Robot autonomous sampling method and system based on global and local information

Abstract

The invention discloses a robot autonomous sampling method and system based on global and local information, and relates to the technical field of intelligent robots. The method comprises the following steps: establishing a Gaussian process model based on spatial variation and correlation of environment variables; the uncertainty of the sampling positions is used as a quantitative index of a global target sampling position, the uncertainty of each sampling position is evaluated based on a Gaussian process model, and a global optimal target sampling position is obtained; determining a local autonomous sampling decision based on the global optimal target sampling position, and performing local sampling according to the local autonomous sampling decision to obtain a local sampling position; and sampling the local sampling position to obtain an observation value. According to the invention, a sampling decision and planning algorithm based on global and local information is constructed, a sampling process based on a statistical model is optimized, and dynamic optimization and adjustment of robot sampling decision and planning are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent robots, and in particular to a robot autonomous sampling method and system based on global and local information. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] The robot uses its mobile perception ability to measure the target environment parameters through sensors to complete the monitoring of the surrounding environment, and can further use the acquired data to reconstruct the complete environmental space. In related research, the target environment is usually described as a spatial scalar field, in which the robot completes autonomous sampling planning and reconstructs the scalar field of the entire environmental space based on the sampled observations, that is, the target environmental field.

[0004] If every location in the monitoring area is sampled, the monitoring task will be limited by the robot's operating time and resources. Traditional sampling methods generally use sampling theory for uniform or random sampling, but this will collect a lot of unnecessary measurement data and waste perception resources. In addition, although mobile robots can sample the environment regularly or randomly, this cannot deeply mine the effective information contained in the environmental variables, resulting in the inability to fully utilize sampling resources to maximize the description and reconstruction of the target environment field. Therefore, the robot's ability to make effective autonomous sampling decisions and planning in the environment is the core part of the robot's unmanned environmental monitoring.

[0005] Autonomous sampling by robots requires mining key information in the environmental model to determine the optimal target sampling point. Key information is generally used to determine the objective function through mathematical statistics methods. Common methods include mutual information, information entropy and their corresponding improved methods. Therefore, the autonomous sampling and scheduling capabilities of robots in the environmental field are the key to achieving efficient and reliable environmental monitoring.

[0006] However, the existing robot autonomous sampling methods, sampling decisions and path planning are not reasonable enough, resulting in a large error between the determined optimal target sampling point and the actual optimal target sampling point, which leads to the inability to establish an accurate environmental field. Therefore, how to achieve more accurate robot autonomous sampling has become an urgent problem to be solved in the existing technology. Summary of the invention

[0007] In view of the shortcomings of the prior art, the purpose of the present invention is to provide a robot autonomous sampling method and system based on global and local information, construct a sampling decision and planning algorithm based on global and local information, optimize the sampling process based on statistical models, and be able to combine historical data with newly acquired data to achieve dynamic optimization and adjustment of robot sampling decisions and planning, thereby ensuring accurate and efficient execution of sampling tasks.

[0008] In order to achieve the above object, the present invention is implemented through the following technical solutions: A first aspect of the present invention provides a robot autonomous sampling method based on global and local information, comprising the following steps: Get the robot's initial position and sampling position set; Gaussian process models were established based on the spatial variation and correlation of environmental variables; The uncertainty of the sampling position is used as a quantitative indicator of the global target sampling position. The uncertainty of each sampling position is evaluated based on the Gaussian process model to obtain the global optimal target sampling position. Determine a local autonomous sampling decision based on the global optimal target sampling position, perform local sampling according to the local autonomous sampling decision, and obtain a local sampling position; Sample the local sampling position to obtain the observed value.

[0009] Furthermore, the specific steps of establishing a Gaussian process model based on the spatial variation and correlation of environmental variables are as follows: Define random variables to describe the spatial variation and correlation of environmental variables; Construct a Gaussian process model consisting of a set of random variables.

[0010] Furthermore, the greater the uncertainty, the greater the information gain, and the position with the largest uncertainty is calculated as the global optimal target sampling position.

[0011] Furthermore, the specific steps to evaluate the uncertainty of each sampling location are: According to the Gaussian process model, the variance of the estimated variable value at any sampling location in the unsampled area is estimated using the Gaussian process fitting method; The moving distance in the global sampling path planning process is considered, and the uncertainty is calculated in combination with the variance of the variable estimates.

[0012] Furthermore, the robot performs local sampling tasks while moving to the global optimal position.

[0013] Furthermore, the specific steps for determining the local autonomous sampling decision based on the global optimal target sampling position are: Obtain the robot's global motion direction based on the global optimal target sampling position; A sector of a preset radius within the robot's global motion direction is defined as an evaluation area; The information gain of all potential sampling points in the evaluation area is evaluated to obtain the local sampling locations.

[0014] Furthermore, the environmental field is reconstructed by a Gaussian process fitting method based on the observations obtained at the local sampling positions.

[0015] A second aspect of the present invention provides a robot autonomous sampling system based on global and local information, comprising: A data acquisition module is configured to obtain an initial position of the robot and a set of sampling positions; a model building module configured to build a Gaussian process model based on the spatial variation and correlation of environmental variables; A global sampling module is configured to use the uncertainty of the sampling position as a quantitative indicator of the global target sampling position, evaluate the uncertainty of each sampling position based on the Gaussian process model, and obtain the global optimal target sampling position; A local sampling module is configured to determine a local autonomous sampling decision based on a global optimal target sampling position, perform local sampling according to the local autonomous sampling decision, and obtain a local sampling position; The observation value calculation module is configured to sample the local sampling position to obtain the observation value.

[0016] The third aspect of the present invention provides a medium having a program stored thereon, which, when executed by a processor, implements the steps of the robot autonomous sampling method based on global and local information as described in the first aspect of the present invention.

[0017] The fourth aspect of the present invention provides a device, including a memory, a processor, and a program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the robot autonomous sampling method based on global and local information as described in the first aspect of the present invention are implemented.

[0018] One or more of the above technical solutions have the following beneficial effects: The present invention discloses a robot autonomous sampling method and system based on global and local information. For environmental monitoring tasks, the mobile robot uses environmental variables (such as temperature) measured by sensors, and uses sampled observations to estimate the spatial numerical distribution of the entire target environmental field, thereby realizing environmental scalar field reconstruction. The robot generally uses key information in the target monitoring environment to determine the target sampling position, and the key information is usually determined by statistical information based on the environmental field. This patent proposes a robot autonomous sampling decision and path planning method based on global-local information. The method describes the environmental space through a Gaussian Process (GP) model and guides the robot to achieve autonomous sampling through multivariate conditional mutual information (MCMI), and finally completes the environmental space field reconstruction. The experimental results show the effectiveness and reliability of the proposed method in robot autonomous sampling and environmental field reconstruction.

[0019] The present invention proposes a global and local information driven robot autonomous sampling decision and planning framework for environmental spatial field sampling and reconstruction.

[0020] We also designed a global optimal sampling position decision based on uncertainty measurement for the robot's global autonomous sampling task, and a local optimal sampling position decision based on multivariate conditional mutual information measurement for the robot's local autonomous sampling task.

[0021] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0023] Figure 1 This is a flow chart of a robot autonomous sampling method based on global and local information in Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of local autonomous sampling decision-making in Embodiment 1 of the present invention; Figure 3 This is a schematic diagram of the experimental setup of the temperature map surveying and mapping robot in the first embodiment of the present invention; Figure 4 Schematic diagram of map comparison using four comparison methods and the method proposed in the present invention. DETAILED DESCRIPTION

[0024] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0025] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "include" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or their combinations; Embodiment 1: Embodiment 1 of the present invention provides a robot autonomous sampling method based on global and local information, such as Figure 1 As shown, the following steps are included: Step 1: Get the robot's initial position and the sampling location set .

[0026] The sampling position set is a collection of all sampling positions in the target space, and the position coordinates can be obtained by spatial gridding.

[0027] Step 2: Establish a Gaussian process model based on the spatial variation and correlation of environmental variables.

[0028] Step 2.1: Define random variables to describe the spatial variation and correlation of environmental variables.

[0029] For environmental monitoring, the environmental field is the target object for robot sampling and reconstruction. In spatial statistics, the changes in environmental characteristics are usually described by random fields defined by random variables. In order to effectively describe the statistical characteristics of the environmental field, the present invention adopts a Gaussian process model to describe the spatial changes and correlations of environmental variables.

[0030] Step 2.2: Construct a Gaussian process model consisting of a set of random variables.

[0031] A set of Gaussian random variables can represent the target Gaussian random field, and the established random field can be expressed as: , .

[0032] in, represents the random field observation, Indicates spatial location The corresponding random variable is represents the target environment space, and n represents the number of sampling locations. represents the established Gaussian process model, is the mean matrix used to describe the spatial average trend, is a covariance matrix that describes the spatial correlation at different locations in the ambient field.

[0033] For Gaussian process models, a common criterion for determining key sampling locations is the ability to obtain more information. Mutual information has been widely used to measure the interdependence between observed and unobserved locations. Given a covariance matrix of a Gaussian process model, the mutual information corresponding to different locations in the environment field can be estimated, thereby measuring the information gain of different sampling locations. The location with the maximum information gain in the environment field can be obtained through the information gain evaluation index, and this location can be designated as the best sampling location in the environment field.

[0034] Step 3: Use the uncertainty of the sampling position as a quantitative indicator of the global target sampling position, evaluate the uncertainty of each sampling position based on the Gaussian process model, and obtain the global optimal target sampling position.

[0035] In a specific implementation, the uncertainty of each unsampled position is estimated based on the sampling position, which can be used as a metric to evaluate the next global target sampling position. In theory, a sampling position with greater uncertainty can bring more information gain, and the greater the uncertainty, the more information gain. The position with the largest uncertainty is calculated as the global optimal target sampling position.

[0036] The specific steps to evaluate the uncertainty at each sampling location are: Step 3.1: Based on the Gaussian process model, use the Gaussian process fitting method to estimate the variance of the variable estimate at any sampling location in the unsampled area.

[0037] Gaussian process model for target environment field , environmental variables at different spatial locations can be represented by high-dimensional random variables, which can be used to design the global optimal sampling location. The statistical characteristics of the random field can be obtained by Gaussian Process Regression (GPR). Given a set of sampled points , and its corresponding observed value , the Gaussian process fitting method can be used to estimate the unsampled area Any sampling position The estimated variance of the variable : .

[0038] in, represents the attribute matrix, represents the correlation between the unsampled locations and other locations, - , Represents variance. , M represents the attribute space transformation matrix. The above formula integrates spatial correlation and attribute information to calculate the estimated uncertainty of unsampled locations.

[0039] Step 3.2: Consider the moving distance during the global sampling path planning process and calculate the uncertainty in combination with the variance of the variable estimates.

[0040] At the same time, considering time and resource efficiency, the global sampling path planning should also consider the global moving distance of the robot, so that the robot has less moving distance when performing global autonomous sampling. Therefore, this embodiment proposes a quantitative indicator based on uncertainty to determine the global autonomous sampling position: For any unsampled position , the proposed quantitative indicators for: . in, Indicates unsampled locations The Euclidean distance from the robot's current position.

[0041] In unsampled areas Any position in By calculating the above index, we can obtain the global optimal target sampling position. : .

[0042] Step 4: Determine the local autonomous sampling decision based on the global optimal target sampling position, perform local sampling according to the local autonomous sampling decision, and obtain the local sampling position.

[0043] In a specific implementation, after determining the global sampling position, the robot performs local sampling tasks while moving to the global optimal position, thereby obtaining more observation values ​​for environmental field estimation, thereby enhancing its environmental field estimation capability and improving the accuracy of environmental field reconstruction. Therefore, this embodiment further designs a local autonomous sampling method for local autonomous sampling decision-making and planning.

[0044] The method first defines a specific sector within the robot's global motion direction as an evaluation region. Then, the information gain of all potential sampling points in the region is evaluated to select local target sampling points to obtain more environmental information. Local autonomous sampling selects the position that can bring the maximum information gain, that is, the position with the largest amount of information in a given local region. Then, the selected position is determined as the robot's next local autonomous sampling target position.

[0045] The specific steps for determining the local autonomous sampling decision based on the global optimal target sampling position are: Step 4.1: Obtain the robot's global motion direction based on the global optimal target sampling position.

[0046] Figure 2 The proposed local autonomous sampling decision and planning strategy is presented. First, the evaluation area needs to be specified for the robot's next local sampling. The current position is represented as , the global target sampling position is expressed as By calculating the angle between these two positions , the direction of travel can be obtained.

[0047] Step 4.2: Define a sector of a preset radius within the robot's global motion direction as the evaluation area.

[0048] Step 4.2.1: Figure 2 As shown, create a The fan-shaped area is centered on . By Increase or decrease specific angles based on , the angular range of the evaluation area can be determined, that is, In the angle range of the fan-shaped area, a series of lengths are evenly created. Vector By adding the vector to the current position, we can get some position coordinates These locations and the current location constitute the area to be evaluated for local sampling decisions and planning.

[0049] Step 4.2.2: Finally, check whether these local sampling positions are within the set environmental field area to ensure their validity. At the same time, it is necessary to delete the positions that have been sampled to avoid repeated and invalid sampling. Relatively fixed, the distance from the current sampling position in the fan-shaped area is The position of is set as the local sampling candidate position. Finally, the local sampling evaluation area can be obtained and candidate locations .

[0050] Step 4.3: Evaluate the information gain of all potential sampling points in the evaluation area to obtain the local sampling locations.

[0051] Step 4.3.1: After obtaining the candidate locations for local sampling, determine which of these locations can provide the most information gain. Based on historical observations, use multivariate conditional mutual information at the current location as an evaluation indicator to determine the local optimal sampling location. The calculation formula for multivariate conditional mutual information is as follows: .

[0052] in, represents mutual information, () represents multivariate conditional mutual information, represents the environmental field scalar of the evaluation area calculated based on the Gaussian process regression model, A random variable representing the candidate position currently being evaluated, Indicates except The remaining random variables outside the evaluation area are is the observed value of the robot's current position, that is, the sampled position.

[0053] Step 4.3.2: The calculation formula of multivariate conditional mutual information can be further derived and converted into the form of entropy, where the first term can be derived as: .

[0054] The second term can be derived as: .

[0055] Step 4.3.3: Calculate the multivariate conditional mutual information of each candidate position as a criterion for evaluating its importance.

[0056] In a specific embodiment, a larger multivariate conditional mutual information indicates that sampling at this location can provide more information for the environmental field estimation. Therefore, the local sampling location can be determined as follows : .

[0057] Step 5: Sample the local sampling position to obtain the observation value, and reconstruct the environmental field through the Gaussian process fitting method based on the observation value obtained at the local sampling position.

[0058] In a specific embodiment, given an observation value, any position The environment variable value can be estimated as: .

[0059] in, Represents the value of an environment variable. represents the attribute matrix, represents the least squares solution matrix, which can be obtained by the generalized least squares method in the Gaussian process fitting method.

[0060] This paper proposes a robot autonomous sampling decision and planning method based on global-local information for active robot sampling and environmental field reconstruction. A global autonomous sampling method based on global uncertainty and a local autonomous sampling method based on local multivariate conditional mutual information are designed to guide the robot to perceive the environment.

[0061] In order to better illustrate the superiority of the method of this embodiment, a robot test experiment was conducted, and the autonomous sampling and environment reconstruction experimental results were obtained.

[0062] To verify the performance, the proposed information path planner is compared with four advanced path planning strategies in the experiment. The comparison methods in the experiment are as follows: Local random sampling method: Randomly select sampling locations for local sampling.

[0063] Local maximum mutual information sampling method: Using mutual information as an indicator, the sampling position with the highest mutual information value is selected locally to form a sampling path.

[0064] Global maximum mutual information sampling method: Using mutual information as an indicator, generate the global optimization positions with the highest mutual information value in the world and connect them as local sampling paths.

[0065] Hierarchical mutual information sampling method: The optimal sensing location is selected based on the global mutual information while running the local mutual information greedy path planning.

[0066] The reconstruction accuracy of the environment field is evaluated by four indicators: mean absolute error (MAE), mean square error (MSE), average reconstruction variance (ARV) and maximum reconstruction variance (MRV), which are defined as follows: , , , .

[0067] in, The following experimental test evaluates the performance of the method of this embodiment and gives specific experimental results.

[0068] Figure 3 The experimental scene setting is shown. The mobile sensing robot is used to conduct experiments to implement the proposed information path planner. The dotted rectangular area represents the target area for verifying the method of this embodiment. The robot uses a temperature sensor to measure the ambient temperature of the monitored area. A thermal imager is installed to capture the temperature map of the environmental field as the true value. Every 5 minutes is set as a monitoring cycle for information path planning and environmental field mapping. A total of 10 sets of data are collected in the experiment to verify the average performance of the proposed method.

[0069] The experimental results obtained by averaging the four indicators for 10 periods are given in Table 1.

[0070] Table 1 Experimental results.

[0071]

[0072] As can be seen from the table, compared with other benchmark methods, the method of this embodiment performs well in MAE, MSE, ARV and MRV, and the environmental reconstruction performance is the most outstanding. This demonstrates the excellent ability of the proposed global-local information path planner and its high accuracy in active sensing and environmental field mapping. Therefore, the method of this embodiment can better guide the mobile robot to perform active perception and reconstruct the target environmental field.

[0073] A real test example of experimental results is as follows Figure 3 In this monitoring cycle, the true value of the environment field reconstructed by the target is as follows: Figure 3 As shown, the temperature distribution map is captured and calibrated by a thermal imager. Figure 4The maps reconstructed using four comparison methods and the method proposed in this embodiment are shown, where the first row from left to right are the true value map of the experimental environment field, the local random sampling method to reconstruct the map and the local mutual information greedy method to reconstruct the map, and the second row from left to right are the global mutual information planner to reconstruct the map, the hierarchical mutual information planner to reconstruct the map and the method of this embodiment to reconstruct the map. The sampling path planned by each method is also shown in the figure. In this test, the MAE corresponding to local random sampling is 0.0643, the MSE is 0.0168, the ARV is 0.0762, and the MRV is 0.5801. The performance of the local mutual information greedy planner corresponds to a MAE of 0.1111, an MSE of 0.0375, an ARV of 0.1226, and an MRV of 0.6490. The performance of the global mutual information planner corresponds to a MAE of 0.0329, an MSE of 0.0044, an ARV of 0.0553, and an MRV of 0.3022. The results of the hierarchical mutual information planner are MAE of 0.0317, MSE of 0.0042, ARV of 0.0543, and MRV of 0.2985. Among them, the method of this embodiment performs best, with MAE of 0.0149, MSE of 0.0008, ARV of 0.0393, and MRV of 0.2309.

[0074] Experimental results show that by combining global sampling based on maximizing global uncertainty with local sampling based on maximizing local multivariate conditional mutual information, the method proposed in this embodiment can provide efficient and reliable robot autonomous sampling and environmental field reconstruction performance.

[0075] Embodiment 2: Embodiment 2 of the present invention provides a robot autonomous sampling system based on global and local information, including: A data acquisition module is configured to obtain an initial position of the robot and a set of sampling positions; a model building module configured to build a Gaussian process model based on the spatial variation and correlation of environmental variables; A global sampling module is configured to use the uncertainty of the sampling position as a quantitative indicator of the global target sampling position, evaluate the uncertainty of each sampling position based on the Gaussian process model, and obtain the global optimal target sampling position; A local sampling module is configured to determine a local autonomous sampling decision based on a global optimal target sampling position, perform local sampling according to the local autonomous sampling decision, and obtain a local sampling position; The observation value calculation module is configured to sample the local sampling position to obtain the observation value.

[0076] Embodiment three: Embodiment 3 of the present invention provides a medium on which a program is stored. When the program is executed by a processor, the steps of the robot autonomous sampling method based on global and local information as described in Embodiment 1 of the present invention are implemented.

[0077] Embodiment 4: Embodiment 4 of the present invention provides a device, including a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, the steps in the robot autonomous sampling method based on global and local information as described in Embodiment 1 of the present invention are implemented.

[0078] The steps involved in the above embodiments 2, 3 and 4 correspond to the method embodiment 1. For the specific implementation methods, please refer to the relevant description part of embodiment 1.

[0079] Those skilled in the art should understand that the modules or steps of the present invention described above can be implemented by a general-purpose computer device, or alternatively, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0080] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A robot autonomous sampling method based on global and local information, characterized in that: The following steps are involved: Get the robot's initial position and sampling position set; Gaussian process models were established based on the spatial variation and correlation of environmental variables; The uncertainty of the sampling position is used as a quantitative indicator of the global target sampling position. The uncertainty of each sampling position is evaluated based on the Gaussian process model to obtain the global optimal target sampling position. Determine a local autonomous sampling decision based on the global optimal target sampling position, perform local sampling according to the local autonomous sampling decision, and obtain a local sampling position; Sample the local sampling position to obtain the observed value.

2. The robot autonomous sampling method based on global and local information as claimed in claim 1, characterized in that: The specific steps to establish a Gaussian process model based on the spatial variation and correlation of environmental variables are: Define random variables to describe the spatial variation and correlation of environmental variables; Construct a Gaussian process model consisting of a set of random variables.

3. The robot autonomous sampling method based on global and local information as claimed in claim 1, characterized in that: The greater the uncertainty, the greater the information gain, and the position with the largest uncertainty is calculated as the global optimal target sampling position.

4. The robot autonomous sampling method based on global and local information as claimed in claim 3, characterized in that: The specific steps to evaluate the uncertainty at each sampling location are: According to the Gaussian process model, the variance of the estimated variable value at any sampling location in the unsampled area is estimated using the Gaussian process fitting method; The moving distance in the global sampling path planning process is considered, and the uncertainty is calculated in combination with the variance of the variable estimates.

5. The robot autonomous sampling method based on global and local information as claimed in claim 1, characterized in that: The robot performs local sampling tasks while moving to the global optimal position.

6. The robot autonomous sampling method based on global and local information as claimed in claim 1, characterized in that: The specific steps for determining the local autonomous sampling decision based on the global optimal target sampling position are: Obtain the robot's global motion direction based on the global optimal target sampling position; A sector of a preset radius within the robot's global motion direction is defined as an evaluation area; The information gain of all potential sampling points in the evaluation area is evaluated to obtain the local sampling locations.

7. The robot autonomous sampling method based on global and local information as claimed in claim 1, characterized in that: The environmental field is reconstructed by the Gaussian process fitting method based on the observations obtained at the local sampling positions.

8. A robot autonomous sampling system based on global and local information, characterized in that: include: A data acquisition module is configured to obtain an initial position of the robot and a set of sampling positions; a model building module configured to build a Gaussian process model based on the spatial variation and correlation of environmental variables; A global sampling module is configured to use the uncertainty of the sampling position as a quantitative indicator of the global target sampling position, evaluate the uncertainty of each sampling position based on the Gaussian process model, and obtain the global optimal target sampling position; A local sampling module is configured to determine a local autonomous sampling decision based on a global optimal target sampling position, perform local sampling according to the local autonomous sampling decision, and obtain a local sampling position; The observation value calculation module is configured to sample the local sampling position to obtain the observation value.

9. A computer-readable storage medium, characterized in that: A plurality of instructions are stored therein, and the instructions are suitable for being loaded by a processor of a terminal device and executing the robot autonomous sampling method based on global and local information as described in any one of claims 1-7.

10. A terminal device, characterized in that: It includes a processor and a computer-readable storage medium, the processor is used to implement each instruction; the computer-readable storage medium is used to store multiple instructions, and the instructions are suitable for being loaded by the processor and executing the robot autonomous sampling method based on global and local information according to any one of claims 1-7.

Images