Multi-agent collaborative path planning decision-making method and device

Abstract

The present application relates to a multi-agent collaborative path planning decision-making method and device, wherein the method includes: determining the sensor viewing angles of multiple target agents and obtaining their environmental observation information; compressing the environmental observation information through a digital filter, obtaining environmental compression features, and reconstructing the environmental compression features to generate environmental observation reconstruction data; obtaining all target agents that meet the preset obstacle detection requirements, and sharing the environmental compression features between neighboring agents corresponding to each target agent, obtaining neighbor reconstruction data of the neighboring agent, so as to calculate the detour direction probability of each target agent and neighboring agent, so as to generate the planning paths of multiple target agents according to the detour direction probability. Thus, the existing synchronous positioning and mapping methods are solved, such as the easy generation of a large amount of high-dimensional data, the need to complete operations such as point cloud matching, which puts great pressure on system communication and computing equipment, and the difficulty of real-time transmission between multi-agent clusters.

Description

Technical Field

[0001] The present application relates to the technical field of intelligent agent collaborative path planning, and in particular to a multi-agent collaborative path planning decision-making method and device. Background Art

[0002] Performing multi-agent collaborative path planning in a complex, unknown environment without external positioning is a challenging task. One of the difficulties of this task is how to make full use of the environmental observations of the agents in the system to find a good local path in the environment and optimize the system path. This requires the robot to find a safe direction from the environmental observations in real time and efficiently complete the transmission and fusion of observations within the cluster.

[0003] Currently commonly used methods such as simultaneous positioning and mapping will generate a large amount of high-dimensional data and require the completion of point cloud matching steps, which puts great pressure on system communication and computing equipment, making it difficult to transmit in real time between multi-agent clusters, and this needs to be solved urgently. Summary of the invention

[0004] The present application provides a multi-agent collaborative path planning decision-making method and device to solve the problems that the existing synchronous positioning and mapping methods are prone to generate a large amount of high-dimensional data, and need to complete operations such as point cloud matching, which puts great pressure on system communication and computing equipment, and is difficult to perform real-time transmission between multi-agent clusters.

[0005] The first aspect of the present application provides a multi-agent collaborative path planning decision-making method, comprising the following steps: determining the sensor perspectives of multiple target agents, and obtaining environmental observation information corresponding to the sensor perspectives; compressing the environmental observation information through a preset digital filter, obtaining environmental compression features corresponding to the environmental observation information, and reconstructing the environmental compression features based on a preset polygonal approximation strategy to generate environmental observation reconstruction data; based on the environmental observation information, obtaining all target agents that meet preset obstacle detection requirements, and sharing the environmental compression features among neighbor agents corresponding to each target agent among all the target agents to obtain neighbor reconstruction data of the neighbor agents, and calculating the detour direction probabilities of each target agent and the neighbor agents through the environmental observation reconstruction data and the neighbor reconstruction data to generate planning paths for the multiple target agents according to the detour direction probabilities.

[0006] Optionally, in one embodiment of the present application, before compressing the environmental observation information through the preset digital filter, it also includes: based on a preset intelligent agent observation model, dividing the environmental observation information into multiple quadrant information; determining at least one target safety sector information in the multiple quadrant information, so as to calculate the safety sector angle according to the at least one target safety sector information; obtaining the number of environmental depth sampling points corresponding to the environmental observation information, and calculating the analog domain cutoff frequency and the digital domain cutoff frequency according to the number of environmental depth sampling points, the sensor viewing angle and the safety sector angle; determining the digital filter based on the analog domain cutoff frequency, the digital domain cutoff frequency and the preset window function.

[0007] Optionally, in one embodiment of the present application, compressing the environmental observation information through a preset digital filter, obtaining the environmental compression features corresponding to the environmental observation information, and reconstructing the environmental compression features based on a preset polygonal approximation strategy to generate environmental observation reconstruction data, includes: filtering the environmental observation information through the digital filter to generate environmental observation filter information; calculating all extreme values ​​of the environmental observation filter information, and obtaining the extreme value direction angles and extreme value distances corresponding to all extreme values, so as to obtain the environmental compression features according to the extreme value direction angles and the extreme value distances; rotating the extreme value direction angles corresponding to the environmental compression features to obtain the target extreme value direction angles; and connecting the vertices of the extreme value points corresponding to all the extreme values ​​based on the target extreme value direction angles and a preset linear analysis strategy to generate the environmental observation reconstruction data.

[0008] Optionally, in one embodiment of the present application, the detour direction probability of each target intelligent agent and the neighbor intelligent agent is calculated through the environment observation reconstruction data and the neighbor reconstruction data to generate the planned paths of the multiple target intelligent agents according to the detour direction probability, including: mapping the environment observation reconstruction data and the neighbor reconstruction data to the probability domain respectively through the detour direction probability to obtain the probability domain environment observation reconstruction data and the probability domain neighbor reconstruction data corresponding to the environment observation reconstruction data and the neighbor reconstruction data; fusing the probability domain environment observation reconstruction data and the probability domain neighbor reconstruction data to generate fused reconstruction data, and determining the fused left turn probability of the multiple target intelligent agents according to the fused reconstruction data to generate the planned paths of the multiple target intelligent agents through the fused left turn probability.

[0009] The second aspect of the present application provides a multi-agent collaborative path planning decision-making device, including: a first determination module, used to determine the sensor perspective of multiple target agents, and obtain environmental observation information corresponding to the sensor perspective; a reconstruction module, used to compress the environmental observation information through a preset digital filter, obtain the environmental compression features corresponding to the environmental observation information, and reconstruct the environmental compression features based on a preset polygonal approximation strategy to generate environmental observation reconstruction data; a decision module, used to obtain all target agents that meet the preset obstacle detection requirements based on the environmental observation information, and share the environmental compression features among the neighboring agents corresponding to each target agent among all the target agents to obtain the neighbor reconstruction data of the neighboring agents, and calculate the detour direction probability of each target agent and the neighboring agent through the environmental observation reconstruction data and the neighbor reconstruction data to generate the planned paths of the multiple target agents according to the detour direction probability.

[0010] Optionally, in one embodiment of the present application, it also includes: a division module, used to divide the environmental observation information into multiple quadrant information based on a preset intelligent agent observation model before compressing the environmental observation information through the preset digital filter; a first calculation module, used to determine at least one target safety sector information in the multiple quadrant information, so as to calculate the safety sector angle based on the at least one target safety sector information; a second calculation module, used to obtain the number of environmental depth sampling points corresponding to the environmental observation information, and calculate the analog domain cutoff frequency and the digital domain cutoff frequency based on the number of environmental depth sampling points, the sensor viewing angle and the safety sector angle; a second determination module, used to determine the digital filter based on the analog domain cutoff frequency, the digital domain cutoff frequency and a preset window function.

[0011] Optionally, in one embodiment of the present application, the reconstruction module includes: a filtering unit, used to filter the environmental observation information through the digital filter to generate environmental observation filtering information; an acquisition unit, used to calculate all extreme values ​​of the environmental observation filtering information, and obtain the extreme value direction angles and extreme value distances corresponding to the extreme values, so as to obtain the environmental compression feature according to the extreme value direction angles and the extreme value distances; a rotation unit, used to rotate the extreme value direction angles corresponding to the environmental compression feature to obtain the target extreme value direction angle; a connection unit, used to connect the vertices of the extreme value points corresponding to all the extreme values ​​based on the target extreme value direction angle and a preset linear analysis strategy to generate the environmental observation reconstruction data.

[0012] Optionally, in one embodiment of the present application, the decision module includes: a mapping unit, used to map the environment observation reconstruction data and the neighbor reconstruction data into the probability domain respectively through the detour direction probability, so as to obtain the probability domain environment observation reconstruction data and the probability domain neighbor reconstruction data corresponding to the environment observation reconstruction data and the neighbor reconstruction data; a fusion unit, used to fuse the probability domain environment observation reconstruction data and the probability domain neighbor reconstruction data to generate fused reconstruction data, and determine the fused left turn probability of the multiple target intelligent bodies according to the fused reconstruction data, so as to generate the planned paths of the multiple target intelligent bodies through the fused left turn probability.

[0013] The third aspect of the present application provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the multi-agent collaborative path planning decision method as described in the above embodiment.

[0014] The fourth aspect of the present application provides a computer-readable storage medium, which stores a computer program that, when executed by a processor, implements the above-mentioned multi-agent collaborative path planning decision-making method.

[0015] The fifth aspect of the present application provides a computer program product, including a computer program, which is executed to implement the above-mentioned multi-agent collaborative path planning decision-making method.

[0016] Therefore, the embodiments of the present application have the following beneficial effects:

[0017] The embodiment of the present application can determine the sensor perspectives of multiple target agents and obtain the environmental observation information corresponding to the sensor perspective; compress the environmental observation information through a preset digital filter, obtain the environmental compression features corresponding to the environmental observation information, and reconstruct the environmental compression features based on a preset polygonal approximation strategy to generate environmental observation reconstruction data; based on the environmental observation information, obtain all target agents that meet the preset obstacle detection requirements, and share the environmental compression features between the neighboring agents corresponding to each target agent in all target agents to obtain the neighbor reconstruction data of the neighbor agent, and calculate the detour direction probability of each target agent and the neighbor agent through the environmental observation reconstruction data and the neighbor reconstruction data, so as to generate the planning path of multiple target agents according to the detour direction probability. In the case of limited perception, no external positioning and environmental map, the present application realizes cluster observation sharing through the local environmental observation of the multi-agent cluster, thereby reducing the path cost of the crossing task. Thus, the existing synchronous positioning and mapping method is easy to generate a large amount of high-dimensional data, and needs to complete operations such as point cloud matching, which puts great pressure on system communication and computing equipment, and is difficult to perform real-time transmission between multi-agent clusters.

[0018] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0020] Figure 1 A flowchart of a multi-agent collaborative path planning decision method provided according to an embodiment of the present application;

[0021] Figure 2 A schematic diagram of an intelligent agent observation model provided for one embodiment of the present application;

[0022] Figure 3 A schematic diagram of a filter time domain and frequency domain image provided for an embodiment of the present application;

[0023] Figure 4 A schematic diagram of the logical architecture of a multi-agent collaborative path planning decision method provided for one embodiment of the present application;

[0024] Figure 5 A schematic diagram of test results of the present application in different scenarios provided for an embodiment of the present application;

[0025] Figure 6 This is an example diagram of a multi-agent collaborative path planning decision-making device according to an embodiment of the present application;

[0026] Figure 7 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application.

[0027] Among them, 10-multi-agent collaborative path planning decision-making device; 100-first determination module, 200-reconstruction module, 300-decision module; 701-memory, 702-processor, 703-communication interface. DETAILED DESCRIPTION

[0028] Embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0029] The following describes the multi-agent collaborative path planning decision method and device of the embodiment of the present application with reference to the accompanying drawings. In view of the problems mentioned in the above background technology, the present application provides a multi-agent collaborative path planning decision method, in which the sensor perspectives of multiple target agents are determined and the environmental observation information corresponding to the sensor perspectives is obtained; the environmental observation information is compressed by a preset digital filter, the environmental compression features corresponding to the environmental observation information are obtained, and the environmental compression features are reconstructed based on the preset polygonal approximation strategy to generate environmental observation reconstruction data; based on the environmental observation information, all target agents that meet the preset obstacle detection requirements are obtained, and the environmental compression features are shared between the neighboring agents corresponding to each target agent in all target agents to obtain the neighbor reconstruction data of the neighbor agent, and the detour direction probability of each target agent and the neighbor agent is calculated through the environmental observation reconstruction data and the neighbor reconstruction data, so as to generate the planning path of multiple target agents according to the detour direction probability. In the case of limited perception, no external positioning and environmental map, the present application realizes cluster observation sharing through the local environmental observation of the multi-agent cluster, thereby reducing the path cost of the crossing task. This solves the problem that existing synchronous positioning and mapping methods easily generate a large amount of high-dimensional data, require operations such as point cloud matching, put great pressure on system communications and computing equipment, and are difficult to transmit in real time between multi-agent clusters.

[0030] Specifically, Figure 1 A flowchart of a multi-agent collaborative path planning decision-making method provided in an embodiment of the present application.

[0031] like Figure 1 As shown, the multi-agent collaborative path planning decision method includes the following steps:

[0032] In step S101, the sensor perspectives of multiple target agents are determined, and environmental observation information corresponding to the sensor perspectives is obtained.

[0033] In step S102, the environment observation information is compressed by a preset digital filter to obtain the environment compression features corresponding to the environment observation information, and the environment compression features are reconstructed based on a preset polygonal approximation strategy to generate environment observation reconstruction data.

[0034] In an embodiment of the present application, the processed environmental observation information of the intelligent agent can be represented or converted into a polar coordinate signal represented by "angle-distance", represented by s(n); assuming that the viewing angle range of the sensor is Θ, s(n) can be represented as an array, and each component in the array represents the observation distance corresponding to each angle along the positive direction (counterclockwise) starting from the observation starting point.

[0035] Afterwards, the embodiment of the present application compresses the environmental observation information through a digital filter to obtain environmental compression features corresponding to the environmental observation information, and uses a polygonal approximation strategy to reconstruct the environmental compression features to generate environmental observation reconstruction data.

[0036] Optionally, in one embodiment of the present application, before compressing the environmental observation information through a preset digital filter, it also includes: dividing the environmental observation information into multiple quadrant information based on a preset intelligent agent observation model; determining at least one target safety sector information in the multiple quadrant information, and calculating the safety sector angle based on the at least one target safety sector information; obtaining the number of environmental depth sampling points corresponding to the environmental observation information, and calculating the analog domain cutoff frequency and the digital domain cutoff frequency based on the number of environmental depth sampling points, the sensor viewing angle and the safety sector angle; determining the digital filter based on the analog domain cutoff frequency, the digital domain cutoff frequency and the preset window function.

[0037] In the actual implementation process, before compressing the environmental observation information through the digital filter, the embodiment of the present application also needs to adopt the following Figure 2 The model shown is used as an intelligent agent observation model, and the environmental observation information within the intelligent agent's field of view is divided into four quadrants. Among them, the environmental observation information in quadrants I and IV is used to determine the safety of the intelligent agent's forward direction. In the embodiment of the present application, the fan of quadrants I and IV is called a safety fan, and the sum of the fan angles is α<Θ. The intelligent agent can only move forward when the environment observed in the safety fan meets the traffic conditions. If an obstacle is observed in the safety fan, the intelligent agent needs to choose a direction to bypass the obstacle.

[0038] In order to achieve efficient compression and transmission of environmental observations between multiple agents, the embodiment of the present application needs to design a digital low-pass filter to filter the original environmental observations (i.e., environmental observation information) in the analog domain according to the size of the safety sector angle α, thereby obtaining an approximate distribution of environmental free space or obstacles with α as the scale. The filter cutoff frequency in the analog domain is Assuming the number of points of s(n) is M, the digital domain cutoff frequency of the filter is:

[0039]

[0040] According to the above cutoff frequency, the embodiment of the present application first designs an ideal low-pass filter with a length of N, and then performs windowing processing on it. Its time domain expression is:

[0041] h(n)=2f c sinc(2f c (n-τ))w N (n)

[0042] in, w N (n) is a window function with a length of N. In the embodiment of the present application, a Blackman window may be used, and its expression is: In the embodiment of the present application, a digital filter example can be constructed when Θ=360°, α=60°, and M=360, and its time domain and frequency domain images are as follows: Figure 3 shown.

[0043] Therefore, the embodiments of the present application improve the design of a digital filter so that the filter parameters correspond to the physical size of the intelligent agent, thereby achieving compression of the intelligent agent's environmental observation.

[0044] Optionally, in one embodiment of the present application, environmental observation information is compressed by a preset digital filter to obtain environmental compression features corresponding to the environmental observation information, and the environmental compression features are reconstructed based on a preset polygonal approximation strategy to generate environmental observation reconstruction data, including: filtering the environmental observation information through a digital filter to generate environmental observation filter information; calculating all extreme values ​​of the environmental observation filter information, and obtaining the extreme value direction angles and extreme value distances corresponding to all extreme values ​​to obtain environmental compression features based on the extreme value direction angles and extreme value distances; rotating the extreme value direction angle corresponding to the environmental compression feature to obtain the target extreme value direction angle; based on the target extreme value direction angle and the preset linear analysis strategy, connecting the vertices of the extreme points corresponding to all extreme values ​​to generate environmental observation reconstruction data.

[0045] It should be noted that for the agent numbered i, the environmental observation compression and reconstruction process of the embodiment of the present application is as follows: Figure 4As shown in the green box, the original environmental observation (i.e. environmental observation information) s i (n) After filtering, the output signal y is obtained i (n), the embodiment of the present application obtains the extreme value thereof, thereby obtaining the compressed environmental characteristics:

[0046] {Σ i =(φ i，j ,d i，j )|j=1,2,...,N i}

[0047] Among them, φ i,j and d i,j Respectively represent the direction angle and distance of the extreme value, N i represents the number of extreme values ​​of the agent numbered i; due to the effect of low-pass filtering, N i The number is around f0. If the float format is used to complete data transmission between intelligent agents, the embodiment of the present application can theoretically limit the amount of transmitted data to around 16f0.

[0048] In the actual implementation process, the embodiment of the present application can use a wireless communication network to transmit environmental features Σ between agents. i , wherein the wireless communication network includes but is not limited to Wi-Fi, Bluetooth, ultra-wideband (UWB) and the like, which are not specifically limited here.

[0049] When the agent receives the environmental features Σ from the neighboring agent i After that, the embodiment of the present application can first rotate φ i,j Variables are aligned with their own coordinate system; then a linear method is used to connect the vertices of all extreme points to obtain a polygonal approximation of the environmental observation reconstruction data.

[0050] Therefore, the embodiment of the present application constructs a digital filter for compressing environmental observation data. The compressed environmental observation information can be transmitted in a wireless communication network in the form of bit-level data, and the embodiment of the present application can use polygonal approximation to reconstruct the received compressed environmental observation information, thereby effectively ensuring the generation of subsequent planning paths for multiple target intelligent agents.

[0051] In step S103, based on the environmental observation information, all target intelligent agents that meet the preset obstacle detection requirements are obtained, and the environmental compression features are shared among the neighbor intelligent agents corresponding to each target intelligent agent among all the target intelligent agents to obtain the neighbor intelligent agent's neighbor reconstruction data, and the detour direction probability of each target intelligent agent and the neighbor intelligent agent is calculated through the environmental observation reconstruction data and the neighbor reconstruction data to generate planned paths for multiple target intelligent agents according to the detour direction probability.

[0052] After generating the environmental observation reconstruction data, further, the embodiments of the present application require that when the intelligent agent encounters an obstacle, it receives the compressed environmental observation information of the neighboring intelligent agent through the wireless communication network, maps the reconstructed environmental data information to the probability domain, thereby calculating the probability of the detour direction when encountering an obstacle, and realizing the fusion decision of the detour direction in the probability domain, making full use of the multi-agent system's observation of the environment and realizing the optimization of the execution path.

[0053] Optionally, in one embodiment of the present application, the detour direction probability of each target intelligent agent and neighbor intelligent agent is calculated through the environment observation reconstruction data and the neighbor reconstruction data to generate planned paths for multiple target intelligent agents according to the detour direction probability, including: mapping the environment observation reconstruction data and the neighbor reconstruction data to the probability domain respectively through the detour direction probability to obtain the probability domain environment observation reconstruction data and the probability domain neighbor reconstruction data corresponding to the environment observation reconstruction data and the neighbor reconstruction data; fusing the probability domain environment observation reconstruction data and the probability domain neighbor reconstruction data to generate fused reconstruction data, and determining the fused left turn probability of multiple target intelligent agents according to the fused reconstruction data to generate planned paths for multiple target intelligent agents by fusing the left turn probabilities.

[0054] It should be noted that the probability-based fusion decision in the embodiment of the present application is as follows: Figure 4 The orange part shows the detour probability P calculated based on its own environmental observations. s and the detour probability P calculated based on neighbor observations i .

[0055] P s The reconstructed self-environment observation is mapped to the probability domain, and its expression is:

[0056]

[0057] in, x0 is a 4-dimensional vector, which represents the estimated value of the free space area observed by the agent in the four observation quadrants; b0 is the bias term; W0 is the parameter vector, which represents the weight of the agent in each observation quadrant.

[0058] In the embodiment of the present application, the area of ​​the reconstructed polygon in the four observation quadrants is used as an approximate description of the traffic conditions of the agent's surrounding environment, thereby simplifying the observation input into a four-dimensional vector, and the value of each dimension of x0 is:

[0059]

[0060] Among them, n 0,k is the number of extreme values ​​in quadrant k, j = 0 and j = n 0,k+1 indicates the situation on the two boundaries of the observation quadrant, and the rest are extreme values ​​within the observation quadrant.

[0061] P i Mapping the received and reconstructed neighbor observations to the probability domain is expressed as:

[0062] P i =σ(x ni +w n )

[0063] Among them, x ni is the observation of the data receiving agent towards the target direction of the safety sector, that is, the observation rotated to the I and IV quadrants in the local coordinate system; w n is the threshold value, when x ni When x is greater than the threshold, the data receiver's agent will tend to move toward the neighboring agent of the data sender, otherwise it will tend to move in the opposite direction. If there is no minimum value in the fan, it means that there is no direct obstacle near the sender's agent, so x ni The maximum value or the maximum boundary value in the sector is used for assignment; if there is a minimum value in the quadrant, it indicates that there may be obstacles near the sender agent. In this case, the embodiment of the present application can use fuzzy evaluation, x ni Assign the value to the mean of the minimum and maximum values, x ni The expression is:

[0064]

[0065] Among them, k is the extreme value index falling within the safety sector.

[0066] The agent completes P s and P i After the calculation, it is also necessary to fuse in the probability domain to select the detour direction when encountering obstacles. Due to the limitation of the viewing angle range, the agent in the embodiment of the present application only fuses the neighboring agents observed in the field of view.

[0067] Suppose the number of neighboring agents in the left quadrants I and II is n l , the number of neighboring agents in quadrants III and IV on the right is n r , then the fusion process can be expressed as:

[0068]

[0069] Among them, P is the probability of left turn after fusion; w0 is the robot's own observation weight; w i is the neighbor observation weight; the fusion process can be written in the form of a matrix, is the weight matrix, the sum of the matrix elements is 1, is a probability matrix. When assigning neighbor weights, w0>w i .

[0070] Therefore, the embodiments of the present application can efficiently complete the efficient compression, sharing and fusion of multi-agent observations of the environment, and complete multi-agent fusion decisions in the probability domain, thereby achieving optimization of the execution path.

[0071] The following is a simulation verification of the multi-agent collaborative path planning decision-making method of the present application.

[0072] In terms of compressed transmission of environmental observations, the embodiment of the present application can use a UWB wireless communication network with a bandwidth of 78.125 Kbps to transmit intelligent agent observation data. The average amount of data transmitted by the intelligent agent is 48 to 64 bytes, and the average transmission delay is within 7 ms.

[0073] In terms of path optimization, the embodiment of the present application can select Gazebo simulation software as an experimental platform to verify the performance of the algorithm in two different situations: dense forest environment and rocky environment. In the experiment, for two types of maps, 20 two-dimensional space maps with a size of 20m×20m are randomly generated as the environment for multi-agent collaborative path planning. In each map, 15 agents are run simultaneously to perform the task of crossing from the bottom to the top of the map.

[0074] The obstacle density in the map is about 5m 2 / , the experimental evaluation criteria are: the average path of the agent to perform the task. The algorithm used for comparison and verification is the Bug algorithm, which is divided into two cases: fixed left turn and fixed right turn. The experimental statistical results are shown in Table 1. The comparison of the running results in the two environments is as follows Figure 5 shown.

[0075] Table 1

[0076]

[0077] It can be seen from Table 1 that the multi-agent collaborative path planning decision-making method of the embodiment of the present application has a statistically significant optimization effect on path length. Figure 5 The execution path examples of the method proposed in this application and two comparative verification algorithms in two environments are shown. Figure 5 It can be seen that the box indicated by the red dotted line can reflect the optimization effect of the algorithm: through the real-time fusion decision-making of multi-agent observations, the agents in the cluster can spontaneously produce a synergistic effect, choose a direction with fewer local obstacles or more stability, and optimize their own execution path.

[0078] In summary, the present application can realize real-time fusion decision-making of perception information between intelligent agents through efficient compression and reconstruction of environmental observations, thereby optimizing the obstacle avoidance direction of the intelligent agent when encountering obstacles, thereby optimizing the execution path of the intelligent agent when performing crossing tasks.

[0079] According to the multi-agent collaborative path planning decision method proposed in the embodiment of the present application, the sensor perspectives of multiple target agents are determined, and the environmental observation information corresponding to the sensor perspectives is obtained; the environmental observation information is compressed by a preset digital filter, the environmental compression features corresponding to the environmental observation information are obtained, and the environmental compression features are reconstructed based on a preset polygonal approximation strategy to generate environmental observation reconstruction data; based on the environmental observation information, all target agents that meet the preset obstacle detection requirements are obtained, and the environmental compression features are shared between the neighboring agents corresponding to each target agent among all target agents to obtain the neighbor reconstruction data of the neighboring agents, and the detour direction probability of each target agent and the neighboring agent is calculated through the environmental observation reconstruction data and the neighbor reconstruction data to generate the planning paths of multiple target agents according to the detour direction probability. In the case of limited perception, no external positioning and environmental maps, the present application realizes cluster observation sharing through local environmental observation of a multi-agent cluster, thereby reducing the path cost of the crossing task.

[0080] Secondly, the multi-agent collaborative path planning decision-making device proposed according to the embodiment of the present application is described with reference to the accompanying drawings.

[0081] Figure 6 It is a block diagram of a multi-agent collaborative path planning and decision-making device according to an embodiment of the present application.

[0082] like Figure 6 As shown, the multi-agent collaborative path planning decision device 10 includes: a first determination module 100, a reconstruction module 200 and a decision module 300.

[0083] The first determination module 100 is used to determine the sensor viewing angles of multiple target intelligent agents and obtain the environmental observation information corresponding to the sensor viewing angles.

[0084] The reconstruction module 200 is used to compress the environmental observation information through a preset digital filter, obtain environmental compression features corresponding to the environmental observation information, and reconstruct the environmental compression features based on a preset polygonal approximation strategy to generate environmental observation reconstruction data.

[0085] The decision module 300 is used to obtain all target intelligent agents that meet the preset obstacle detection requirements based on the environmental observation information, and share the environmental compression features among the neighbor intelligent agents corresponding to each target intelligent agent among all the target intelligent agents to obtain the neighbor reconstruction data of the neighbor intelligent agent, and calculate the detour direction probability of each target intelligent agent and the neighbor intelligent agent through the environmental observation reconstruction data and the neighbor reconstruction data to generate the planned paths of the multiple target intelligent agents according to the detour direction probability.

[0086] Optionally, in one embodiment of the present application, the multi-agent collaborative path planning decision-making device 10 of the embodiment of the present application further includes: a division module, a first calculation module, a second calculation module and a second determination module.

[0087] Among them, the division module is used to divide the environmental observation information into multiple quadrant information based on a preset agent observation model before compressing the environmental observation information through the preset digital filter.

[0088] The first calculation module is used to determine at least one target safety sector information among the multiple quadrant information, so as to calculate the safety sector angle according to the at least one target safety sector information.

[0089] The second calculation module is used to obtain the number of environmental depth sampling points corresponding to the environmental observation information, and calculate the analog domain cutoff frequency and the digital domain cutoff frequency according to the number of environmental depth sampling points, the sensor viewing angle and the safety sector angle.

[0090] The second determination module is used to determine the digital filter based on the analog domain cutoff frequency, the digital domain cutoff frequency and a preset window function.

[0091] Optionally, in one embodiment of the present application, the reconstruction module 200 includes: a filtering unit, an acquisition unit, a rotation unit and a connection unit.

[0092] The filtering unit is used to perform a filtering operation on the environmental observation information through the digital filter to generate environmental observation filtered information.

[0093] An acquisition unit is used to calculate all extreme values ​​of the environmental observation filter information, and obtain extreme value direction angles and extreme value distances corresponding to all the extreme values, so as to obtain the environmental compression feature according to the extreme value direction angles and the extreme value distances.

[0094] The rotation unit is used to rotate the extreme value direction angle corresponding to the environmental compression feature to obtain a target extreme value direction angle.

[0095] A connection unit is used to connect the vertices of the extreme value points corresponding to all the extreme values ​​based on the target extreme value direction angle and a preset linear analysis strategy to generate the environmental observation reconstruction data.

[0096] Optionally, in one embodiment of the present application, the decision module 300 includes: a mapping unit and a fusion unit.

[0097] Among them, the mapping unit is used to map the environmental observation reconstruction data and the neighbor reconstruction data into the probability domain respectively through the detour direction probability, so as to obtain the probability domain environmental observation reconstruction data and probability domain neighbor reconstruction data corresponding to the environmental observation reconstruction data and the neighbor reconstruction data.

[0098] A fusion unit is used to fuse the probability domain environment observation reconstruction data and the probability domain neighbor reconstruction data to generate fused reconstruction data, and determine the fused left-turn probabilities of the multiple target intelligent agents based on the fused reconstruction data, so as to generate the planned paths of the multiple target intelligent agents through the fused left-turn probabilities.

[0099] It should be noted that the aforementioned explanation of the embodiment of the multi-agent collaborative path planning decision-making method is also applicable to the multi-agent collaborative path planning decision-making device of this embodiment, and will not be repeated here.

[0100] According to the multi-agent collaborative path planning decision device proposed in the embodiment of the present application, it includes a first determination module, which is used to determine the sensor perspective of multiple target agents and obtain the environmental observation information corresponding to the sensor perspective; a reconstruction module, which is used to compress the environmental observation information through a preset digital filter, obtain the environmental compression features corresponding to the environmental observation information, and reconstruct the environmental compression features based on a preset polygonal approximation strategy to generate environmental observation reconstruction data; a decision module, which is used to obtain all target agents that meet the preset obstacle detection requirements based on the environmental observation information, and share the environmental compression features between the neighboring agents corresponding to each target agent in all target agents to obtain the neighbor reconstruction data of the neighboring agent, and calculate the detour direction probability of each target agent and the neighboring agent through the environmental observation reconstruction data and the neighbor reconstruction data, so as to generate the planning path of multiple target agents according to the detour direction probability. In the case of limited perception, no external positioning and environmental map, the present application realizes cluster observation sharing through the local environmental observation of the multi-agent cluster, thereby reducing the path cost of the crossing task.

[0101] Figure 7 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. The electronic device may include:

[0102] A memory 701 , a processor 702 , and a computer program stored in the memory 701 and executable on the processor 702 .

[0103] When the processor 702 executes the program, the multi-agent collaborative path planning decision method provided in the above embodiment is implemented.

[0104] Furthermore, the electronic device further comprises:

[0105] The communication interface 703 is used for communication between the memory 701 and the processor 702 .

[0106] The memory 701 is used to store computer programs that can be executed on the processor 702 .

[0107] The memory 701 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0108] If the memory 701, the processor 702 and the communication interface 703 are implemented independently, the communication interface 703, the memory 701 and the processor 702 can be connected to each other through a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 7 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0109] Optionally, in a specific implementation, if the memory 701, the processor 702 and the communication interface 703 are integrated on a chip, the memory 701, the processor 702 and the communication interface 703 can communicate with each other through an internal interface.

[0110] The processor 702 may be a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application.

[0111] An embodiment of the present application also provides a computer-readable storage medium on which a computer program is stored, and when the program is executed by a processor, the multi-agent collaborative path planning decision method as described above is implemented.

[0112] An embodiment of the present application also provides a computer program product, including a computer program, which, when executed, is used to implement the above-mentioned multi-agent collaborative path planning decision-making method.

[0113] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or N embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0114] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0115] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or N executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

[0116] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or N wirings (electronic devices), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways as necessary and then storing it in a computer memory.

[0117] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above embodiment, the N steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. If implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0118] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0119] In addition, each functional unit in each embodiment of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0120] The storage medium mentioned above may be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present application. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A multi-agent collaborative path planning decision method, characterized in that: The following steps are involved: Determine sensor perspectives of multiple target agents and obtain environmental observation information corresponding to the sensor perspectives; Compressing the environmental observation information through a preset digital filter, obtaining environmental compression features corresponding to the environmental observation information, and reconstructing the environmental compression features based on a preset polygonal approximation strategy to generate environmental observation reconstruction data; Based on the environmental observation information, all target agents that meet the preset obstacle detection requirements are obtained, and the environmental compression features are shared among the neighbor agents corresponding to each target agent among all the target agents to obtain the neighbor reconstruction data of the neighbor agent, and the detour direction probabilities of each target agent and the neighbor agent are calculated through the environmental observation reconstruction data and the neighbor reconstruction data, so as to generate the planned paths of the multiple target agents according to the detour direction probabilities; The step of calculating the detour direction probabilities of each target intelligent agent and the neighbor intelligent agent through the environment observation reconstruction data and the neighbor reconstruction data, so as to generate the planned paths of the multiple target intelligent agents according to the detour direction probabilities, comprises: The detour direction probability of each target intelligent agent is calculated through the environmental observation reconstruction data, and the detour direction probability of the neighbor intelligent agent is calculated through the neighbor reconstruction data, and the environmental observation reconstruction data is mapped into a probability domain using the detour direction probability of each target intelligent agent to obtain probability domain environmental observation reconstruction data corresponding to the environmental observation reconstruction data, and the neighbor reconstruction data is mapped into the probability domain using the detour direction probability of the neighbor intelligent agent to obtain probability domain neighbor reconstruction data corresponding to the neighbor reconstruction data; Fusing the probability domain environment observation reconstruction data and the probability domain neighbor reconstruction data to generate fused reconstruction data, and determining the fused left turn probabilities of the multiple target intelligent agents according to the fused reconstruction data, so as to generate the planning paths of the multiple target intelligent agents through the fused left turn probabilities; The mathematical expression for mapping the environmental observation reconstruction data to the probability domain is: Where P s is the detour direction probability of each target agent; sigmoid function x0 is a 4-dimensional vector, representing the estimated value of the free space area observed by each target agent in the four observation quadrants; b0 is a bias term; W0 is a parameter vector, representing the weight of each target agent in each observation quadrant; The mathematical expression for mapping the neighbor reconstruction data to the probability domain is: P i =σ(x ni +w n ) Where P i is the probability of the neighboring agent's detour direction; ni is the observation of the data receiving agent’s safe sector toward the target direction; n is the threshold value.

2. The method according to claim 1, characterized in that Before compressing the environmental observation information through the preset digital filter, the method further includes: Based on a preset agent observation model, the environmental observation information is divided into a plurality of quadrant information; Determine at least one target safety sector information among the plurality of quadrant information to calculate a safety sector angle according to the at least one target safety sector information; Obtaining the number of environmental depth sampling points corresponding to the environmental observation information, and calculating the analog domain cutoff frequency and the digital domain cutoff frequency according to the number of environmental depth sampling points, the sensor viewing angle, and the safety sector angle; The digital filter is determined based on the analog domain cutoff frequency, the digital domain cutoff frequency and a preset window function.

3. The method according to claim 2, characterized in that The step of compressing the environmental observation information by using a preset digital filter, obtaining environmental compression features corresponding to the environmental observation information, and reconstructing the environmental compression features based on a preset polygonal approximation strategy to generate environmental observation reconstruction data includes: Performing a filtering operation on the environmental observation information by using the digital filter to generate environmental observation filtered information; Calculating all extreme values ​​of the environmental observation filtering information, and obtaining extreme value direction angles and extreme value distances corresponding to all the extreme values, so as to obtain the environmental compression feature according to the extreme value direction angles and the extreme value distances; Rotating the extreme value direction angle corresponding to the environmental compression feature to obtain a target extreme value direction angle; Based on the target extreme value direction angle and a preset linear analysis strategy, the vertices of the extreme value points corresponding to all the extreme values ​​are connected to generate the environmental observation reconstruction data.

4. A multi-agent collaborative path planning decision-making device, characterized in that: include: A first determination module is used to determine the sensor perspectives of multiple target agents and obtain environmental observation information corresponding to the sensor perspectives; A reconstruction module, used to compress the environmental observation information through a preset digital filter, obtain environmental compression features corresponding to the environmental observation information, and reconstruct the environmental compression features based on a preset polygonal approximation strategy to generate environmental observation reconstruction data; A decision module, configured to obtain all target agents that meet the preset obstacle detection requirements based on the environmental observation information, and share the environmental compression features among neighboring agents corresponding to each target agent among all the target agents to obtain neighbor reconstruction data of the neighboring agents, and calculate the detour direction probabilities of each target agent and the neighboring agents through the environmental observation reconstruction data and the neighbor reconstruction data, so as to generate planned paths for the multiple target agents according to the detour direction probabilities; Wherein, the decision module includes: A mapping unit, used to calculate the detour direction probability of each target intelligent agent through the environment observation reconstruction data, and calculate the detour direction probability of the neighbor intelligent agent through the neighbor reconstruction data, and use the detour direction probability of each target intelligent agent to map the environment observation reconstruction data into a probability domain to obtain probability domain environment observation reconstruction data corresponding to the environment observation reconstruction data, and map the neighbor reconstruction data into the probability domain through the detour direction probability of the neighbor intelligent agent to obtain probability domain neighbor reconstruction data corresponding to the neighbor reconstruction data; A fusion unit, used to fuse the probability domain environment observation reconstruction data and the probability domain neighbor reconstruction data to generate fused reconstruction data, and determine the fused left turn probabilities of the multiple target intelligent agents according to the fused reconstruction data, so as to generate the planned paths of the multiple target intelligent agents through the fused left turn probabilities; The mathematical expression for mapping the environmental observation reconstruction data to the probability domain is: Where P s is the detour direction probability of each target agent; sigmoid function x0 is a 4-dimensional vector, representing the estimated value of the free space area observed by each target agent in the four observation quadrants; b0 is a bias term; W0 is a parameter vector, representing the weight of each target agent in each observation quadrant; The mathematical expression for mapping the neighbor reconstruction data to the probability domain is: P i =σ(x ni +w n ) Where P i is the probability of the neighboring agent's detour direction; ni is the observation of the data receiving agent’s safe sector toward the target direction; n is the threshold value.

5. The device according to claim 4, characterized in that Also includes: A division module, used for dividing the environmental observation information into a plurality of quadrant information based on a preset agent observation model before compressing the environmental observation information through the preset digital filter; A first calculation module, used to determine at least one target safety sector information among the plurality of quadrant information, so as to calculate a safety sector angle according to the at least one target safety sector information; A second calculation module is used to obtain the number of environmental depth sampling points corresponding to the environmental observation information, and calculate the analog domain cutoff frequency and the digital domain cutoff frequency according to the number of environmental depth sampling points, the sensor viewing angle and the safety sector angle; The second determination module is used to determine the digital filter based on the analog domain cutoff frequency, the digital domain cutoff frequency and a preset window function.

6. The device according to claim 5, characterized in that The reconstruction module comprises: A filtering unit, configured to perform a filtering operation on the environmental observation information through the digital filter to generate environmental observation filtered information; An acquisition unit, used for calculating all extreme values ​​of the environmental observation filtering information, and acquiring extreme value direction angles and extreme value distances corresponding to all the extreme values, so as to obtain the environmental compression feature according to the extreme value direction angles and the extreme value distances; A rotation unit, used for rotating the extreme value direction angle corresponding to the environmental compression feature to obtain a target extreme value direction angle; A connection unit is used to connect the vertices of the extreme value points corresponding to all the extreme values ​​based on the target extreme value direction angle and a preset linear analysis strategy to generate the environmental observation reconstruction data.

7. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the multi-agent collaborative path planning decision method as described in any one of claims 1 to 3.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: The program is executed by a processor to implement the multi-agent collaborative path planning decision method as described in any one of claims 1-3.

9. A computer program product, comprising a computer program, characterized in that The computer program is executed to implement the multi-agent collaborative path planning decision method as described in any one of claims 1-3.

Images