CN-121979230-A - Unmanned aerial vehicle autonomous sampling planning method for three-dimensional space multi-environment variable prediction

CN121979230ACN 121979230 ACN121979230 ACN 121979230ACN-121979230-A

Abstract

The invention provides an unmanned aerial vehicle autonomous sampling planning method for three-dimensional space multi-environment variable prediction, which relates to the technical field of path planning and comprises the steps of generating a global target point through a global prediction variance optimal sampling method based on the current position of an unmanned aerial vehicle in a three-dimensional space and acquired multi-environment variable sample data; the method comprises the steps of constructing a conical evaluation area which takes the current position as a vertex and faces to a global target point, carrying out optimal sampling of local mutual information in the area to determine local sampling points, controlling an unmanned aerial vehicle to fly to the local sampling points to collect multi-environment variable sample data, and carrying out next iteration again until the condition of iteration stopping is met after the acquisition is completed.

Inventors

LI TENG
YANG BINGYU
WU TIANHAO
ZHAO RUKUN

Assignees

山东大学

Dates

Publication Date: 20260505
Application Date: 20251225

Claims (10)

1. The unmanned aerial vehicle autonomous sampling planning method for three-dimensional space multi-environment variable prediction is characterized by comprising the following steps of iteratively executing to form a closed-loop autonomous planning flow: generating a global target point by a global prediction variance optimal sampling method based on the current position of the unmanned aerial vehicle in the three-dimensional space and the acquired multi-environment variable sample data; constructing a conical evaluation area which takes the current position as a vertex and faces to a global target point, carrying out optimal sampling of local mutual information in the area, and determining local sampling points; Controlling the unmanned aerial vehicle to fly to the local sampling point to collect multi-environment variable sample data, and after the collection is completed, carrying out the next iteration again until the condition of iteration stopping is met; When the global target point is generated, an adaptive sampling strategy fused with the pareto optimal rule is adopted, adaptive weights are calculated according to the overall uncertainty duty ratio of the current environment variable prediction field, the multi-target utility values of each point in the pareto front edge point set are weighted and summed, and the point with the maximum comprehensive information gain after weighting is selected as the global target point.
2. The unmanned aerial vehicle autonomous sampling planning method for three-dimensional spatial multi-environment variable prediction of claim 1, wherein the prediction variance is for each environment variable, and a spatial location-based gaussian process model is built to predict the variance of that environment variable at the non-sampled point.
3. The unmanned aerial vehicle autonomous sampling planning method for three-dimensional space multi-environment variable prediction according to claim 2, wherein the prediction variance is expressed as: Wherein, the Representing non-sampled points The variance information of the self-body, Representing non-sampled points And sampled points The correlation of the cores between them, For the covariance matrix between the sample points, Representing the noise contribution in the observed data.
4. The unmanned aerial vehicle autonomous sampling planning method for three-dimensional space multi-environment variable prediction according to claim 1, wherein the generating a global target point by the global prediction variance optimal sampling method is specifically as follows: constructing an evaluation index for each environmental variable based on the prediction variance; Screening out a pareto front point set by using the pareto optimal rule and taking evaluation indexes of all non-sampled points for all environmental variables as optimization targets; and selecting a point with the maximum comprehensive information gain from the pareto front edge point set as a global target point by adopting a self-adaptive sampling strategy fused with the pareto optimal rule.
5. The unmanned aerial vehicle autonomous sampling planning method for three-dimensional space multi-environment variable prediction according to claim 4, wherein the self-adaptive selection mode of the global target point is expressed as: Wherein, the 、 The predicted variances of the two environmental variables are respectively, In order for the weights to be adaptive, 、 Respectively the evaluation indexes of two environmental variables, For the pareto front edge point set, Is the number of spatially sampable positions.
6. The unmanned aerial vehicle autonomous sampling planning method for three-dimensional space multi-environment variable prediction according to claim 1, wherein the constructing of the conical evaluation area comprises the steps of taking the current position as an apex, taking a vector towards the global target point as an axis direction, generating a discrete conical space grid in the three-dimensional space according to preset conical height and half apex angle parameters, and eliminating points located outside the boundary of the monitoring area and at sampled positions to obtain the evaluation area.
7. The unmanned aerial vehicle autonomous sampling planning method for three-dimensional space multi-environment variable prediction according to claim 1, wherein the performing local mutual information optimal sampling in the region to determine local sampling points comprises: The method comprises the steps of calculating mutual information gain values of candidate position points in an evaluation area aiming at various environment variables to form a multi-target optimization problem, screening a local pareto front edge point set by utilizing a pareto optimal rule, determining local sampling points according to a distance nearest principle, namely selecting a point nearest to Euclidean distance of the current position of the unmanned aerial vehicle from the local pareto front edge point set, and determining the local sampling points.
8. A computer program product comprising a computer program, characterized in that the computer program, when executed by a processor, implements the unmanned aerial vehicle autonomous sampling planning method for three-dimensional space multi-environment variable prediction according to any of claims 1 to 7.
9. A non-transitory computer readable storage medium for storing computer instructions which, when executed by a processor, implement the unmanned aerial vehicle autonomous sampling planning method for three-dimensional space multi-environment variable prediction of any of claims 1-7.
10. An electronic device comprising a processor, a memory and a computer program, wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory, so that the electronic device executes the unmanned aerial vehicle autonomous sampling planning method for three-dimensional space multi-environment variable prediction according to any one of claims 1 to 7.

Description

Unmanned aerial vehicle autonomous sampling planning method for three-dimensional space multi-environment variable prediction Technical Field The invention relates to the technical field of path planning, in particular to an unmanned aerial vehicle autonomous sampling planning method for three-dimensional space multi-environment variable prediction. Background In the environment monitoring task in which the unmanned aerial vehicle participates, the environment space is a main object of unmanned aerial vehicle sampling and estimation. The environment model is a set of scalar values describing the physical world environment distribution, and is the basis for unmanned aerial vehicle sampling and predicting the environment variable value distribution (such as temperature and humidity). Environmental monitoring tasks are generally not limited to two-dimensional scenes, and in three-dimensional scenes, environmental field prediction by using an unmanned aerial vehicle has significant advantages compared with the traditional method due to large sampling difficulty. In general, the mobility of the unmanned aerial vehicle in the three-dimensional space is stronger than that of other robots (such as unmanned vehicles and unmanned dogs), so the sampling of the same area is usually completed by one unmanned aerial vehicle. However, unmanned aerial vehicles have limited maneuverability and sampling capability due to limitations of resources such as fuel, electricity, etc., and limited distribution and number of sites where sampling observations can be made. Compared with the scale of a monitoring environment, the unmanned aerial vehicle has a limited sampling range, so that the unmanned aerial vehicle needs to utilize a small amount of sampling observation data to finish estimation of overall environment information, active sensing is a mainstream unmanned aerial vehicle sampling and path planning method, the basic idea is to guide the unmanned aerial vehicle to finish an observation task by analyzing the sampling position with the most information in a target environment field space, active sensing is the core of the unmanned aerial vehicle environment monitoring task, an information driven active sampling strategy can remarkably improve the value of the sampling data to environment estimation, and therefore, the research on an efficient unmanned aerial vehicle autonomous sampling planning algorithm plays an important role in improving the efficiency and accuracy of environment estimation. For the prediction tasks of different environmental field variables in the same area, the traditional algorithm requires that the unmanned aerial vehicle only aims at one environmental variable in each sampling, so that the sampling strategy has low efficiency and cannot consider the accuracy of environmental field prediction. Disclosure of Invention In order to solve the problems, the invention provides an unmanned aerial vehicle autonomous sampling planning method for three-dimensional space multi-environment variable prediction, and adopts a hierarchical information path planning strategy to enable the unmanned aerial vehicle to efficiently plan a high-quality sampling track of a three-dimensional multivariable field, thereby improving the accuracy and the effectiveness of three-dimensional space multi-environment variable prediction. According to some embodiments, the present invention employs the following technical solutions: the unmanned aerial vehicle autonomous sampling planning method for three-dimensional space multi-environment variable prediction comprises the following steps of iteratively executing to form a closed-loop autonomous planning flow: generating a global target point by a global prediction variance optimal sampling method based on the current position of the unmanned aerial vehicle in the three-dimensional space and the acquired multi-environment variable sample data; constructing a conical evaluation area which takes the current position as a vertex and faces to a global target point, carrying out optimal sampling of local mutual information in the area, and determining local sampling points; Controlling the unmanned aerial vehicle to fly to the local sampling point to collect multi-environment variable sample data, and after the collection is completed, carrying out the next iteration again until the condition of iteration stopping is met; When the global target point is generated, an adaptive sampling strategy fused with the pareto optimal rule is adopted, adaptive weights are calculated according to the overall uncertainty duty ratio of the current environment variable prediction field, the multi-target utility values of each point in the pareto front edge point set are weighted and summed, and the point with the maximum comprehensive information gain after weighting is selected as the global target point. According to some embodiments, the present invention employs the following technical solutions: a c