A sensor trajectory planning method and system based on an adaptive field model

Abstract

The application discloses a kind of sensor trajectory planning method and system based on adaptive field model, the described including the following steps: S1.different sensor data is aligned on coordinate and time;S2.received signal power is recorded, and noise level is estimated, and the uncertainty of each sampling point is also output;S3.log distance model is used to fit the average fading trend of signal with distance and height, and then Gaussian process is used to model residual;S4.calculate information gain, and form comprehensive utility function in combination with flight energy consumption or time cost, generate candidate point set in flyable area;S5.based on candidate point set, under the constraints of energy and time, the optimal sampling path is planned.The application can consider flight efficiency and sampling quality, and can intelligently adjust the trajectory planning according to the actual characteristics of field.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of automation sensing and robotics, in particular to a sensor trajectory planning method and system based on an adaptive field model. BACKGROUND

[0002] Currently, the sampling and reconstruction of spatial physical fields using unmanned devices mainly relies on preset trajectory planning strategies.

[0003] 1. Boustrophedon Path: This is the most commonly used coverage path strategy. Its advantages are simple path planning and high execution efficiency. However, when the sensor is a point sensor (such as a signal strength meter), this method results in extremely dense sampling points in the direction of the path and very sparse sampling points between paths, forming "anisotropic" sampling. This leads to serious unevenness in the accuracy of the reconstructed field model, resulting in large uncertainties and distortions in the areas between the flight lines.

[0004] 2. Uniform Grid Sampling: To overcome the above problems, an improved method is to define a uniform two-dimensional grid point set in advance, and then plan a shortest path (such as using the TSP algorithm) to visit these points. This method ensures the spatial uniformity of the sampling points. However, its main defects are "blindness" and "inefficiency": it treats all areas of the field equally, resulting in a waste of a large amount of flight and sampling resources in areas where the field changes smoothly, while in critical areas where the field changes dramatically, the sampling density may still be insufficient. SUMMARY

[0005] The present application aims to overcome the shortcomings of the prior art and provide a sensor trajectory planning method and system based on an adaptive field model, which can balance flight efficiency and sampling quality.

[0006] The purpose of the present application is achieved by the following technical solution: a sensor trajectory planning method based on an adaptive field model, comprising the following steps:

[0007] S1. Convert the original positioning data collected by the unmanned aerial vehicle into ECEF, then into ENU coordinates, and interpolate and synchronize the time stamps of RTK, IMU, and received signal power to achieve time alignment of the data;

[0008] S2. Construct an uncertainty model of the observation value of the received signal power, and record the received signal power when the unmanned aerial vehicle hovers or passes through the specified sampling points, and construct an observation set representing the uncertainty of the sampling points;

[0009] S3. Fitting the average fading trend of signal with distance and height by using log-distance model, and modeling the residual by using Gaussian process to obtain the prediction model of the mean and variance of the received signal power of the sampling points;

[0010] S4. Calculating information gain, and combining with flight energy consumption or time cost to form a comprehensive utility function, and generating a candidate point set in the flyable area;

[0011] S5. Planning the optimal sampling path under the energy and time constraints based on the candidate point set.

[0012] Preferably, the application further comprises a trajectory updating step S6, the UAV flies according to the path , stays for sampling; the new observation is added to the data set D, and the GP model is updated to obtain a new ; checking the trigger condition: the global uncertainty is still large; the local interference suddenly increases; the power / time is insufficient; if triggered, returning to steps S4 to S5 for re-planning, so as to roll the updated trajectory and prediction map.

[0013] A sensor trajectory planning system based on an adaptive field model, comprising:

[0014] An unmanned aerial vehicle flight platform for carrying sensors and communication modules to perform low-altitude flight tasks;

[0015] A positioning and attitude perception module for providing centimeter-level positioning through RTK-GPS and providing attitude and speed information through IMU for spatial alignment and track control;

[0016] A communication signal collection module for receiving target frequency band signals, extracting received signal power, and an uncertain observation set;

[0017] A field modeling and decision-making module, an embedded computer / ground station, responsible for establishing a prediction model of the mean and variance of the received signal power of the sampling points by fitting the average fading trend of the signal with distance and height by using a log-distance model, and modeling the residual by using a Gaussian process;

[0018] A path planning and control module for trajectory planning to generate a sampling path based on information gain and energy consumption constraints, and driving the UAV to perform flight tasks.

[0019] The application has the beneficial effects that: the application can take into account flight efficiency and sampling quality by generating a candidate point set and planning the optimal sampling path under the energy and time constraints; and through the updating of the trajectory, the application can iteratively find and fill in the missing points, actively find and encrypt abnormal points and dramatic boundary points in the sampling field, and finally reconstruct a field model with higher precision, more details and fewer artifacts. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 A method flowchart of the present application. DETAILED DESCRIPTION

[0021] The technical solutions of the present application will be described in further detail below in conjunction with the drawings, but the scope of protection of the present application is not limited to the following description.

[0022] As shown in the figure, a sensor trajectory planning method based on an adaptive field model includes the following steps: Figure 1

[0023] S1. The original positioning data collected by the unmanned aerial vehicle is uniformly converted into ECEF, and then into ENU coordinates, and the timestamps of RTK, IMU and received signal power are interpolated and synchronized to realize the time alignment of the data.

[0024] S101. Given the reference point geodetic coordinates: and the unmanned aerial vehicle RTK geodetic coordinates: ; wherein respectively represent the longitude, latitude and ellipsoidal height of the reference point, respectively represent the longitude, latitude and ellipsoidal height of the unmanned aerial vehicle;

[0025] Under the given latitude , the calculation formula of the curvature radius of the prime vertical circle is determined.

[0026]

[0027] wherein is the length of the long semi-axis of the earth, in meters; represents the first eccentricity square, , represents the flattening of the earth as an ellipsoid, ;

[0028] S102. Given the conversion formula of the reference point geodetic coordinate system and the unmanned aerial vehicle RTK geodetic coordinate system to the ECEF coordinate system:

[0029]

[0030] is the coordinate obtained by conversion to the ECEF coordinate system; the ECEF coordinate system refers to the Earth-Centered Earth-Fixed rectangular coordinate system;

[0031] When is taken as , the obtained is the conversion result of the reference point geodetic coordinate system to the ECEF coordinate system, denoted as = ( ,​

[0032] When Take , the resulting is the conversion result of the UAV RTK geodetic coordinate system to the ECEF coordinate system, denoted as ;

[0033] S103. Convert the ECEF coordinate system to the local ENU coordinate system:

[0034] Calculate the relative displacement of the UAV and the reference point in the ECEF coordinate system :

[0035]

[0036] The rotation matrix of the reference point ECEF coordinate system to the ENU coordinate system with the reference point as the origin is:

[0037]

[0038] Calculate the position of the UAV in the ENU coordinate system with the reference point as the origin :

[0039] .

[0040] S104. Time alignment: align the timestamps of RTK, IMU and received signal power to the unified time t with the UAV master clock interpolation, to ensure the consistency of the time series, where UAV refers to the unmanned aerial vehicle.

[0041] S2. Construct the uncertainty model of the observation value of the received signal power. When the unmanned aerial vehicle hovers or passes through the specified sampling point, record the received signal power and construct the observation set representing the uncertainty of the sampling point.

[0042] S201. Construct the uncertainty model of the received power as:

[0043]

[0044] Where:

[0045] denotes the logarithmic distance fading and height correction function:

[0046]

[0047] Where, , represents a parameter vector composed of reference power, path loss exponent, and height correction coefficient; is the reference power, which represents the average received power at the reference distance.​ is the path loss exponent, the rate at which the control signal attenuates with distance; is the height correction coefficient, used to compensate the influence of UAV height on signal strength; is the height of the location x from the reference point, is the horizontal distance of the location from the reference point;

[0048] represents the residual term, used to capture shadow fading and interference;

[0049] represents the measurement noise, embodied by the variance of the received power measurements;

[0050] S202. At each UAV sampling point, multiple received power measurements are taken to obtain an observation set wherein, is the location of the i-th sampling point, embodied in the form of ENU coordinates; is the observation value of the i-th sampling point, taking the value of the mean of the received power measurements at the i-th sampling point; is the variance of the received power measurements at the i-th sampling point, i = 1, 2, …, n, n representing the number of sampling points.

[0051] S3. The average fading trend of signal with distance and height is fitted using the log-distance model, and the residual is modeled using the Gaussian process to obtain the prediction model of the mean and variance of the received signal power at the sampling points;

[0052] S301. The average received power fading trend of all sampling points is fitted using the log-distance fading model and the height correction function to obtain a vector consisting of the average received power, the path loss exponent and the height correction coefficient;

[0053] The true value of the parameter vector cannot be directly measured in the actual environment, and the estimated value of the parameter vector needs to be obtained by performing least squares or weighted least squares fitting on the collected data :

[0054] ;

[0055] wherein, ;

[0056] is the design matrix, the i-th row of which is composed of the feature row vector , wherein:

[0057] , i = 1, 2, …, n

[0058] S302. , define a prior model of Gaussian process to describe the spatial distribution characteristics of the residual in the signal field, the residual refers to the part of the observation value that cannot be explained by the physical model, for any two sampling points , the kernel function is:

[0059]

[0060] wherein, is the correlation length scale, the difference variance scale;

[0061] Construct the covariance matrix , denotes the vector composed of the positions of the sampling points, the matrix The element in the ith row and jth column of the matrix ; denotes the position of the ith sampling point and the position of the jth sampling point;

[0062] Calculate the residual of the ith sampling point:

[0063]

[0064] Build the residual vector R:

[0065]

[0066] S303. Let the position of the sampling point to be predicted be , give the prediction model of the mean and variance of the received signal power:

[0067] The predicted mean of the received signal power is:

[0068]

[0069] wherein,

[0070]

[0071] X denotes the vector composed of the positions of the training samples, containing the ENU coordinates of all the sampled points; denotes the covariance vector between the test point and all the training points, the ith item in the covariance vector is , i=1,2,…,n;

[0072] The predicted variance of the received signal power is:

[0073] .

[0074] S4. Calculate information gain, and combine with flight energy or time cost, form comprehensive utility function, generate candidate point set in flyable area;

[0075] S401. Calculate single point information gain:

[0076] For any sampling point in the flyable area, according to its position , according to step S3, the received signal power variance prediction is carried out, and the prediction result is recorded as , and combined with the measurement noise corresponding to the position of the sampling point , the single point information gain is calculated:

[0077] S402. Comprehensive utility function: combine the cost of reaching the point Form the utility ratio:

[0078]

[0079] Wherein is the unit energy consumption or time cost of moving from the current position to x and sampling;

[0080] S403. Candidate point generation: set a threshold value for the utility ratio of the flyable area, select the points in the flyable area whose utility ratio is greater than the set threshold value, and generate a candidate point set.

[0081] The present application distributes the points or adaptively refines according to , and outputs: candidate point set ; According to the prediction variance given by the GP model and the measurement noise estimation, the expected information gain of each position is calculated, and the comprehensive utility is formed by combining the energy / time cost of reaching the position. The system selects the points with the highest utility in the entire flyable area and performs moderate spatial sparsification and encryption processing, and outputs the candidate point set S for path planning.

[0082] S5. Based on the candidate point set, plan the optimal sampling path under the energy and time constraints.

[0083] S501. From the candidate point set, select a group of points to generate a path p;

[0084] Establish 0 / 1 selection variable , to maximize the information benefit to build the objective function:

[0085]

[0086] The goal is to select a group from the candidate points so that the sum of the information benefits is maximum; wherein, is the jth point in the path p, is a selection variable for the jth point in path p, = 0 means not visiting the jth point in path p, = 1 means visiting the jth point in path p;

[0087] Constraints:

[0088]

[0089] where the trajectory energy cost of path p :

[0090]

[0091] is the energy cost model coefficient, is the maximum trajectory energy cost constraint, represents the flight time of path p, is the maximum flight time constraint, , represents the speed and acceleration at time t;

[0092] Curvature constraint: The curvature of the trajectory cannot exceed the maximum maneuvering capability of the UAV ;

[0093] S502. Obtain the path of the sampling point, i.e., the planned trajectory, by solving the objective function constructed to maximize the information gain under the constraints in step S501.

[0094] In an embodiment of the present application, a model updating and re-planning step S6 is further included:

[0095] The UAV starts to execute the planned path, obtains new observations when passing through the sampling points, and updates the Gaussian process model in real time; after updating, if there are still areas with high uncertainty, sudden interference, or insufficient energy / time, re-planning is performed according to steps S4-S5.

[0096] The step S6 includes:

[0097] S601. The UAV flies according to the path and stays at the sampling points;

[0098] S602. New observations are added to the data set D;

[0099] S603. Update the GP model to obtain a new ;

[0100] S604. Check the trigger condition:

[0101] The global uncertainty is still large;

[0102] local disturbance surge;

[0103] insufficient power / time;

[0104] If triggered, return to step S4 to step S5 to re-plan;

[0105] S605. Rolling update of the trajectory and prediction map.

[0106] After completing the UAV sampling and updating the Gaussian process model, the system performs real-time checks on the update results to determine whether to enter a new round of path planning. The specific logic is as follows:

[0107] 1. Global uncertainty determination

[0108] The system calculates the average uncertainty by statistically analyzing the prediction variance or entropy values of all sampled regions in the current prediction field. If the average uncertainty is still higher than the preset threshold, it indicates that the current sampling is insufficient to accurately model the entire survey area, triggering the generation of new candidate points and path supplementation.

[0109] 2. Local disturbance surge determination

[0110] By comparing the latest observation values with the model prediction values, if the residual exceeds a certain multiple of the noise standard deviation (e.g., 3σ), and such anomalies appear continuously in space, the system considers that new disturbance hotspots or shielding effects have been discovered, and preferentially generates dense candidate points near the area for supplementary measurement.

[0111] 3. Energy and time determination

[0112] Real-time monitoring of the remaining power of the UAV and the time flown, if it is predicted that the current path cannot be completed or cannot be safely returned, the subsequent point sampling is aborted, and a return path or simplified subsequent trajectory is immediately generated to ensure task safety.

[0113] 4. Re-planning trigger

[0114] If any condition is met, the system will call steps S4 and S5: recalculate the global information gain distribution, generate a new candidate point set and optimize the path, splice into a new remaining flight plan, replace the current unexecuted path segment, and realize the rolling planning of MPC (Model Predictive Control).

[0115] The trajectory planning method further comprises:

[0116] When the model prediction uncertainty falls below the threshold, or the UAV energy and task time reach the upper limit, it is automatically terminated; the final output includes: low-altitude signal field prediction map, uncertainty distribution map, and UAV flight trajectory and observation log;

[0117] In particular, the stopping condition comprises:

[0118] Global average variance:

[0119]

[0120] Interference hotspot residual variance:

[0121]

[0122] Resource depletion:

[0123] ;

[0124] The final output comprises:

[0125] Communication signal field prediction map ;

[0126] Uncertainty distribution map ;

[0127] UAV flight trajectory and observation log.

[0128] In summary, the present application concentrates limited flight resources (time, energy) into the area with the most complex field changes and the highest information value, avoiding redundant sampling in flat areas, and greatly improving overall sampling efficiency. The present application can actively discover and encrypt abnormal points and dramatic change boundaries in the sampling field through iterative "leakage and deficiency filling", and the finally reconstructed field model has higher accuracy, more details and fewer artifacts. The present application does not rely on prior knowledge of the field, and has good adaptability to various unknown or complex field distributions, and can "adapt to local conditions" for intelligent detection. The present application not only provides a high-precision field distribution map, but also through the final uncertainty map, the user can clearly understand the reliability of the predicted value at each position in the entire region.

[0129] A sensor trajectory planning system based on an adaptive field model, comprising:

[0130] An unmanned aerial vehicle flight platform for carrying sensors and communication modules to perform low-altitude flight tasks;

[0131] A positioning and attitude perception module provides centimeter-level positioning through RTK-GPS and provides attitude and speed information through IMU for spatial alignment and track control;

[0132] A communication signal acquisition module for receiving target frequency band signals, extracting received signal power, and an uncertain observation set;

[0133] The field modeling and decision module, embedded computer / ground station, is responsible for establishing the average fading trend of the signal with distance and height by fitting the log-distance model, and then modeling the residual by Gaussian process to obtain the prediction model of the mean and variance of the received signal power of the sampling points.

[0134] The path planning and control module generates a sampling path by trajectory planning based on information gain and energy consumption constraints, and drives the UAV to perform flight tasks.

[0135] In the embodiments of the present application, the field modeling and decision module is also responsible for real-time updating of the Gaussian process model and re-planning of the trajectory.

[0136] In the embodiments of the present application, the system further comprises a data management and output module for storing the sampling trajectory, observation data, generating a final communication field map and an uncertainty distribution map.

[0137] The above description shows and describes one preferred embodiment of the present application, but as previously described, it should be understood that the present application is not limited to the form disclosed herein, should not be considered as excluding other embodiments, and can be used in various other combinations, modifications and environments, and can be modified within the scope of the inventive concept described herein by the above teaching or related art or knowledge. Any modification and change made by those skilled in the art without departing from the spirit and scope of the present application shall be within the protection scope of the appended claims of the present application.

Claims

1. A sensor trajectory planning method based on an adaptive field model, characterized in that: Includes the following steps: S1. Convert the raw positioning data collected by the UAV into ECEF, then into ENU coordinates, and interpolate and synchronize the timestamps of RTK, IMU and received signal power to achieve time alignment of the data; S2. Construct an uncertainty model for the observed values ​​of the received signal power. When the UAV hovers or passes over a designated sampling point, it records the received signal power and constructs an observation set representing the uncertainty of the sampling point. Step S2 includes: S201. The uncertainty model for the received power is constructed as follows: ； in: Representing the logarithmic distance fading and altitude correction function: ； in, , representing the parameter vector consisting of reference power, path loss exponent, and height correction factor; Reference power, representing the average received power at a reference distance; The path loss exponent controls the rate at which the signal decays with distance. This is a height correction factor used to compensate for the impact of UAV height on signal strength; The x-axis represents the height of the position from the reference point. The horizontal distance between the location and the reference point; This represents the residual term, used to capture shadow fading and interference; The measurement noise is represented by the variance of the received power measurement results; S202. At each UAV sampling point, perform multiple received power measurements to obtain the observation set. ,in, The position of the i-th sample point is represented in ENU coordinates. Let be the observed value at the i-th sampling point, and let be the mean value of the received power measurement results at the i-th sampling point. Let be the variance of the received power measurement result at the i-th sampling point, where i = 1, 2, ..., n, and n represents the number of sampling points; S3. The average fading trend of the signal with distance and altitude is fitted using a logarithmic distance model, and the residual is modeled using a Gaussian process to obtain a prediction model for the mean and variance of the received signal power at the sampling point. Step S3 includes: S301. Utilizing the logarithmic distance fading model and height correction function By fitting the average received power fading trend of all sampling points, a parameter vector consisting of average received power, path loss exponent, and height correction coefficient is obtained. ; The collected data are fitted with least squares or weighted least squares to obtain the estimated values ​​of the parameter vector. : ； in, ; To design a matrix, its i-th row is composed of eigenrow vectors. Composition, in which: ,i=1,2,…,n S302. Define a prior model of a Gaussian process to describe the spatial distribution characteristics of the residuals in the signal field, for any two sampling points. kernel function for: ； in, It is a relevant length scale. Variance metric; Construct the covariance matrix , A vector or matrix representing the positions of the sampling points. The element in the i-th row and j-th column ; This indicates the positions of the i-th and j-th sampling points; Calculate the residual at the i-th sampling point: ； Construct the residual vector R: ； S303. Let the location of the sampling point to be predicted be... ; The predicted mean of the received signal power is: ； in, ； X represents a vector consisting of the positions of the training samples, containing the ENU coordinates of all sampled points; This represents the covariance vector between the test point and all training points, where the i-th term in the covariance vector is... , i=1,2,…,n; The prediction variance of the received signal power is: ； S4. Based on the prediction results of the received signal power variance, calculate the information gain of any sampling point within the flightable area, and combine it with flight energy consumption or time cost to form a comprehensive utility function. Based on the comprehensive utility function, generate a candidate point set within the flightable area. S5. Based on the candidate point set, plan the optimal sampling path under energy and time constraints.

2. The sensor trajectory planning method based on an adaptive field model according to claim 1, characterized in that: Step S1 includes: S101. Given the geodetic coordinates of the reference point: And the drone's RTK geodetic coordinates: ;in, These represent the longitude, latitude, and ellipse height of the reference point, respectively. These represent the longitude, latitude, and elliptic altitude of the UAV, respectively. At a given latitude Next, determine the radius of curvature of the zonal loop. The calculation formula is as follows: ； in, The length of the Earth's semi-major axis, in meters; Represents the square of the first eccentricity. , This indicates the oblateness of the Earth as an ellipsoid. ; S102. Transformation formula from a given reference point geodetic coordinate system and the UAV RTK geodetic coordinate system to the ECEF coordinate system: ； This refers to the coordinates obtained by transforming to the ECEF coordinate system; the ECEF coordinate system is the geocentric rectangular coordinate system. when Pick , obtained The transformation result from the reference point geodetic coordinate system to the ECEF coordinate system is denoted as... =( , ); when Pick , obtained The transformation result from the UAV RTK geodetic coordinate system to the ECEF coordinate system is denoted as... ; S103. Transform the ECEF coordinate system to the local ENU coordinate system: Calculate the relative displacement between the UAV and the reference point in the ECEF coordinate system. : ； The rotation matrix to transform the reference point ECEF coordinate system to the ENU coordinate system with the reference point as the origin is: ； Calculate the position of the UAV in the ENU coordinate system with the reference point as the origin. : ； S104. Time Alignment: The timestamps of RTK, IMU, and received signal power are interpolated to a unified time t using the UAV master clock to ensure time series consistency, where UAV refers to unmanned aerial vehicles.

3. The sensor trajectory planning method based on an adaptive field model according to claim 2, characterized in that: Step S4 includes: S401. Calculate single-point information gain: For any sampling point within the flyable area, based on its location Perform received signal power variance prediction according to step S3, and record the prediction result as... And combined with the measurement noise corresponding to the sampling point location Calculate single-point information gain : ; S402. Comprehensive Utility Function: Incorporating the cost of reaching that point. Formation utility ratio: ； in The energy or time cost per unit of time to move from the current position to x and sample; S403. Candidate Point Generation: Set a threshold for the utility ratio within the flyable area, and select the utility ratio within the flyable area. Points exceeding a set threshold are used to generate a candidate point set.

4. The sensor trajectory planning method based on an adaptive field model according to claim 3, characterized in that: Step S5 includes: S501. Select a set of points from the candidate point set and generate path p; Establish 0 / 1 selection variables To maximize information gain, the objective function is constructed as follows: ； The objective is to select a set of candidate points that maximizes the sum of their information gains; where, Let j be the j-th point in path p. Let j be the selection variable for the j-th point in path p. =0 means that the j-th point in path p is not visited. =1 indicates that the j-th point in path p is visited; Constraints: ； Among them, the trajectory energy consumption of path p : ； These are the coefficients for the energy consumption model. As a constraint on the maximum energy consumption of the trajectory, This represents the flight time of path p. As a constraint on the maximum flight time, , Represents the velocity and acceleration at time t; Curvature constraint: The curvature of the trajectory must not exceed the maximum maneuverability of the UAV. ; S502. Under the constraints of step S501, solve the objective function constructed to maximize information gain to obtain the path of the sampling points, i.e. the planned trajectory.

5. A sensor trajectory planning system based on an adaptive field model, wherein the method described in any one of claims 1 to 4 is characterized in that: include: Unmanned aerial vehicle (UAV) flight platforms are used to carry sensors and communication modules to perform low-altitude flight missions. The positioning and attitude perception module provides centimeter-level positioning via RTK-GPS and attitude and velocity information via IMU for spatial alignment and trajectory control. The communication signal acquisition module is used to receive signals in the target frequency band, extract the received signal power, and obtain an uncertain set of observations; The field modeling and decision module, an embedded computer / ground station, is responsible for establishing a predictive model of the mean and variance of the received signal power at the sampling points by fitting the average fading trend of the signal with distance and altitude using a logarithmic distance model, and then using a Gaussian process to model the residuals. The path planning and control module, based on information gain and energy consumption constraints, performs trajectory planning to generate sampling paths and drives the UAV to perform flight missions.

Images