set for training the TS navigation model is collected by rotating a circular circuit board of the laser signal navigator, obtained by laser illumination signals received by multiple groups of sensors, training tags corresponding to the training sample set for training the TS navigation model are determined by the driving direction of the unmanned aerial vehicle when it is illuminated by laser, and rules of each layer in the rule layer of the TS navigation model are determined by hierarchically performing calculations to minimize direction prediction errors based on the sample sets of the TS navigation model. Effect of the Invention

[0015] The unmanned aerial vehicle navigation method and system based on a laser signal navigator provided in an embodiment of the present invention is based on the laser signal navigator structure and converts the relative position problem of the unmanned aerial vehicle into a laser signal distribution problem, thereby making the device structure simple, easy to process, low hardware costs, and higher real-time performance compared to vision-based algorithms. The turning direction is determined by the TS navigation model, and the good prediction output is used to positively guide training, improve the influence of important features, and further minimize the prediction error hierarchically, thereby approaching the target output layer by layer and ensuring the accuracy and reliability of the model. [Brief description of the drawings]

[0016] In order to more clearly describe the technical solutions in the embodiments of the present application, the drawings used in the embodiments of the present application will be briefly introduced below.

[0017] [Figure 1] 1 is a flowchart of a navigation method for an unmanned aerial vehicle based on a laser signal navigator provided in an embodiment of the present application. [Diagram 2] FIG. 2 is a schematic diagram of the structure of a laser signal navigator provided in one embodiment of the present application. [Diagram 3]4 is a flowchart of a method for obtaining inputs for a TS navigation model provided in an embodiment of the present application; [Figure 4] 1 is a flowchart of a training process for a TS navigation model provided in an embodiment of the present application. [Diagram 5] FIG. 1 is a schematic diagram of the structure of an unmanned aerial vehicle navigation system based on a laser signal navigator provided in one embodiment of the present application. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0018] Hereinafter, the technical solutions in the embodiments of the present application will be described with reference to the drawings in the embodiments of the present application.

[0019] In the following drawings, like symbols and letters indicate like items, and therefore, once an item is defined in one drawing, it does not need to be further defined and interpreted in the following drawings. Also, in the description of this application, the terms "first", "second", etc. are merely for distinguishing the description, and cannot be understood as indicating or implying relative importance.

[0020] Fig. 1 is a flow chart of an unmanned aerial vehicle navigation method based on a laser signal navigator provided in an embodiment of the present application. The unmanned aerial vehicle navigation method needs to be implemented based on a laser signal navigator of a specific structure, and Fig. 2 shows a schematic diagram of the structure of the laser signal navigator.

[0021] The unmanned aerial vehicle in the embodiment of the present invention may be various unmanned aerial vehicles such as fixed wing, rotary wing, military, and civilian drones as generally understood, or may be underwater unmanned vehicles such as unmanned submarines. The navigator, as one of the functional modules of the unmanned aerial vehicle, may be built into the unmanned aerial vehicle device as a functional module, or may be assembled into the unmanned aerial vehicle body as a mountable module. The navigator can provide the unmanned aerial vehicle with functions such as positioning, navigation, route planning, attitude control, obstacle avoidance, and data recording and transmission, and is the key to ensuring the safety of the unmanned aerial vehicle and efficiently completing flight tasks.

[0022] In the prior art, a visual sensor navigation device is generally used as an important solution for the recovery of unmanned aerial vehicles, which obtains environmental information around the unmanned aerial vehicle through a visual sensor, and uses image processing and computer vision technology to analyze and recognize images, thereby realizing precise positioning of the unmanned aerial vehicle and precise judgment of its attitude, providing important technical support for the recovery of unmanned aerial vehicles. However, unmanned aerial vehicles based on conventional visual sensor navigation devices have problems such as a complicated mechanical structure of the system, high hardware costs, and insufficient real-time calculation.

[0023] Different from the prior art, the navigator in the embodiment of the present invention is mainly based on collecting and processing laser signals. As shown in FIG. 2, the laser signal navigator includes a circular circuit board and laser sensors numbered 1 to 16, which are sequentially distributed around the circumference of the circular circuit board at a uniform angle. The sensors are arranged along a straight line perpendicular to the circuit board with the center of the circuit board as an axis, and rotate sequentially at intervals of 22.5° to detect laser signals in each direction around the circumference. In operation, the circular circuit board is rotated around the perpendicular line to the center of the circle by a rotation mechanism, so that the sensor can be rotated to the required direction. The sensor is mounted on the circuit board by a surface mounting method. Each sensor is effectively light-shielded according to the circuit board and the housing. A sufficient cylindrical space is left as a light passing area directly above each sensor, and a hemispherical lens and a spherical crown mask are further arranged thereon. These lenses and masks are tightly coupled and are mounted in a position opposite to the receiving direction of the sensor, thereby avoiding the interference of the laser signal with the sensor on the back side of the receiving surface. In addition, the laser signal navigator further includes commonly used components such as a main control chip, resistors, peripheral circuit components, housing, etc., and the related functions are not described in the embodiments of the present invention.

[0024] Based on the above structure of the laser signal navigator, the flowchart of the navigation method for an unmanned aerial vehicle based on a laser signal navigator provided in the embodiment of the present invention shown in FIG. 1 specifically includes the following steps S110 to S140.

[0025] In S110, within a specific time resolution, the signals obtained when the laser sensors numbered 1 to 16 maintain their current orientation and receive laser irradiation 10 times are converted into byte information to become 10 input byte signals, and the time resolution is determined by the current speed of the unmanned aerial vehicle.

[0026] In an embodiment of the present invention, any laser sensors in the laser signal navigator are ordered in a clockwise direction starting from "1". Furthermore, the third sensor may be designated as the left front (1), the seventh sensor as the right front (2), the eleventh sensor as the right rear (3), and the fifteenth sensor as the left rear (4). The above method of ordering the laser sensors is merely an example, and it is not necessary to assign a specific number to each sensor.

[0027] The signal state of the laser sensor after the laser is irradiated is collected sequentially. The signal obtained by receiving the laser irradiation 10 times within a specific time resolution is used as the original input. These signals are then converted into byte information and output. Specifically, the signals obtained by the laser sensors numbered 1 to 16 after receiving the laser irradiation 10 times are converted into byte information by a register provided in the laser signal navigator.

[0028] The navigator obtains the speed parameters of the aircraft through the data bus, and the time resolution is directly bound to the speed of the aircraft. This is realized by correspondingly shortening the time interval for the navigator to obtain the original input when the aircraft speed is fast. In this way, the data continuity between successive collected signals can be guaranteed, and the difference between the actual position of the unmanned aerial vehicle between a group of collected data is avoided from being too large.

[0029] In S120, the Z-score method is adopted to remove non-key features in the 10 input byte signals to obtain input samples, and the input samples are input into the trained TS navigation model to obtain direction indicators output from the TS navigation model.

[0030] Specifically, as shown in FIG. 3, the input in step S120 may be obtained by the following steps S121 to S124.

[0031] In S121, ten input byte signals are stored in a first array; In S122, the 10 input byte signals are converted into 10 decimal numbers and stored in a second array; In S123, using a Z-score method to recognize and remove non-key values ​​in the second array, and set the corresponding positions of the non-key values ​​in the first array to binary 0; In S124, the average value of each element in the updated first array is calculated, and these average values ​​are stored in a third array corresponding to one independent input sample in sequence.

[0032] Z-score selection is a simple and fast feature selection method, the principle of which is to measure the importance of each feature in the data set based on its Z-score index. If a feature value is significantly different from the overall level and exceeds or falls below a preset threshold, it indicates that the feature is less important and should be discarded. In this way, by setting a threshold, it is common to be able to quickly select important features in data.

[0033] Specifically, we filter and cull the original input, set the Z-score threshold to α, and convert the set of bytes into a second array A2 of decimal numbers, where A2 = {a1, a2, . . . , a 10}. The formula z = (x i Calculate the Z-score value based on x −μ) / σ, where x i is the i-th element in the array, μ is the mean value of the array, and σ is the variance of the array. We define the numbers whose Z-score value is greater than the absolute value of α as non-key values, and set their corresponding positions in the first array to binary 0. Furthermore, i Extract the values ​​of each digit of the one-dimensional matrix a i to obtain matrix A1, that is,

number

[0034] After obtaining the input samples, the input samples can be input to the trained TS navigation model to obtain the direction indicators output from the TS navigation model.

[0035] The embodiment of the present invention calculates the navigation direction by training a TS navigation model. The corresponding Takagi-Sugeno (TS) fuzzy system is a commonly used system for complex nonlinear dynamic processing based on fuzzy logic modeling and control, and was proposed by Takagi and Sugeno in 1985. The distinguishing property of the TS fuzzy system is that it builds a model by partitioning the local linear relationship between the input and output of the system. In the TS fuzzy system, polynomial linear equations are used to represent the conclusion, and the linear combination of these rules achieves a good linear description of the global output of the nonlinear system. The training method of the TS navigation model will be described in detail later.

[0036] At S130, the directional indicator is transmitted to a main control chip of the unmanned aerial vehicle to guide the turning of the unmanned aerial vehicle by analyzing the relative position between a laser source of the laser illumination and the unmanned aerial vehicle.

[0037] The basic principle of the laser signal navigator is to help the unmanned aerial vehicle determine the relative positional relationship between the unmanned aerial vehicle and a specific target by emitting a laser beam. After the laser beam is emitted, each laser sensor displays a different laser signal, and in the embodiment of the present invention, the laser source is not installed integrally with the aircraft, but is installed at the navigation target position of the unmanned aerial vehicle. Therefore, the navigator reflects the laser reception characteristics received by different sensors under the irradiation of the laser source at the target position. After transmitting the characteristic information to the main control chip of the laser signal navigator, the relative position between the laser source and the unmanned aerial vehicle is calculated to determine the turning direction of the unmanned aerial vehicle and perform the next route planning. The specific calculation process can adopt the coordinate transformation algorithm and route planning algorithm in the prior art, and the embodiment of the present invention is only used to determine the key real-time turning information, and other aspects are not specifically limited.

[0038] At S140, the above process is repeated until the unmanned aerial vehicle reaches the target position.

[0039] The unmanned aerial vehicle is controlled in real time to gradually move to the target position by the above steps S110 to S130, and the final navigation operation is completed. After approaching the target position, the mode may be switched from auto-navigation mode to manual control mode to complete reaching the target position.

[0040] Another key technology of the embodiment of the present invention is the process of training the TS navigation model. The training sample set for training the TS navigation model is collected by rotating the circular circuit board of the laser signal navigator, and is obtained by the laser illumination signal received by a group of sensors. The training tag corresponding to the training sample set for training the TS navigation model is determined by the driving direction of the unmanned aerial vehicle when it is illuminated by the laser, and the rules of each layer in the rule layer of the TS navigation model are determined by hierarchically performing a direction prediction error minimization calculation based on the sample set of the TS navigation model. The training process can be combined with the structural features of the laser signal navigator of the embodiment of the present invention, and can positively guide the training by using good prediction output, improve the influence of important features, and further minimize the prediction error hierarchically to approach the target output layer by layer, thereby ensuring the accuracy and reliability of the model. The following is a detailed description.

[0041] The TS navigation model in the embodiment of the present invention includes an input layer, a membership layer, a rule layer having a multi-layer structure, and an output layer, and the training process thereof includes the following steps S410 to S430, as shown in FIG.

[0042] In S410, the training sample set is input to the TS navigation model through the input layer to obtain input feature values; At S420, the membership layer maps input feature values ​​to output values ​​of a membership function to obtain membership values ​​corresponding to the training sample set; In S430, the rule layer generates rules and calculates corresponding fitness values ​​according to the input feature values ​​and corresponding membership values, and outputs prediction results from the TS navigation model through an output layer; Here, when optimizing a target rule layer among the rule layers, a good quality output is determined from the predicted output of the layer immediately before the target rule layer obtained based on the training sample set and introduced into the original space of the target rule layer; if the target rule layer is the last layer of the rule layers, the target rule layer is fixed, and the training result is calculated.

[0043] Specifically, the training process of the TS model is described in detail below: The training process first needs to collect a sample set and a test set required for training the model.

[0044] Based on the ordering of the laser sensors, the rotation mechanism rotates the circular circuit board around the circle center perpendicular line to rotate the sensors to the required orientation, and the sensors in the 1st to 4th directions (laser sensors ordered as left front, right front, right rear, left rear) are sequentially made to face the laser source for a certain time interval. Within the same time resolution, a sample set X is constructed by stitching multiple third sequences and adding a code number of the corresponding direction when receiving the laser to the end of each third sequence as the corresponding training tag. The sample set is randomly sorted, and the sample set is divided into two parts with a ratio of 7:3, and the last columns of them are extracted to obtain the actual outputs of the training sample set and the test sample set.

number

number

number

number

[0045] Then, the training sample set X' can be used to build a TS navigation model, which is a deep TS fuzzy neural network, where it includes an input layer, a membership layer, a rule layer and an output layer. The number of layers of the rule layer is set to K, and the number of the remaining layers is all 1. In the training process, a fuzzy rule set needs to be established by the training set X'.

[0046] The training samples are transmitted from the input layer of the TS navigation model to the inside of the network, so that the input and output of the input layer do not change and are both feature values ​​X.

[0047] The membership layer maps the input features to the output values ​​of the membership functions, and also introduces nonlinear and uncertain features. In general, the fuzzification is performed using Gaussian membership functions, where the membership value μ of the jth feature of the ith sample is given by:

number

number

number

number

number

number

number

[0048] Here, R k is the number of rules in the kth rule layer,

number

number

[0049] After calculating the fitness and antecedent parameters of each rule, the fitness and antecedent parameters of the p-th fuzzy rule are

number

number

number

number

[0050] where p=1,2,…R k ,

number

number

[0051] Then, optimize and guide the input space of the kth rule layer, and introduce some of the prediction outputs of the previous layer into the original space as disturbances. When k is the maximum number of rule layers K, fix the fuzzy rule base and related parameters to calculate the training result.

number

number

[0052] Loss function for target and predicted outputs

number

number

number

number

number

[0053] where p=1,2,…R k , λ is one superparameter, I is the identity matrix,

number

number

[0054] Preferably, S7 specifically calculates an error E k We use θ to represent the error between the predicted output and the target output for each current rule layer.

number

[0055] where k = 1, 2, …, K and p = 1, 2, …, R k , y is the prediction output of the kth layer, and

number

number

number

[0056] where k = 1, 2, …, K and p = 1, 2, …, R k , E k is the error between the measured output of the current layer and the target output, and e k is the error threshold and Q is the rule layer number corresponding to good quality prediction output.

[0057] Select the good quality outputs and optimize the input space of the current layer using the following formula:

number

[0058] where k = 1, 2, …, K, p ∈ R Q , R Q represents the maximum rule layer number selected,

number

number

number

number

[0059] The unmanned aerial vehicle navigation method based on a laser signal navigator provided in the embodiment of the present invention proposes a laser signal navigator structure, and converts the relative position problem of the unmanned aerial vehicle into a laser signal distribution problem, so that the device structure is simple, easy to process, the hardware cost is low, and the real-time performance is higher than that of vision-based algorithms. The turning direction is determined by the TS navigation model, the high-quality prediction output is used to positively guide training, the influence of important features is improved, and the prediction error is minimized hierarchically, so as to approach the target output layer by layer, and the accuracy and reliability of the model are ensured.

[0060] As shown in FIG. 5, an embodiment of the present invention further provides a navigation system for an unmanned aerial vehicle based on a laser signal navigator.

[0061] The laser signal navigator includes a circular circuit board and laser sensors numbered 1 to 16, which are sequentially distributed around the circumference of the circular circuit board at a uniform angle. The system includes: A collection module 501 for converting signals obtained by the laser sensors numbered 1 to 16 being irradiated with laser ten times while maintaining the current orientation within a specific time resolution into byte information to obtain ten input byte signals, the time resolution being determined by the current speed of the unmanned aerial vehicle; a navigation module 502 for adopting a Z-score method to remove non-key features in the 10 input byte signals to obtain input samples, inputting the input samples into a trained TS navigation model, and obtaining a direction indicator output from the TS navigation model; a steering module 503 for transmitting the directional indicator to a main control chip of the unmanned aerial vehicle and guiding the steering of the unmanned aerial vehicle by analyzing a relative position between a laser source of the laser illumination and the unmanned aerial vehicle; and a cycling module 504 for repeating the above process until the unmanned aerial vehicle reaches a target location. Here, a navigation system for an unmanned aerial vehicle based on a laser signal navigator is provided, in which a training sample set for training the TS navigation model is collected by rotating a circular circuit board of the laser signal navigator, obtained by laser illumination signals received by multiple groups of sensors, training tags corresponding to the training sample set for training the TS navigation model are determined by the driving direction of the unmanned aerial vehicle when it is illuminated by laser, and rules of each layer in the rule layer of the TS navigation model are determined by hierarchically performing calculations to minimize direction prediction errors based on the sample sets of the TS navigation model.

[0062] The unmanned aerial vehicle navigation system based on the laser signal navigator provided in the embodiment of the present invention proposes a laser signal navigator structure, and converts the relative position problem of the unmanned aerial vehicle into a laser signal distribution problem, so that the device structure is simple, easy to process, the hardware cost is low, and the real-time performance is higher than that of vision-based algorithms. The turning direction is determined by the TS navigation model, the high-quality prediction output provides positive guidance for training, improves the influence of important features, and further minimizes the prediction error hierarchically, thereby approaching the target output layer by layer and ensuring the accuracy and reliability of the model.

[0063] The above-described device embodiments are merely illustrative, and the units described as separate components therein may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one place or distributed among multiple network units. Some or all of the modules may be selected according to actual needs to achieve the objectives of the solutions of the present embodiment. Those skilled in the art can understand and implement them without any creative effort.

[0064] From the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be realized by adding a required general-purpose hardware platform to software, and of course by hardware. Based on such understanding, the above technical solutions can be essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product is stored in a computer-readable storage medium such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes a plurality of instructions for making a computer device (which may be a personal computer, a server, a network device, etc.) execute the method described in each embodiment or a certain part of the embodiment.

[0065] Finally, it should be noted that the above embodiments are only for illustrating the technical solutions of the present invention, and are not intended to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art can still modify the technical solutions described in the above embodiments or replace some of the technical features with equivalents, and such modifications or replacements do not cause the essence of the corresponding technical solutions to depart from the spirit and scope of the technical solutions of the embodiments of the present invention.

[0066] (Additional Note) (Appendix 1) A navigation method for an unmanned aerial vehicle based on a laser signal navigator, the laser signal navigator includes a circular circuit board and laser sensors numbered 1 to 16 sequentially distributed on the circumference of the circular circuit board with a uniform angle, the method comprising: A step of converting the signals obtained by the laser sensors numbered 1 to 16 being irradiated with laser ten times while maintaining the current orientation within a specific time resolution into byte information to obtain ten input byte signals, the time resolution being determined by the current speed of the unmanned aerial vehicle; adopting Z-score method to remove non-key features in the 10 input byte signals to obtain input samples, inputting the input samples into a trained TS navigation model, and obtaining a direction indicator output from the TS navigation model; transmitting the directional indicator to a main control chip of the unmanned aerial vehicle and guiding the turning of the unmanned aerial vehicle by analyzing a relative position of a laser source of the laser illumination and the unmanned aerial vehicle; repeating the process until the unmanned aerial vehicle reaches a target location; Here, the navigation method for an unmanned aerial vehicle based on a laser signal navigator is characterized in that a training sample set for training the TS navigation model is collected by rotating the circular circuit board of the laser signal navigator and is obtained by laser illumination signals received by multiple groups of sensors, training tags corresponding to the training sample set for training the TS navigation model are determined by the driving direction of the unmanned aerial vehicle when it is illuminated by laser, and the rules of each layer in the rule layer of the TS navigation model are determined hierarchically by performing a directional prediction error minimization calculation based on the sample set of the TS navigation model.

[0067] (Appendix 2) The TS navigation model includes an input layer, a membership layer, a rule layer having a multi-layer structure, and an output layer, and the training process thereof includes: inputting the training sample set into the TS navigation model through the input layer to obtain input feature values; said membership layer mapping input feature values ​​to membership function output values ​​to obtain membership values ​​corresponding to said training sample set; The rule layer generates rules and calculates corresponding fitness values ​​according to the input feature values ​​and corresponding membership values, and outputs prediction results from the TS navigation model by an output layer; In this case, when optimizing a target rule layer among the rule layers, a good quality output is determined from the predicted output of the layer immediately before the target rule layer obtained based on the training sample set, and introduced into the original space of the target rule layer; if the target rule layer is the last layer of the rule layers, the target rule layer is fixed, and the training result is calculated.

[0068] (Appendix 3) 3. The method of claim 2, wherein the quality output is determined based on a difference between the predicted output of the layer immediately preceding the target rule layer and the corresponding training tag.

[0069] (Appendix 4) The method described in Appendix 2, characterized in that a target rule layer among the rule layers includes a plurality of rules, each rule being divided into an antecedent portion and a consequent portion, the antecedent portion being calculated based on feature values ​​and membership values ​​corresponding to the training sample set, and the consequent portion being calculated from the output of the target rule layer and the predicted output of the layer immediately preceding the target rule layer.

[0070] (Appendix 5) A training test set for training the TS navigation model is collected by rotating the circular circuit board of the laser signal navigator, and the laser illumination signals are obtained by receiving laser illumination signals from a plurality of groups of sensors; After the steps of fixing the target rule layer and calculating the training result, the training process of the TS navigation model includes: The method of claim 4, further comprising the steps of storing parameters of the target rule layer if the prediction output accuracy rate corresponding to the training test set is higher than a preset value, and otherwise, re-collecting a training sample set and a training test set and training the TS navigation model.

[0071] (Appendix 6) Specifically, the training test set for training the TS navigation model is obtained by rotating the circular circuit board of the laser signal navigator to collect laser illumination signals received by a plurality of groups of sensors. rotating the circular circuit board of the laser signal navigator to sequentially point different sensors directly at the laser source for a fixed time interval; The method described in Appendix 5, characterized in that it includes a step of stitching together multiple third sequences generated within the same time resolution, adding a directional marker of the corresponding unmanned aerial vehicle when a laser illumination signal is received to the end of each third sequence, and obtaining the training sample set and the test sample set.

[0072] (Appendix 7) The step of removing non-key features in the 10 input byte signals by adopting Z-score method to obtain input samples specifically includes: storing ten input byte signals in a first array; converting the ten input byte signals into ten decimal numbers and storing them in a second array; using a Z-score method to recognize and remove non-key values ​​in the second array and set their corresponding positions in the first array to binary 0; calculating an average value for each element in the updated first array and storing these average values ​​in a third array, each corresponding to one independent input sample.

[0073] (Appendix 8) The method according to claim 1, characterized in that the signals obtained by the laser sensors numbered 1 to 16 upon receiving laser irradiation 10 times are converted into byte information by a register provided in a laser signal navigator.

[0074] (Appendix 9) Specifically, the time resolution is determined by the current speed of the unmanned aerial vehicle. The method described in Appendix 1, characterized in that the faster the current speed of the unmanned aerial vehicle, the shorter the time interval for acquiring a laser irradiation signal, represented by the time resolution.

[0075] (Appendix 10) A navigation system for an unmanned aerial vehicle based on a laser signal navigator, the laser signal navigator includes a circular circuit board and laser sensors numbered 1 to 16 sequentially distributed on the circumference of the circular circuit board with a uniform angle, the system comprising: A collection module for converting signals obtained by the laser sensors numbered 1 to 16 being irradiated with laser ten times while maintaining the current orientation within a specific time resolution into byte information to obtain ten input byte signals, the time resolution being determined by the current speed of the unmanned aerial vehicle; a navigation module for adopting a Z-score method to remove non-key features in the 10 input byte signals to obtain input samples, inputting the input samples into a trained TS navigation model, and obtaining a direction indicator output from the TS navigation model; a steering module for transmitting the directional indicator to a main control chip of the unmanned aerial vehicle and guiding the steering of the unmanned aerial vehicle by analyzing a relative position of a laser source of the laser illumination and the unmanned aerial vehicle; and a cycling module for repeating the process until the unmanned aerial vehicle reaches a target location. Here, a training sample set for training the TS navigation model is collected by rotating the circular circuit board of the laser signal navigator, obtained by laser illumination signals received by multiple groups of sensors, training tags corresponding to the training sample set for training the TS navigation model are determined by the driving direction of the unmanned aerial vehicle when it is illuminated by laser, and the rules of each layer in the rule layer of the TS navigation model are determined by hierarchically performing a directional prediction error minimization calculation based on the sample set of the TS navigation model, characterized in that:

Claims

1. A navigation method for an unmanned aerial vehicle based on a laser signal navigator, the laser signal navigator includes a circular circuit board and laser sensors numbered 1 to 16 sequentially distributed on the circumference of the circular circuit board with a uniform angle, the method comprising: A step of converting the signals obtained by the laser sensors numbered 1 to 16 being irradiated with laser ten times while maintaining the current orientation within a specific time resolution into byte information to obtain ten input byte signals, the time resolution being determined by the current speed of the unmanned aerial vehicle; adopting Z-score method to remove non-key features in the 10 input byte signals to obtain input samples, inputting the input samples into a trained TS navigation model, and obtaining a direction indicator output from the TS navigation model; transmitting the directional indicator to a main control chip of the unmanned aerial vehicle and guiding the turning of the unmanned aerial vehicle by analyzing a relative position of a laser source of the laser illumination and the unmanned aerial vehicle; repeating the process until the unmanned aerial vehicle reaches a target location; Wherein, a training sample set for training the T-S navigation model is obtained by collecting laser illumination signals received by a plurality of groups of sensors by rotating the circular circuit board of the laser signal navigator, and a training tag corresponding to the training sample set for training the T-S navigation model is determined by the driving direction of the unmanned aerial vehicle when it is illuminated by laser, and the rules of each layer in the rule layer of the T-S navigation model are determined by hierarchically performing a direction prediction error minimization calculation based on the sample set of the T-S navigation model; The T-S navigation model includes an input layer, a membership layer, a rule layer having a multi-layer structure, and an output layer, and the training process thereof includes: inputting the training sample set into the TS navigation model through the input layer to obtain input feature values; the membership layer mapping input feature values ​​to output values ​​of a membership function to obtain membership values ​​corresponding to the training sample set; The rule layer generates rules and calculates corresponding fitness values ​​based on the input feature values ​​and corresponding membership values, and outputs prediction results from the T-S navigation model through an output layer; Here, when optimizing a target rule layer among the rule layers, a good quality output is determined from the predicted output of the layer immediately preceding the target rule layer obtained based on the training sample set and introduced into the original space of the target rule layer, and when the target rule layer is the last layer of the rule layers, the target rule layer is fixed and the training result is calculated. This is a navigation method for an unmanned aerial vehicle based on a laser signal navigator, characterized in that:

2. 2. The method of claim 1, wherein the quality output is determined based on a difference between a predicted output of a layer prior to the target rule layer and a corresponding training tag.

3. 2. The method of claim 1, wherein a target rule layer among the rule layers includes a plurality of rules, each rule being divided into an antecedent portion and a consequent portion, the antecedent portion being calculated based on feature values ​​and membership values ​​corresponding to the training sample set, and the consequent portion being calculated from an output of the target rule layer and a predicted output of a layer immediately preceding the target rule layer.

4. A training test set for training the TS navigation model is acquired by a plurality of groups of sensors receiving laser illumination signals collected by rotating a circular circuit board of the laser signal navigator; After the steps of fixing the target rule layer and calculating the training result, the training process of the TS navigation model includes: The method of claim 3, further comprising the steps of: storing the parameters of the target rule layer if the prediction output accuracy rate corresponding to the training test set is higher than a preset value; otherwise, recollecting a training sample set and a training test set, and training the T-S navigation model.

5. The step of employing a Z-score method to remove non-key features in the 10 input byte signals to obtain input samples specifically comprises: storing ten input byte signals in a first array; converting the ten input byte signals into ten decimal numbers and storing them in a second array; using a Z-score method to recognize and remove non-key values ​​in the second array and set their corresponding positions in the first array to binary 0; calculating an average value for each element in the updated first array and storing these average values ​​in a third array, each corresponding to one independent input sample in turn; Specifically, the training test set for training the TS navigation model is obtained by the laser illumination signals received by a plurality of groups of sensors, which are collected by rotating the circular circuit board of the laser signal navigator. rotating the circular circuit board of the laser signal navigator to sequentially point different sensors directly at the laser source for a fixed time interval; The method of claim 4, further comprising the step of stitching together a plurality of the third sequences generated within the same time resolution, and adding a directional marker of the corresponding unmanned aerial vehicle at the end of each third sequence when a laser illumination signal is received, to obtain the training sample set and the training test set.

6. The step of removing non-key features in the 10 input byte signals by adopting Z-score method to obtain input samples specifically includes: storing ten input byte signals in a first array; converting the ten input byte signals into ten decimal numbers and storing them in a second array; using a Z-score method to recognize and remove non-key values ​​in the second array and set their corresponding positions in the first array to binary 0; and calculating an average value for each element in the updated first array and storing these average values ​​in a third array, each corresponding to one independent input sample in turn.

7. 2. The method according to claim 1, wherein the signals obtained by the laser sensors numbered 1 to 16 receiving laser irradiation ten times are converted into byte information by a register provided in a laser signal navigator.

8. Specifically, the time resolution is determined by the current speed of the unmanned aerial vehicle. The method of claim 1, further comprising the step of: the faster the current speed of the unmanned aerial vehicle, the shorter the time interval for acquiring the laser illumination signal, which represents the time resolution.

9. A navigation system for an unmanned aerial vehicle based on a laser signal navigator, the laser signal navigator includes a circular circuit board and laser sensors numbered 1 to 16 sequentially distributed on the circumference of the circular circuit board with a uniform angle, the system comprising: A collection module for converting signals obtained by the laser sensors numbered 1 to 16 being irradiated with laser ten times while maintaining a current orientation within a certain time resolution into byte information to obtain ten input byte signals, the time resolution being determined by the current speed of the unmanned aerial vehicle; a navigation module for adopting a Z-score method to remove non-key features in the 10 input byte signals to obtain input samples, inputting the input samples into a trained T-S navigation model, and obtaining a direction indicator output from the T-S navigation model; a steering module for transmitting the directional indicator to a main control chip of the unmanned aerial vehicle and guiding the steering of the unmanned aerial vehicle by analyzing a relative position of a laser source of the laser illumination and the unmanned aerial vehicle; and a cycling module for repeating the process until the unmanned aerial vehicle reaches a target location. Wherein, a training sample set for training the T-S navigation model is obtained by collecting laser illumination signals received by a plurality of groups of sensors by rotating the circular circuit board of the laser signal navigator, and a training tag corresponding to the training sample set for training the T-S navigation model is determined by the driving direction of the unmanned aerial vehicle when it is illuminated by laser, and the rules of each layer in the rule layer of the T-S navigation model are determined by hierarchically performing a direction prediction error minimization calculation based on the sample set of the T-S navigation model; The T-S navigation model includes an input layer, a membership layer, a rule layer having a multi-layer structure, and an output layer, and the training process thereof includes: inputting the training sample set into the TS navigation model through the input layer to obtain input feature values; the membership layer mapping input feature values ​​to output values ​​of a membership function to obtain membership values ​​corresponding to the training sample set; The rule layer generates rules and calculates corresponding fitness values ​​based on the input feature values ​​and corresponding membership values, and outputs prediction results from the T-S navigation model through an output layer; Here, when optimizing a target rule layer among the rule layers, a good quality output is determined from the predicted output of the layer immediately preceding the target rule layer obtained based on the training sample set and introduced into the original space of the target rule layer, and if the target rule layer is the last layer of the rule layers, the target rule layer is fixed and the training result is calculated.This is a navigation system for an unmanned aerial vehicle based on a laser signal navigator, characterized in that:

Images