A method for identifying working conditions of an amphibious vehicle

Abstract

The invention relates to a working condition identification method of an amphibious vehicle, belongs to the field of vehicle perception, and solves the problem that the existing amphibious vehicle cannot make a flexible transition when switching between water and land. The working condition identification method of the amphibious vehicle includes: determining a first distance between the amphibious vehicle and the shoreline; if the first distance satisfies a first threshold range, determining the weight of the amphibious vehicle carried by the water body; if the initial working condition of the amphibious vehicle is a land working condition and the weight is greater than a weight threshold, or if the initial working condition of the amphibious vehicle is a water working condition and the absolute value of the difference between the weight and the weight of the amphibious vehicle is greater than the weight threshold, using a first machine learning model to map the suspension attitude angle data of the amphibious vehicle to an intermediate working condition of the amphibious vehicle; determining a land dynamic model matching rate and a water dynamic model matching rate of the amphibious vehicle; and determining the final working condition of the amphibious vehicle based on the intermediate working condition, the land dynamic model matching rate and the water dynamic model matching rate.

Description

Technical Field

[0001] The present invention relates to the field of vehicle perception, and in particular to a method for identifying the working condition of an amphibious vehicle. Background Art

[0002] As a cutting-edge key technology in the current development of the vehicle industry, the application scope of unmanned driving is no longer limited to traditional pure land transportation tools, but has gradually expanded to the field of general-purpose land and water platforms. Amphibious unmanned vehicles are intelligent special vehicles that have the ability to travel on land, navigate on water, and operate in the land-water interface. Compared with traditional unmanned vehicles or unmanned ships, it can not only be applied to a single land or water condition, but also has the application needs of land-water transition conditions. That is, during the speed change of amphibious vehicles, the land-water conditions cannot be directly switched to each other, and an intermediate land-water condition is required for transition, that is, the process of landing or launching. Therefore, high-precision, low-latency cross-domain stage identification is of great significance to improving the stability of the cross-domain process of amphibious unmanned vehicles and realizing flexible transition of land-water switching.

[0003] Considering the complexity of cross-domain working conditions, a single sensor cannot identify each stage of cross-domain working conditions in fine granularity, so the judgment of water-land transition needs to combine multiple information. How to integrate multi-source sensor information such as lidar, images, sonar, and prior information such as electronic charts and perform relevant working condition identification, so as to effectively improve the accuracy and achieve the purpose of controlling the relevant actuators of the vehicle platform to perform the corresponding actions at the corresponding time, is a key problem that needs to be solved urgently for current amphibious unmanned vehicles. Summary of the invention

[0004] In view of the above analysis, an embodiment of the present invention aims to provide a method for identifying the working condition of an amphibious vehicle, so as to solve the problem that the existing amphibious vehicles cannot flexibly transition when switching between water and land.

[0005] On the one hand, an embodiment of the present invention provides a method for identifying an operating condition of an amphibious vehicle, comprising:

[0006] determining a first distance between the amphibious vehicle and a shoreline;

[0007] If the first distance satisfies a first threshold range, determining the weight of the amphibious vehicle carried by the water body;

[0008] If the initial operating condition of the amphibious vehicle is a land operating condition and the weight is greater than a weight threshold, or if the initial operating condition of the amphibious vehicle is a water operating condition and the absolute value of the difference between the weight and the weight of the amphibious vehicle is greater than the weight threshold, using a first machine learning model, mapping the suspension attitude angle data of the amphibious vehicle to an intermediate operating condition of the amphibious vehicle;

[0009] Determining a land dynamics model matching rate and a water dynamics model matching rate of the amphibious vehicle; and

[0010] The final operating condition of the amphibious vehicle is determined according to the intermediate operating condition, the land dynamics model matching rate and the water dynamics model matching rate.

[0011] Based on the further improvement of the above method, the training process of the first machine learning model includes:

[0012] Acquire training samples from a training sample set for the first machine learning model, wherein the training samples include suspension attitude angle data of an amphibious vehicle and operating condition labels of the amphibious vehicle, and the operating condition labels include: on-land operating condition, land-to-water operating condition, water-to-land operating condition, and on-water operating condition;

[0013] generating an operating condition of the amphibious vehicle using the first machine learning model; and

[0014] Based on the comparison between the operating condition and the operating condition label, the first machine learning model is updated.

[0015] Based on a further improvement of the above method, if the initial working condition of the amphibious vehicle is a land working condition, if the first distance satisfies a first threshold range, the method further includes:

[0016] Calculating a second distance between the amphibious vehicle and the shoreline based on the camera data and the laser point cloud data of the amphibious vehicle;

[0017] and,

[0018] If the weight is greater than a weight threshold and if the second distance satisfies a second threshold range, mapping the suspension attitude angle data of the amphibious vehicle to an intermediate operating condition of the amphibious vehicle using a first machine learning model;

[0019] The maximum value of the second threshold range is smaller than the maximum value of the first threshold range.

[0020] Based on the further improvement of the above method, if the initial working condition of the amphibious vehicle is the working condition on water,

[0021] If the first distance satisfies a first threshold range, the method further includes calculating the height of the amphibious vehicle from the ground based on the sonar data of the amphibious vehicle;

[0022] Furthermore, if the absolute value of the difference between the weight and the weight of the amphibious vehicle is greater than a weight threshold and the height above the ground is less than or equal to a height threshold, the suspension attitude angle data of the amphibious vehicle is mapped to an intermediate working condition of the amphibious vehicle using a first machine learning model.

[0023] Based on the further improvement of the above method, the determining the final working condition of the amphibious vehicle according to the intermediate working condition, the land dynamic model matching rate and the water dynamic model matching rate includes:

[0024] If the intermediate operating condition is a land operating condition and the land dynamics model matching rate satisfies a third threshold range and the water dynamics model matching rate satisfies a fourth threshold range, determining a travel distance of the amphibious vehicle after entering the water-to-land operating condition;

[0025] If the difference between the driving distance and the driving distance threshold satisfies a fifth threshold range, it is determined that the final operating condition of the amphibious vehicle is a land operating condition.

[0026] Based on the further improvement of the above method, the determining the final working condition of the amphibious vehicle according to the intermediate working condition, the land dynamic model matching rate and the water dynamic model matching rate includes:

[0027] If the intermediate operating condition is a land-to-water operating condition and both the land dynamics model matching rate and the water dynamics model matching rate satisfy a sixth threshold range, determining a travel distance of the amphibious vehicle after entering the land-to-water operating condition;

[0028] If the driving distance is greater than a driving distance threshold, it is determined that the final operating condition of the amphibious vehicle is an on-water operating condition.

[0029] Based on the further improvement of the above method, the determining the final working condition of the amphibious vehicle according to the intermediate working condition, the land dynamic model matching rate and the water dynamic model matching rate includes:

[0030] If the intermediate working condition is a water-to-land working condition and both the on-land dynamic model matching rate and the on-water dynamic model matching rate satisfy a sixth threshold range, determining a travel distance of the amphibious vehicle after entering the water-to-land working condition;

[0031] If the driving distance is greater than a driving distance threshold, it is determined that the final operating condition of the amphibious vehicle is a land operating condition.

[0032] Based on the further improvement of the above method, the determining the final working condition of the amphibious vehicle according to the intermediate working condition, the land dynamic model matching rate and the water dynamic model matching rate includes:

[0033] If the intermediate working condition is an on-water working condition and the on-land dynamic model matching rate satisfies a fourth threshold range and the on-water dynamic model matching rate satisfies a third threshold range, determining a travel distance of the amphibious vehicle after entering the land-to-water working condition;

[0034] If the difference between the driving distance and the driving distance threshold satisfies a fifth threshold range, it is determined that the final operating condition of the amphibious vehicle is an on-water operating condition.

[0035] Based on the further improvement of the above method, the training process of the first machine learning model includes:

[0036] Selecting feature parameters related to the working condition label for training from the suspension attitude angle data through feature engineering;

[0037] According to the characteristic parameters and the operating condition labels, the kernel function parameters and the penalty parameters of the support vector machine are optimized using a particle swarm algorithm; and

[0038] If the operating condition identification accuracy of the support vector machine reaches a preset threshold, the first machine learning model is obtained.

[0039] Based on a further improvement of the above method, the determining the weight of the amphibious vehicle carried by the water body comprises:

[0040] The weight of the amphibious vehicle carried by the water body is determined according to the pressure sensor data of the amphibious vehicle.

[0041] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects:

[0042] 1. The solution of the present invention integrates multi-source sensor information such as lidar, images, sonar, and prior information such as electronic charts, thereby effectively improving the accuracy of working condition identification.

[0043] 2. The solution of the present invention integrates active detection and passive contact, solves the linear measurement of the vehicle's state in the water-land transition state, guides the continuous conversion control of the vehicle in the water-land transition state, and improves the adaptability of the universal water and land unmanned vehicle under different cross-domain working conditions.

[0044] In the present invention, the above-mentioned technical solutions can also be combined with each other to achieve more preferred combination solutions. Other features and advantages of the present invention will be described in the subsequent description, and some advantages can become obvious from the description, or can be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained through the contents particularly pointed out in the description and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The accompanying drawings are only used for the purpose of illustrating specific embodiments and are not to be considered as limiting the present invention. In the entire drawings, the same reference symbols represent the same components;

[0046] Figure 1 Schematic diagram of a flow chart of a method for identifying an operating condition of an amphibious vehicle according to an embodiment of the present invention.

[0047] Figure 2 The figure is a schematic diagram of a process of identifying working conditions of an amphibious vehicle during launching according to another embodiment of the present invention.

[0048] Figure 3 The figure is a schematic diagram of a process of identifying working conditions of an amphibious vehicle during landing according to another embodiment of the present invention. DETAILED DESCRIPTION

[0049] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not used to limit the scope of the present invention.

[0050] Figure 1 Schematic diagram of a flow chart of a method for identifying an operating condition of an amphibious vehicle according to an embodiment of the present invention.

[0051] Combine the following Figure 1 An embodiment of the present invention will be described.

[0052] like Figure 1 As shown, the working condition identification method of the amphibious vehicle includes the following steps:

[0053] Step 101: Determine a first distance between the amphibious vehicle and a shoreline.

[0054] In this embodiment, the amphibious vehicle (hereinafter referred to as "vehicle") can have four operating conditions: on-land operating condition, land-to-water operating condition, water-to-land operating condition, and on-water operating condition. The vehicle can determine the current operating condition in real time based on the currently sensed information.

[0055] The first distance here can be a rough distance between the current vehicle and the coastline. In this embodiment, the rough distance da between the current vehicle's position and the coastline can be determined by establishing a global map and matching the high-precision positioning result. The current vehicle's position can be obtained through a global positioning system (GPS).

[0056] Step 102: If the first distance satisfies a first threshold range, determine the weight of the amphibious vehicle carried by the water body.

[0057] In this embodiment, the first distance can be compared with the first threshold range. If the first distance meets the first threshold range, it means that the vehicle may currently have a land-water cross-domain requirement. Let da0 be the first threshold, and the first threshold range can be less than or equal to 1.2da0. In some embodiments, the value range of da0 is 30-80 meters, and preferably, da0 is 50 meters.

[0058] In this embodiment, after comparing the first distance with the first threshold range, if the first distance does not meet the first threshold range, it means that the current vehicle is still in the initial working condition. Depending on the different initial working conditions of the current vehicle, the working condition output by the vehicle is also different. For example, if the initial working condition of the current vehicle is a land working condition, after determining that the first distance does not meet the first threshold range, the working condition output by the current vehicle is a land working condition; if the initial working condition of the preceding vehicle is a water working condition, after determining that the first distance does not meet the first threshold range, the working condition output by the current vehicle is a water working condition.

[0059] It should be noted that the first distance satisfying the first threshold range only means that the vehicle may currently have a need to cross land and water. As for whether to switch between land and water, it is necessary to further determine whether and when to enter the land and water switching condition.

[0060] In this embodiment, the weight of the vehicle body currently carried by the water body can be determined by the pressure sensor data distributed on the vehicle body. Depending on the current initial working condition of the vehicle, the initial value of the weight of the vehicle body currently carried by the water body obtained by the pressure sensor data is also different. For example, if the current initial working condition of the vehicle is a land working condition, the initial value of the weight of the vehicle body currently carried by the water body obtained by the pressure sensor data is 0; if the current initial working condition of the vehicle is a water working condition, the initial value of the weight of the vehicle body currently carried by the water body obtained by the pressure sensor data is the vehicle body weight m.

[0061] Depending on the current initial working condition of the vehicle, the vehicle weight carried by the current water body obtained from the pressure sensor data is also different. If the current initial working condition of the vehicle is a land working condition, the vehicle weight carried by the current water body obtained from the pressure sensor data can be m1, and the initial value of m1 is 0. When the vehicle enters the water-land switching working condition, m1 begins to increase; if the current initial working condition of the vehicle is a water working condition, the vehicle weight carried by the current water body obtained from the pressure sensor data can be m2. When the vehicle enters the water-land switching working condition, m2 begins to decrease.

[0062] In some optional embodiments, if the initial operating condition of the vehicle is a land operating condition and if the first distance satisfies a first threshold range, the following steps are further performed:

[0063] The second distance between the vehicle and the shoreline is calculated based on the vehicle's camera data and laser point cloud data. In this embodiment, the vehicle's camera equipment can be used to obtain camera data such as images, and the vehicle's laser radar can be used to obtain laser point cloud data of the environment. If the first distance meets the first threshold range, it indicates that the current vehicle has a land and water cross-domain requirement, and the precise distance between the vehicle and the shoreline can be determined by the fusion information of the laser point cloud and the image. For image information, the shoreline can be identified using an image-based edge detection algorithm. For laser point cloud information, the shoreline can be identified based on the difference in echo intensity caused by the different land and water media. After identifying the shoreline, the precise distance d between the vehicle and the shoreline can be calculated from the perspective of the vehicle's land observation. b The second distance here can be the precise distance d between the vehicle and the shoreline calculated based on the vehicle's camera data and laser point cloud data. b .

[0064] In some optional embodiments, if the initial operating condition of the vehicle is an on-water condition and if the first distance satisfies a first threshold range, the following steps are further performed:

[0065] The height of the vehicle from the ground is calculated based on the sonar data of the vehicle. In this embodiment, if the first distance meets the first threshold range, it indicates that the current vehicle has a need to cross land and water. When the vehicle has a need to cross land and water, the height h between the vehicle and the bottom of the water, that is, the height of the vehicle from the ground can be obtained based on the sonar data of the vehicle.

[0066] Step 103: If the initial operating condition of the amphibious vehicle is a land operating condition and the weight is greater than a weight threshold, or if the initial operating condition of the amphibious vehicle is a water operating condition and the absolute value of the difference between the weight and the weight of the amphibious vehicle is greater than the weight threshold, use a first machine learning model to map the suspension attitude angle data of the amphibious vehicle to an intermediate operating condition of the amphibious vehicle.

[0067] In this embodiment, the weight threshold may be m0, and the value range of m0 is between 0.1m-0.2m, where m is the vehicle body weight. If the initial working condition of the vehicle is the land working condition, when the vehicle body weight m1 carried by the water body is greater than 0, it means that the vehicle body is partially carried by the water body at this time; when the vehicle body weight m1 carried by the water body is greater than m0, it means that the vehicle body weight carried by the water body has changed significantly at this time, and it can be determined through subsequent steps whether the vehicle has entered the land-to-water working condition. If the initial working condition of the vehicle is the water working condition, when the vehicle body weight m2 carried by the water body is less than the vehicle body weight m, it means that the vehicle body weight is no longer completely carried by the water body at this time; when the absolute value of the difference between the vehicle body weight m2 carried by the water body and the vehicle body weight m is greater than m0, it means that the weight carried by the vehicle tires or tracks has changed significantly at this time, and it can be determined through subsequent steps whether the vehicle has entered the water-to-land working condition.

[0068] In this embodiment, in order to prevent errors in the judgment results of environmental information and to more accurately control the position information of the vehicle during landing or entering the water, even if it is determined that the weight of the vehicle body carried by the current water body has changed significantly, the current vehicle will not enter the water-land switching condition, and a comprehensive judgment is required based on the vehicle's own status information.

[0069] In this embodiment, the time sequence of the changes in the vehicle suspension attitude angle can be obtained through the data of multiple suspension attitude sensors distributed at the bottom of the vehicle, and then the progressive state of the vehicle when it is currently landing or entering the water can be obtained. Assuming that the vehicle suspension attitude angle is α, an accurate identification model can be obtained by training limited data samples to identify the current land-water crossing state of the vehicle. The first machine learning model in this embodiment can be the aforementioned trained identification model.

[0070] The training process of the aforementioned recognition model is described below.

[0071] In this embodiment, the recognition model can be based on a support vector machine (SVM), given the classification information of the current working condition (onshore working condition, land-to-water working condition, water working condition, water-to-land working condition), and trained with limited data sample information. The resulting recognition model can accurately complete the classification of the four labels. The specific process is as follows:

[0072] (1) The data set related to the vehicle's suspension attitude sensor information may include the relative suspension displacement d of the left and right, front and rear four road wheels. a d b d c d d In order to avoid the situation where the change of the intuitive observation quantity cannot reflect all the characteristics of the complex working process, this embodiment can calculate the relative displacement d ad b d c d d Data processing is performed to obtain a set of candidate feature parameters for subsequent feature parameter selection. For example, the feature parameter set may include: the maximum displacement of each road wheel suspension, the maximum and minimum displacement difference, the maximum displacement change rate, the average displacement change rate and the starting point displacement value, etc. Then, the feature parameter with the highest correlation with the working condition identification is selected through the feature selection method for subsequent training.

[0073] It should be noted that the relative displacement d of the four road wheels is a d b d c d d The characteristic parameters used for subsequent training are obtained by feature engineering these intuitive observation data. For example, the maximum displacement of each road wheel suspension refers to the displacement from d a (t), d b (t), d c (t), d d (t) Select the maximum displacement d a (t a ), d b (t b ), d c (t c ), d d (t d ), where d a (t), d b (t), d c (t), d d (t) represents the relative suspension displacement of the four road wheels. a d b d c d d Time series variation function, d a (t a ), d b (t b ), d c (t c ), d d (t d ) represent the maximum extreme values ​​of the time series variation function.

[0074] (2) The kernel function parameter g and penalty parameter C of the support vector machine are optimized by particle swarm optimization. Specifically, the training validation set is divided into 4 groups, 3 of which are used as training data samples. The prediction performance of each group (g, C) is measured by the average of the mean square error of the prediction errors obtained from the 4 validations. After searching, the set of parameters with the smallest mean error variance is finally determined as the optimal parameter.

[0075] (3) Through training and testing of experimental data, when the model recognition accuracy can stably reach more than 95%, the training is completed and the required recognition model is obtained.

[0076] Now back to step 103, if the weight change of the vehicle body carried by the water body is greater than the weight threshold, the suspension attitude angle data of the current vehicle can be mapped to the intermediate working condition of the vehicle through the above-mentioned trained identification model. It should be noted that the intermediate working condition here means that the working condition currently output by the identification model is not the final working condition of the vehicle, and it is necessary to further determine whether and when to enter the water-land switching working condition.

[0077] In some optional embodiments, if the initial working condition of the vehicle is a land working condition and if the first distance satisfies the first threshold range, the first machine learning model can be used to map the suspension attitude angle data of the vehicle to the intermediate working condition of the vehicle only when the vehicle body weight m1 carried by the water body is greater than the weight threshold m0 and the second distance satisfies the second threshold range. In this embodiment, the maximum value of the second threshold range is less than the maximum value of the first threshold range. For example, if the first threshold range is [0,1.2da0], the second threshold range is [0,db0], where da0 is greater than db0. When da0 is 50 meters, db0 can be 30 meters.

[0078] In some optional embodiments, if the initial working condition of the vehicle is the working condition on water and if the first distance satisfies the first threshold range, the first machine learning model can be used to map the suspension attitude angle data of the vehicle to the intermediate working condition of the vehicle only when the vehicle weight m1 carried by the water is greater than the weight threshold m0 and the vehicle's ground clearance h is less than or equal to the height threshold h0. Exemplarily, the height threshold h0 is 5 meters.

[0079] Step 104: Determine a land dynamics model matching rate and a water dynamics model matching rate of the amphibious vehicle.

[0080] It should be noted that the intermediate working condition output in step 103 is the result obtained based on the prior information. In step 104, it is also necessary to identify the state of the current vehicle entering and exiting the water by calculating the feedback information of the current vehicle during the water-land switching process, so as to form a verification relationship with the prior result in step 103.

[0081] In this embodiment, a simplified dynamic model for driving on land (hereinafter referred to as the "land dynamic model") and a simplified dynamic model for driving on water (hereinafter referred to as the "water dynamic model") can be established respectively to obtain the dynamic response characteristics in the fully attached and fully floating states. Then, according to the land dynamic model and the water dynamic model, the matching rate of the land dynamic model and the matching rate of the water dynamic model are calculated in combination with the feedback information of the current vehicle.

[0082] The following is an analysis of the vehicle's dynamics when driving on land and on sea.

[0083] (1) First, the dynamics analysis of the vehicle on land. Considering the motion state of the amphibious vehicle during the land-water switching process, the ramp dynamics model of the tracked vehicle can be established by analyzing the steering dynamics of the tracked vehicle when driving on a ramp as follows:

[0084]

[0085] Wherein, T is the output torque of the engine; η is the mechanical transmission efficiency; i is the comprehensive transmission ratio; r is the radius of the driving wheel; G is the vehicle's own gravity; f is the ground friction coefficient; μ is the steering resistance coefficient; α is the ramp angle value; β is the azimuth angle of the vehicle on the slope, It is the total azimuth of the vehicle's rotation around the z' axis at time t, including the azimuth of the vehicle at its initial position and the azimuth of the vehicle's rotation around the z' axis from t0 to t The z′ axis is the direction axis perpendicular to the slope; C ,y C The lateral and longitudinal distances between the center of gravity and the geometric center of the vehicle are F l is the driving force on the left side of the vehicle, F r The driving force for the right side of the vehicle.

[0086] The current vehicle's left and right driving forces F l and F r As an unknown quantity, the output torque T of the engine of the current vehicle can be obtained by solving it. Then the actual torque T′ of the engine of the current vehicle is obtained. Finally, the output torque T of the engine and the actual torque T′ of the engine are compared to obtain the matching rate of the land dynamics model.

[0087] (2) Then comes the dynamic analysis of the amphibious vehicle in water. Considering that the autonomous navigation of amphibious vehicles mainly focuses on the heading and position information of the vehicle body, and there is a lack of effective control devices and control methods for the roll, pitch and heave motion of the vehicle body, although the amphibious vehicle has six degrees of freedom in water motion, only the pitch, roll and bow are considered in actual motion control, so the three-degree-of-freedom model of the amphibious vehicle is established as follows:

[0088]

[0089] in: is the position and angle vector of the vehicle body, υ=[u,V,r] T is the velocity and angular velocity vector of the vehicle body; υ=[u r, v r ,0] T represents the ocean current velocity vector, R(η) is the rotation matrix between the two coordinate systems; τ=[τ u , τ v , τ r ] T is the thrust vector of the vehicle; τ b =[τ b1 , τ b2 , τ b3 ] T is the environmental disturbance force matrix, M is the vehicle inertia force matrix, C(υ) is the centripetal force matrix, and D(υ) is the damping matrix.

[0090] Considering that the driving system of the amphibious vehicle is a double-push type, it is propelled by two water jet propulsion units symmetrically arranged at the rear of the amphibious vehicle. It realizes sea navigation control by controlling the flow of the water jet propulsion units and the opening and closing of the water gates: if the water gate on one side is closed so that the reverse waterway is connected, the greater the speed of the propulsion units on both sides, the greater the steering change. Therefore, the thrust calculation of the water jet propulsion unit can be obtained:

[0091]

[0092] Among them, P e is the engine power, n is the engine speed, ρ ω is the density of the fluid, Q is the flow rate at the water jet outlet, v j is the flow rate of the fluid at the outlet, α s is the inflow influence coefficient, v s is the speed of the vehicle; θ2 is the angle between the right waterjet thrust and the vehicle heading, θ1 is the angle between the left waterjet thrust and the vehicle heading; l2 is the distance between the right waterjet thrust and the vehicle drift center, l1 is the distance between the left waterjet thrust and the vehicle drift center.

[0093] Taking the current vehicle's engine speed as an unknown quantity and solving it, we can get the current vehicle's engine power P e Then obtain the actual power P of the current vehicle engine e '. Finally, the calculated engine power P e and the actual power P of the transmitter e 'By comparing, we can get the matching rate of the water dynamics model.

[0094] Step 105: Determine the final operating condition of the amphibious vehicle according to the intermediate operating condition, the land dynamics model matching rate and the water dynamics model matching rate.

[0095] In this embodiment, the current working condition identification result can be determined based on the changing rules of the onshore dynamics model matching efficiency and the onshore dynamics model matching efficiency during the land-water switching process.

[0096] Specifically, if the intermediate operating condition is a land operating condition and the land dynamic model matching rate meets the third threshold range and the water dynamic model matching rate meets the fourth threshold range, the final operating condition of the amphibious vehicle is determined to be a land operating condition. In some embodiments, the third threshold range may be greater than 0.9, and the fourth threshold range may be less than 0.1.

[0097] If the intermediate operating condition is the land-to-water operating condition and both the land dynamics model matching rate and the water dynamics model matching rate meet the sixth threshold range, the final operating condition of the amphibious vehicle is determined to be the land-to-water operating condition. In some embodiments, the sixth threshold range may be greater than 0.3 and less than 0.9.

[0098] If the intermediate operating condition is a water-to-land operating condition and both the on-land dynamic model matching rate and the on-water dynamic model matching rate satisfy the sixth threshold range, the final operating condition of the amphibious vehicle is determined to be a water-to-land operating condition. In some embodiments, the sixth threshold range may be greater than 0.3 and less than 0.9.

[0099] If the intermediate operating condition is an on-water operating condition and the on-land dynamics model matching rate satisfies the fourth threshold range and the on-water dynamics model matching rate satisfies the third threshold range, it is determined that the final operating condition of the amphibious vehicle is an on-water operating condition.

[0100] The working condition identification method of an amphibious vehicle provided in an embodiment of the present invention combines two methods: active detection and passive contact. It should be noted that the active detection method here refers to the vehicle's perception of the external environment, such as position positioning. The passive contact method here refers to the vehicle's own state feedback, such as pressure sensors, suspension attitude angle sensors, dynamic model matching, etc. The working condition identification method of an amphibious vehicle provided in an embodiment of the present invention solves the linear measurement of the vehicle's state in the water-land transition state, guides the continuous conversion control of the vehicle in the water-land transition state, and improves the adaptability of the amphibious general unmanned vehicle under different cross-domain working conditions. The linear measurement here means that the working condition changes are all linear, not jumps. Taking the landing process in the water-land cross-domain as an example, the linear measurement means that during the landing process, the identification result must follow a linear process such as "water-water to land-land", rather than a nonlinear process such as "water-land-water to land-land".

[0101] In some optional embodiments, a safety mechanism may be provided to prevent the vehicle from entering the land-to-water condition for a long time without changing the current condition to a pure water condition. The current vehicle speed may be used to obtain the driving distance X of the vehicle after entering the land-to-water condition. When X is greater than a given threshold value X0, the current vehicle has completely entered the floating condition, and the identification result "water condition" is directly outputted, and the condition identification of the amphibious vehicle entering the water is completed.

[0102] In some optional embodiments, considering the complexity of the land-water cross-domain working condition, even if the final working condition outputted in step 105 is the water working condition, a safety mechanism can be set. The driving distance X after the vehicle enters the land-to-water working condition can be obtained through the current vehicle speed. Only when the absolute value of the difference between X and the given threshold value X0 is less than the set threshold value ΔX, can it be determined that the current vehicle has completed the land-to-water switch and entered the fully floating working condition. At this time, the identification result "water working condition" is output. Otherwise, it is necessary to continue driving until |X-X0|>ΔX before it can enter the fully floating working condition. The working condition identification of the amphibious vehicle entering the water is completed.

[0103] In some optional embodiments, a safety mechanism may be provided to prevent the vehicle from entering the water-to-land condition for a long time without changing the current condition to the pure land condition. The current vehicle speed may be used to obtain the driving distance X of the vehicle after entering the water-to-land condition. When X is greater than a given threshold value X0, it is considered that the current vehicle has completely entered the pure land condition, and the identification result "land condition" is directly outputted, and the condition identification of the amphibious vehicle landing process is completed.

[0104] In some optional embodiments, considering the complexity of the water-land cross-domain working condition, even if the final working condition outputted in step 105 is the land working condition, a safety mechanism can be set. The driving distance X after the vehicle enters the water-to-land working condition can be obtained through the current vehicle speed. Only when the absolute value of the difference between X and the given threshold value X0 is less than the set threshold value ΔX, can it be determined that the current vehicle has completed the water-to-land switch and entered the full adhesion working condition. At this time, the identification result "land working condition" is output. Otherwise, it is necessary to continue driving until |X-X0|>ΔX before entering the full adhesion working condition. The working condition identification of the amphibious vehicle landing process is completed.

[0105] Figure 2 This is a schematic diagram of a process for identifying working conditions of an amphibious vehicle during launching according to another embodiment of the present invention. Figure 2 Provide explanation.

[0106] Step S1: First, the vehicle can be positioned by GPS. Then, by matching the global electronic map with the GPS positioning results, the rough distance da between the current vehicle's location and the shoreline can be determined. Generally, when real-time dynamic differential positioning (RTK) is not enabled, the GPS positioning accuracy is 2.2m. At the same time, considering the occlusion problem, the threshold da0 can be set to 50m. When da is much larger than the given threshold da0 (for example, da>1.2da0), it is determined that the current vehicle does not need to go into the water and is still in a purely land operation state. At this time, the identification result "1" is output; when da is close to or less than da0 (for example, da≤1.2da0), it is considered that the vehicle may have a water-land cross-domain demand, and it is necessary to combine the next steps to determine whether and when to enter the water-land switching condition.

[0107] Step S2: When the vehicle needs to cross land and water, the precise distance between the vehicle and the shoreline can be determined by the fusion information of the laser radar and the image. For the image information, the shoreline can be identified using the image-based edge detection algorithm. For the laser point cloud information, the shoreline can be identified based on the difference in echo intensity caused by the different land and water media. After the shoreline is identified, the precise distance d between the vehicle and the shoreline can be calculated from the perspective of the vehicle on land. b Since the stable sensing distance inferred based on existing technology is within 50m, the threshold d can be set b0 is 30m. b Still greater than the given threshold d b0 When d is , the vehicle is still considered to be in a pure land condition, and the output identification result is "1", and the system continues to return to step S1 for positioning and matching. b Less than or equal to d b0 , the vehicle is considered to be approaching the land-to-water condition.

[0108] Step S3: Considering the complexity of the cross-domain working condition between water and land, in order to prevent the accuracy of shoreline identification from being affected and making it difficult to trigger further working condition judgment, a safety mechanism can be set through the pressure sensor data distributed on the vehicle body. The pressure sensor data can be used to obtain the weight m1 of the vehicle body currently carried by the water body and then judge the current state of the vehicle body entering the water. When the vehicle is purely on land, the value of the weight m1 of the vehicle body carried by the water body is 0. When the pressure data feedback shows that the vehicle body is partially carried by the water body at this time, that is, the weight m1 of the vehicle body carried by the water body is greater than the threshold m0, it is determined that the weight of the vehicle body carried by the water body at this time has changed significantly, and subsequent steps can be used to determine whether the vehicle has entered the land-to-water working condition. m0 can take the value [0.1m1, 0.2m1].

[0109] Step S4: In order to prevent errors in the judgment results of environmental information and to more accurately control the location information of the landing process of the amphibious vehicle, it is also necessary to make a judgment based on the status information of the vehicle itself. The time sequence law of the successive state changes of the vehicle suspension angle can be obtained through the data of multiple suspension attitude sensors distributed at the bottom of the vehicle, and then the progressive state of the vehicle when it enters the water can be obtained. Assuming the vehicle suspension attitude angle is α, an accurate identification model can be obtained by training limited data samples to determine the current cross-domain state of the vehicle (i.e., the "intermediate working condition" in this article). The specific method has been discussed in the previous article and will not be repeated here.

[0110] Step S5: Establish a simplified dynamic model for on-land driving and a simplified dynamic model for on-water driving, respectively, and obtain the dynamic response characteristics in the fully attached and fully floating states. The current vehicle's state of entering and exiting the water is identified by calculating the feedback information of the current vehicle during the land-water switching process, and a verification relationship is formed for the priori results obtained in step S4. Specifically, the matching efficiency E of the on-land dynamic model is a Matching efficiency E with the hydrodynamic model b The changing rules in the process of water-land switching determine the current working condition identification results. The matching efficiency E of the land dynamic model a Matching efficiency E with the hydrodynamic model b The calculation method of has been discussed in the previous article and will not be repeated here. a Greater than 0.3 and less than 0.9, hydrodynamic model E b When it is greater than 0.3 and less than 0.9, it is determined that the current vehicle is in the land-to-water condition, and the identification result "2" is output.

[0111] Step S6: In order to prevent the above judgment from not placing the current working condition in a pure water working condition after a long period of time after entering the land-to-water working condition, a safety mechanism can be set here. The driving distance X of the vehicle after entering the land-to-water working condition can be obtained through the current vehicle speed. When X is greater than the given threshold value X0, it is considered that the current vehicle has completely entered the floating working condition, and the identification result "4" is directly output, and the working condition identification of the amphibious vehicle entering the water process is completed.

[0112] Step S7: Without triggering the forced classification into floating condition in step S6, if the current classification result of the identification model is "4" and the matching efficiency of the onshore dynamics model is E a Greater than 0.3, hydrodynamic model E bWhen it is greater than 0.9, a safety mechanism can still be set. This can be achieved by judging whether the driving distance meets the requirements. Only when the difference between the driving distance X and the given distance X0 is less than the threshold ΔX can it be considered that the current vehicle has completed the switch from land to water and entered the fully floating condition. At this time, the identification result "4" is output; otherwise, it is necessary to continue driving until |X-X0|>ΔX before it can enter the fully floating condition. At this point, the working condition identification of the amphibious vehicle entering the water is completed.

[0113] Figure 3 This is a schematic diagram of a process flow for identifying working conditions of an amphibious vehicle during landing according to another embodiment of the present invention. Figure 3 Provide explanation.

[0114] Step S1: First, the vehicle can be positioned by GPS. Then, by matching the global electronic map with the GPS positioning results, the rough distance da between the current vehicle's location and the shoreline can be determined. Generally, when real-time dynamic differential positioning (RTK) is not enabled, the GPS positioning accuracy is 2.2m. At the same time, considering the occlusion problem, the threshold da0 can be set to 50m. When da is much larger than the given threshold da0 (for example, da>1.2da0), it is determined that the current vehicle does not need to go into the water and is still in a purely land operation state. At this time, the identification result "1" is output; when da is close to or less than da0 (for example, da≤1.2da0), it is considered that the vehicle may have a water-land cross-domain demand, and it is necessary to combine the next steps to determine whether and when to enter the water-land switching condition.

[0115] Step S2: When the vehicle has a need to cross water and land, the height h between the vehicle and the bottom of the water can be obtained based on the sonar data of the vehicle. The threshold h0 can be set to 5m. When h is less than or equal to the given threshold h0, it is considered that the vehicle is close to the water-to-land condition.

[0116] Step S3: Considering that there is no shoreline identification data during the landing process, it is difficult to ensure accuracy by relying solely on the height determination of the sonar. A safety mechanism can be set up through the pressure sensor data distributed on the vehicle body. The pressure sensor data can be used to obtain the vehicle body weight m1 currently carried by the water body and then judge the current landing status of the vehicle body. When the pressure data feedback shows that the vehicle body weight m1 is no longer completely carried by the water body, that is, when the difference Δm between the current water body weight and the vehicle weight m is greater than the given threshold m0, it is determined that the vehicle's tire or track weight has changed significantly, and it can be judged later whether the vehicle has entered the water-to-land condition. m0 can take the value [0.1m1, 0.2m1].

[0117] Step S4: In order to prevent errors in the judgment results of environmental information and to more accurately control the location information of the landing process of the amphibious vehicle, it is also necessary to make a judgment based on the status information of the vehicle itself. The time sequence law of the successive state changes of the vehicle suspension angle can be obtained through the data of multiple suspension attitude sensors distributed at the bottom of the vehicle, and then the progressive state of the vehicle at the current landing can be obtained. Assuming the vehicle suspension attitude angle is α, an accurate identification model can be obtained by training limited data samples to determine the cross-domain state of the current vehicle (i.e., the "intermediate working condition" in this article). The specific method has been discussed in the previous article and will not be repeated here.

[0118] Step S5: Establish a simplified dynamic model for on-land driving and a simplified dynamic model for on-water driving, respectively, and obtain the dynamic response characteristics in the fully attached and fully floating states. The current vehicle's state of entering and exiting the water is identified by calculating the feedback information of the current vehicle during the land-water switching process, and a verification relationship is formed for the priori results obtained in step S4. Specifically, the matching efficiency E of the on-land dynamic model is a Matching efficiency E with the hydrodynamic model b The changing rules in the process of water-land switching determine the current working condition identification results. The matching efficiency E of the land dynamic model a Matching efficiency E with the hydrodynamic model b The calculation method of has been discussed in the previous article and will not be repeated here. a Greater than 0.3 and less than 0.9, hydrodynamic model E b When it is greater than 0.3 and less than 0.9, it is determined that the current vehicle is in a water-to-land condition, and the identification result "3" is output.

[0119] Step S6: In order to prevent the above judgment from not placing the current working condition in a pure land working condition after a long period of time after entering the water-to-land working condition, a safety mechanism can be set here. The driving distance X of the vehicle after entering the water-to-land working condition can be obtained through the current vehicle speed. When X is greater than the given threshold value X0, it is considered that the current vehicle has completely entered the land working condition, and the identification result "1" is directly output, and the working condition identification of the amphibious vehicle landing process is completed. It should be noted that the threshold value X0 here is a priori value, and different values ​​can be set according to different waters.

[0120] Step S7: Under the premise that the forced classification into the onshore condition in step S6 is not triggered, if the current classification result of the identification model is "1" and the onshore dynamic model matching efficiency E a Greater than 0.9, hydrodynamic model E bWhen it is less than 0.1, a safety mechanism can still be set. This can be achieved by judging whether the driving distance meets the requirements. Only when the difference between the driving distance X and the given distance X0 is less than the threshold ΔX can it be considered that the current vehicle has completed the switch from water to land and entered the full land condition. At this time, the identification result "1" is output; otherwise, it is necessary to continue driving until |X-X0|>ΔX before it can enter the full land condition. At this point, the working condition identification of the amphibious vehicle landing process is completed. It should be noted that the thresholds X0 and ΔX here are a priori values, and different values ​​can be set according to different waters.

[0121] Compared with the prior art, the working condition identification method of an amphibious vehicle provided by the embodiment of the present invention can achieve at least one of the following beneficial effects:

[0122] 1. The solution of the present invention integrates multi-source sensor information such as lidar, images, sonar, and prior information such as electronic charts, thereby effectively improving the accuracy of working condition identification.

[0123] 2. The solution of the present invention integrates active detection and passive contact, solves the linear measurement of the vehicle's state in the water-land transition state, guides the continuous conversion control of the vehicle in the water-land transition state, and improves the adaptability of the universal water and land unmanned vehicle under different cross-domain working conditions.

[0124] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. A method for identifying the working condition of an amphibious vehicle, characterized in that: include: determining a first distance between the amphibious vehicle and a shoreline; If the first distance satisfies a first threshold range, determining the weight of the amphibious vehicle carried by the water body; If the initial operating condition of the amphibious vehicle is a land operating condition and the weight is greater than a weight threshold, or if the initial operating condition of the amphibious vehicle is a water operating condition and the absolute value of the difference between the weight and the weight of the amphibious vehicle is greater than the weight threshold, using a first machine learning model, mapping the suspension attitude angle data of the amphibious vehicle to an intermediate operating condition of the amphibious vehicle; Determining a land dynamics model matching rate and a water dynamics model matching rate of the amphibious vehicle; and The final operating condition of the amphibious vehicle is determined according to the intermediate operating condition, the land dynamics model matching rate and the water dynamics model matching rate.

2. The method for identifying the working condition of an amphibious vehicle according to claim 1, characterized in that: The training process of the first machine learning model includes: Acquire training samples from a training sample set for the first machine learning model, wherein the training samples include suspension attitude angle data of an amphibious vehicle and operating condition labels of the amphibious vehicle, and the operating condition labels include: on-land operating condition, land-to-water operating condition, water-to-land operating condition, and on-water operating condition; generating an operating condition of the amphibious vehicle using the first machine learning model; and Based on the comparison between the operating condition and the operating condition label, the first machine learning model is updated.

3. The method for identifying the working condition of an amphibious vehicle according to claim 2, characterized in that: If the initial working condition of the amphibious vehicle is a land working condition, and if the first distance satisfies a first threshold range, the method further includes calculating a second distance between the amphibious vehicle and the shoreline based on the camera data and the laser point cloud data of the amphibious vehicle; and, if the weight is greater than a weight threshold and if the second distance satisfies a second threshold range, mapping the suspension attitude angle data of the amphibious vehicle to an intermediate operating condition of the amphibious vehicle using a first machine learning model; The maximum value of the second threshold range is smaller than the maximum value of the first threshold range.

4. The method for identifying the working condition of an amphibious vehicle according to claim 2, characterized in that: If the initial working condition of the amphibious vehicle is the working condition on water, and if the first distance satisfies a first threshold range, the method further includes calculating the height of the amphibious vehicle from the ground according to the sonar data of the amphibious vehicle; Furthermore, if the absolute value of the difference between the weight and the weight of the amphibious vehicle is greater than a weight threshold and the height above the ground is less than or equal to a height threshold, the suspension attitude angle data of the amphibious vehicle is mapped to an intermediate working condition of the amphibious vehicle using a first machine learning model.

5. The method for identifying the working condition of an amphibious vehicle according to claim 2, characterized in that: Determining the final operating condition of the amphibious vehicle according to the intermediate operating condition, the land dynamics model matching rate, and the water dynamics model matching rate includes: If the intermediate operating condition is a land operating condition and the land dynamics model matching rate satisfies a third threshold range and the water dynamics model matching rate satisfies a fourth threshold range, determining a travel distance of the amphibious vehicle after entering the water-to-land operating condition; If the difference between the driving distance and the driving distance threshold satisfies a fifth threshold range, it is determined that the final operating condition of the amphibious vehicle is a land operating condition.

6. The method for identifying the working condition of an amphibious vehicle according to claim 2, characterized in that: Determining the final operating condition of the amphibious vehicle according to the intermediate operating condition, the land dynamics model matching rate, and the water dynamics model matching rate includes: If the intermediate operating condition is a land-to-water operating condition and both the land dynamics model matching rate and the water dynamics model matching rate satisfy a sixth threshold range, determining a travel distance of the amphibious vehicle after entering the land-to-water operating condition; If the driving distance is greater than a driving distance threshold, it is determined that the final operating condition of the amphibious vehicle is an on-water operating condition.

7. The method for identifying the working condition of an amphibious vehicle according to claim 2, characterized in that: Determining the final operating condition of the amphibious vehicle according to the intermediate operating condition, the land dynamics model matching rate, and the water dynamics model matching rate includes: If the intermediate working condition is a water-to-land working condition and both the on-land dynamic model matching rate and the on-water dynamic model matching rate satisfy a sixth threshold range, determining a travel distance of the amphibious vehicle after entering the water-to-land working condition; If the driving distance is greater than a driving distance threshold, it is determined that the final operating condition of the amphibious vehicle is a land operating condition.

8. The method for identifying the working condition of an amphibious vehicle according to claim 2, characterized in that: Determining the final operating condition of the amphibious vehicle according to the intermediate operating condition, the land dynamics model matching rate, and the water dynamics model matching rate includes: If the intermediate working condition is an on-water working condition and the on-land dynamic model matching rate satisfies a fourth threshold range and the on-water dynamic model matching rate satisfies a third threshold range, determining a travel distance of the amphibious vehicle after entering the land-to-water working condition; If the difference between the driving distance and the driving distance threshold satisfies a fifth threshold range, it is determined that the final operating condition of the amphibious vehicle is an on-water operating condition.

9. The method for identifying the working condition of an amphibious vehicle according to claim 2, characterized in that: The training process of the first machine learning model includes: Selecting feature parameters related to the working condition label for training from the suspension attitude angle data through feature engineering; According to the characteristic parameters and the operating condition labels, the kernel function parameters and the penalty parameters of the support vector machine are optimized using a particle swarm algorithm; and If the operating condition identification accuracy of the support vector machine reaches a preset threshold, the first machine learning model is obtained.

10. The method for identifying the working condition of an amphibious vehicle according to claim 1, characterized in that: Determining the weight of the amphibious vehicle carried by the water body includes: The weight of the amphibious vehicle carried by the water body is determined according to the pressure sensor data of the amphibious vehicle.

Images