CN112488153B - Autonomous searching and positioning method for odor sources based on clustering tendency principle - Google Patents

Autonomous searching and positioning method for odor sources based on clustering tendency principle Download PDF

Info

Publication number
CN112488153B
CN112488153B CN202011232403.0A CN202011232403A CN112488153B CN 112488153 B CN112488153 B CN 112488153B CN 202011232403 A CN202011232403 A CN 202011232403A CN 112488153 B CN112488153 B CN 112488153B
Authority
CN
China
Prior art keywords
source
cluster
clustering
searching
particles
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202011232403.0A
Other languages
Chinese (zh)
Other versions
CN112488153A (en
Inventor
陈彬
赵勇
陈海亮
朱正秋
季雅泰
何华
郭金林
汤俊
陈宇宁
黄生俊
金光
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
National University of Defense Technology
Original Assignee
National University of Defense Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by National University of Defense Technology filed Critical National University of Defense Technology
Priority to CN202011232403.0A priority Critical patent/CN112488153B/en
Publication of CN112488153A publication Critical patent/CN112488153A/en
Application granted granted Critical
Publication of CN112488153B publication Critical patent/CN112488153B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/25—Design optimisation, verification or simulation using particle-based methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/08—Probabilistic or stochastic CAD
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2113/00—Details relating to the application field
    • G06F2113/08—Fluids

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Artificial Intelligence (AREA)
  • Probability & Statistics with Applications (AREA)
  • Computer Hardware Design (AREA)
  • Geometry (AREA)
  • Manipulator (AREA)

Abstract

The invention discloses an autonomous searching and positioning method of odor sources based on a clustering tendency principle, which comprises the steps of constructing a gas forward diffusion model and a sensor detection model; modeling a search process of the odor source, and dividing the search process into source item estimation, action decision and execution action; the posterior probability density function of the source item estimation is represented by particle filter approximation, and the update of the source item estimation is converted into the update of the particle filter; clustering the particles projected on the two-dimensional plane by using a density-based clustering algorithm; and designing a clustering tendency action strategy for the action of the mobile robot by using the clustering cluster information. The invention adopts a clustering tendency principle, utilizes a density-based clustering algorithm to extract cluster information in source item estimation, and guides the action of the mobile robot based on the information.

Description

Autonomous searching and positioning method for odor sources based on clustering tendency principle
Technical Field
The technology can be used for the independent searching and positioning tasks of multiple scenes and various odor sources, such as harmful gas source positioning, gaseous hazardous chemical substance leakage source independent searching and the like.
Background
In the productive life of the human society, there are many tasks that require the search and accurate location of unknown odor sources. For example, natural gas leaks are searched to avoid more serious accidents; the leakage source is quickly searched and positioned in the sudden gas-state dangerous chemical leakage accident, so that the emergency disposal work is timely carried out, and the like. The existing method generally utilizes a mobile robot carrying a sensor to search, and adopts a corresponding autonomous search algorithm to guide the mobile robot to sense and search behaviors, and the method is generally called as a sourcing method.
Since the 80 s of the last century, researchers began researching various autonomous search algorithms. In general, the autonomous search algorithm can be classified into three types: gradient-based algorithms, algorithms based on the principles of bionics and algorithms based on probabilities and mappings. The gradient-based algorithm and the bionic principle-based algorithm are simpler in principle and less in required calculation amount, but both of the gradient-based algorithm and the bionic principle depend on concentration gradient. In reality, however, the steady gas plume will be divided into intermittent, sporadic gas slugs due to the effects of turbulence effects. In such a case, the above two autonomous search algorithms would fail seriously.
The probability and mapping based algorithm estimates source position equivalent parameters by using a probability density function, and updates the probability density function by continuously collecting related information by using a mobile sensor, thereby obtaining more accurate estimation. The method can model the gas diffusion mode under turbulent flow conditions, so that accurate source searching can be carried out under the conditions. The Infotaxis algorithm is a typical probability and mapping based algorithm that employs an information entropy based reward function to balance utilization and development issues and to guide the mobile sensor to choose to move to the nearest location where the information gain is greatest at each step. Some researchers further model the source searching process as a part of observable Markov decision process, improve the original grid-based Infotaxi algorithm by using particle filtering, and simultaneously refer to the autonomous search algorithm using information entropy as a reward function as a cognitive search strategy. The entotaxis algorithm adopts a sequential Monte Carlo framework of the Inotaxis algorithm, and designs a reward function with higher calculation efficiency according to the maximum entropy sampling principle. The reward function is used to calculate the uncertainty of future detection values, so as to guide the mobile sensor to go to the adjacent position with the maximum uncertainty in each step, so as to collect more source item information.
The cognitive search strategy requires more complex calculations at each step to support the decision, resulting in an increase in overall sourcing time. In practical sourcing applications, the sourcing speed is often as important as the sourcing accuracy. Therefore, the cognitive search strategy will be severely limited in practical applications. However, the sequential Monte Carlo framework based on particle filtering adopted by the method can fuse the collected discontinuous and sporadic source item information into the estimation of the source item parameters, so that the cognitive search algorithm can keep better search performance in the turbulent flow effect.
Disclosure of Invention
The technical problems to be solved by the invention are as follows: aiming at the problems in the prior art, the invention provides an autonomous searching and positioning method of odor sources based on the clustering tendency principle.
In order to solve the technical problems, the invention adopts the technical scheme that:
a scent source autonomous searching and positioning method based on a clustering tendency principle comprises the following steps:
1) constructing a gas forward diffusion model and a sensor detection model;
2) modeling a search process of an odor source, and dividing the search process into a source item estimation part, an action decision part and an execution action part;
3) approximating a posterior probability density function for representing source item estimation by using particle filtering, and converting the update of the source item estimation in the searching process of the odor source into the update of the particle filtering;
4) clustering the particles in the particle filter projected on the two-dimensional plane by using a density-based clustering algorithm to obtain clustering cluster information;
5) and designing a clustering tendency action strategy for the action of the mobile robot by using the obtained clustering cluster information.
Optionally, the functional expression of the gas forward diffusion model constructed in step 1) is:
Figure BDA0002765640510000021
in the above formula, R (R | θ)0) Indicating the sensor at the r ═ { x, y } position and the coordinate r0＝{x0,y0The average number of contacts of gas molecules emitted by the odor source, where a is the radius of the sensor, Q represents the rate at which the odor source leaks gas molecules, and D is the number of contacts of the odor sourceDiffusivity of gas molecules, theta0＝{r0Q is a source term parameter to be estimated, V is the wind speed, λ represents an intermediate parameter, and τ represents the average lifetime of the gas molecules;
the function expression of the sensor detection model constructed in the step 1) is as follows:
Figure BDA0002765640510000022
in the above formula, P (d (r) | θ0) The probability that the actual number of contacts of the sensor with the gas molecules per unit time is d, R (R | θ [ ])0) Representing the velocity Q and coordinate θ of a sensor with radius a at the r ═ { x, y } position0＝{x0,y0The average number of contacts of gas molecules emanating from the odor source of (d) is the number of contacts.
Optionally, the step of modeling the search process of the scent source in step 2) comprises:
2.1) source term estimation: constructing an initial source term estimate from prior information about the odor source, the initial source term estimate using a probability density function P (theta)0) Represents;
2.2) action decision: obtaining an action instruction according to a preset action strategy according to initial source item estimation;
2.3) performing an action: the mobile robot executes the action instruction and moves to a new position, and the mobile robot collects new information; updating the source item estimation based on the Bayesian posterior probability formula according to the new information to obtain a new source item estimation P (theta)k|D1:k) Before the next time step begins, the mobile robot judges whether source confirmation can be carried out or source searching resources are exhausted, wherein the source searching resources refer to energy or time, if the source confirmation can be carried out, the source searching is finished, and odor source position information is output; if the scent source is not determined and the remaining source searching resources exist, skipping to execute the step 2.2) after the next time step; and if the source searching resources are exhausted, judging that the source searching fails, ending and exiting.
Optionally, step 3) comprises:
3.1) Source term estimation representation based on particle Filtering: for the posterior probability density function P (theta)k|D1:k) Sampling to generate N random samples with weights
Figure BDA0002765640510000031
Wherein theta iskPoint estimates representing parameters of the source item, D1:kRepresenting information collected cumulatively from an initial time to a current time,
Figure BDA0002765640510000032
the point estimate of the source term parameter for the ith particle representing the kth step,
Figure BDA0002765640510000033
is the weight value corresponding to the particle, and has
Figure BDA0002765640510000034
Obtaining N random samples with weight
Figure BDA0002765640510000035
Approximating the posterior probability density function P (θ) as estimated source term as N particles according tok|D1:k)：
Figure BDA0002765640510000036
In the above formula, δ (·) represents a dirichlet function;
3.2) updating the source term estimation based on particle filtering: introducing an importance function q (theta)k|D1:k)＝P(θk-1|D1:k-1) The updating formula of the particle filter is simplified as follows:
Figure BDA0002765640510000037
in the above formula, the first and second carbon atoms are,
Figure BDA0002765640510000038
representing the non-normalized weights at time k,
Figure BDA0002765640510000039
representing the non-normalized weights at time k-1,
Figure BDA00027656405100000310
the function of the importance is represented by,
Figure BDA00027656405100000311
representing the weight normalized at the moment k, wherein N is the number of particles;
3.3) resampling step: under the condition of keeping the total number of the particles unchanged, copying the particles with large weight according to the weight to replace invalid particles so as to achieve the purposes of deleting small-weight particles and copying large-weight particles;
3.4) Markov chain Monte Carlo step: the Metropolis-Hastings sampling algorithm based on the Markov chain Monte Carlo method is adopted to increase the diversity of the particles.
Optionally, the step of clustering the particles in the particle filter projected on the two-dimensional plane by using a density-based clustering algorithm in step 4) includes: utilizing a density-based clustering algorithm to cluster and identify the particles in the particle filter projected on a two-dimensional plane as a cluster, wherein the cluster is a region with dense particle distribution in the particle filter, and the parameter of the cluster is an Eps neighborhood, wherein the definition function expression of the Eps neighborhood is as follows:
NEps(p)＝{q∈D|dist(p,q)≤Eps}
in the above formula, NEps(p) represents an Eps neighborhood of a sample p given a sample p e D, D is a sample set, dist (p, q) represents a distance between the sample p and q, and Eps is a set distance threshold; the arbitrary cluster C is a non-empty subset of the clustered sample set D, is a maximum density connected sample set derived from a density reachable relation, and satisfies the following conditions: condition 1: for all p, q ∈ D, if p ∈ C and q is reachable from p, the sample q ∈ C; condition 2: for all p, q ∈ C, sample p and sample q are density-connected.
Optionally, the detailed step in step 5) includes:
5.1) carrying out effective cluster judgment on the obtained cluster: if the proportion of particles in a certain cluster to the total number of particles exceeds a threshold value, judging the cluster as an effective cluster;
5.2) selecting a target cluster: if the effective cluster exists in the scene at a certain source searching stage, the area where the maximum effective cluster is located is the target cluster; if no effective cluster exists in the scene, the whole particle cluster is taken as a target cluster;
5.3) the mobile robot moves to the mean point of the target cluster by a fixed step length, and once the mean point is reached, the whole target cluster is explored by adopting square search until the next target cluster appears or the source searching is terminated.
In addition, the invention also provides a system for automatically searching and positioning the odor source based on the clustering tendency principle, which comprises the following steps:
the model building program unit is used for building a gas forward diffusion model and a sensor detection model;
the search process modeling program unit is used for modeling the search process of the odor source and dividing the search process into a source item estimation part, an action decision part and an execution action part;
a particle filter updating program unit for approximating a posterior probability density function for representing the source item estimation by using the particle filter and converting the update of the source item estimation in the searching process of the odor source into the update of the particle filter;
the density clustering program unit is used for clustering the particles in the particle filter projected on the two-dimensional plane by using a density-based clustering algorithm to obtain clustering cluster information;
and the action strategy design program unit is used for designing a clustering tendency action strategy for the action of the mobile robot by using the obtained clustering cluster information.
In addition, the invention also provides a scent source autonomous searching and positioning system based on the clustering tendency principle, which comprises a computer device, wherein the computer device comprises a microprocessor and a memory which are connected with each other, and the microprocessor is programmed or configured to execute the steps of the scent source autonomous searching and positioning method based on the clustering tendency principle.
In addition, the invention also provides a scent source autonomous searching and positioning system based on the clustering tendency principle, which comprises a computer device, wherein the computer device comprises a microprocessor and a memory which are connected with each other, and a computer program which is programmed or configured to execute the scent source autonomous searching and positioning method based on the clustering tendency principle is stored in the memory.
Furthermore, the present invention also provides a computer-readable storage medium having stored therein a computer program programmed or configured to execute the cluster tropism principle based scent source autonomous search and location method.
Compared with the prior art, the invention has the following advantages: the method comprises the steps of constructing a gas forward diffusion model and a sensor detection model; modeling a search process of an odor source, and dividing the search process into a source item estimation part, an action decision part and an execution action part; approximating a posterior probability density function for representing source item estimation by using particle filtering, and converting the update of the source item estimation in the searching process of the odor source into the update of the particle filtering; clustering the particles in the particle filter projected on the two-dimensional plane by using a density-based clustering algorithm to obtain clustering cluster information; and designing a clustering tendency action strategy for the action of the mobile robot by using the obtained clustering cluster information. The invention adopts a method based on a clustering tendency principle, utilizes a clustering algorithm based on density to extract cluster information in source item estimation, and guides the action of the mobile robot based on the information.
Drawings
FIG. 1 is a schematic diagram of a basic flow of a method according to an embodiment of the present invention.
Fig. 2 shows an autonomous sourcing process in step 2) of the method according to an embodiment of the present invention.
Fig. 3 shows a particle filter updating process in step 3) of the method according to the embodiment of the invention.
Fig. 4 is a flowchart of the density-based clustering algorithm in step 4) of the method according to the embodiment of the present invention.
Fig. 5 is a schematic diagram of the maximum effective cluster in step 5) of the method according to the embodiment of the present invention.
Fig. 6 is a schematic diagram of square search in step 5) of the method according to the embodiment of the present invention.
FIG. 7 is a flowchart of an action decision process based on cluster information in step 5) of the method according to the embodiment of the present invention.
Detailed Description
The following description will be made in detail on the autonomous searching and positioning method of odor sources based on the clustering tendency principle of the present embodiment, taking the task of searching and positioning harmful gas leakage sources as an example. It should be noted that, the embodiment is only described by taking the task of searching and locating the harmful gas leakage source as an example, and needless to say, the embodiment can also be applied to other various scenes in which the odor source needs to be searched and located, and details are not described here.
As shown in fig. 1, the scent source autonomous searching and positioning method based on the clustering tendency principle of the present embodiment includes:
1) constructing a gas forward diffusion model and a sensor detection model;
2) modeling a search process of an odor source, and dividing the search process into a source item estimation part, an action decision part and an execution action part;
3) approximating a posterior probability density function for representing source item estimation by using particle filtering, and converting the update of the source item estimation in the searching process of the odor source into the update of the particle filtering;
4) clustering the particles in the particle filter projected on the two-dimensional plane by using a density-based clustering algorithm to obtain clustering cluster information;
5) and designing a clustering tendency action strategy for the action of the mobile robot by using the obtained clustering cluster information.
In the step 1) of the embodiment, a gas forward diffusion model is specifically established according to a convection diffusion model, and a sensor detection model under the influence of turbulence factors and sensor measurement errors is considered.
The function expression of the gas forward diffusion model constructed in the step 1) is as follows:
Figure BDA0002765640510000061
in the above formula, R (R | θ)0) Indicating the sensor at the r ═ { x, y } position and the coordinate r0＝{x0,y0The average number of contacts of gas molecules emanating from the odor source, where a is the radius of the sensor, Q represents the rate at which the odor source leaks gas molecules, D is the diffusivity of the gas molecules, and θ0＝{r0Q is a source term parameter to be estimated, V is the wind speed, λ represents an intermediate parameter, and τ represents the average lifetime of the gas molecules;
if an unknown odor source exists in the search scene, the coordinate of the unknown odor source is theta0＝{x0,y0}. The odor source leaks gas into the scene at a steady rate Q, D being the diffusivity of the gas. A stable wind field exists in a scene, and the wind speed is constant and is V. If the average lifetime of a gas molecule is τ, a sensor with radius a at the R ═ { x, y } position will contact the gas molecule at a certain rate, and the average number of contacts R (R | θ ═ R) will contact the gas molecule0) I.e. calculated according to the above formula.
The function expression of the sensor detection model constructed in the step 1) is as follows:
Figure BDA0002765640510000062
in the above formula, P (d (r) | θ0) The probability that the actual number of contacts of the sensor with the gas molecules per unit time is d, R (R | θ [ ])0) Representing the velocity Q and coordinate θ of a sensor with radius a at the r ═ { x, y } position0＝{x0,y0The average number of contacts of gas molecules emanating from the odor source of (d) is the number of contacts. However, due to turbulence effects and errors in sensor accuracy, the sensor interacts with the gas molecules in a unit of timeThe actual number of contacts is a random variable. A poisson distribution can be used to model this random variable, i.e. the probability that the actual number of contacts is d is shown in the above equation.
As shown in fig. 2, the step of modeling the search process of the scent source in step 2) in the present embodiment includes:
2.1) source term estimation: constructing an initial source item estimate from a priori information about the odor source, the initial source item estimate employing a probability density function P (θ)0) Represents;
2.2) action decision: obtaining an action instruction according to a preset action strategy according to initial source item estimation;
2.3) performing an action: the mobile robot executes the action instruction and moves to a new position, and the mobile robot collects new information; updating the source item estimation based on the Bayesian posterior probability formula according to the new information to obtain a new source item estimation P (theta)k|D1:k) Before the next time step starts, the mobile robot judges whether source confirmation can be carried out or source searching resources are exhausted currently, wherein the source searching resources refer to energy or time, if the source confirmation can be carried out, the source searching is finished, and source position information of the odor is output; if the scent source is not determined and the remaining source searching resources exist, skipping to execute the step 2.2) after the next time step; and if the source searching resources are exhausted, judging that the source searching fails, ending and exiting.
As shown in fig. 3, step 3) in this embodiment includes:
3.1) Source term estimation representation based on particle Filtering: since the source term estimate based on the a posteriori probability density function in step 2) is non-linear, it is difficult to find an analytical solution, and therefore particle filtering is used to approximate the source term estimate. In this embodiment, the posterior probability density function P (theta)k|D1:k) Sampling to generate N random samples with weights
Figure BDA0002765640510000071
Wherein theta iskPoint estimates representing parameters of the source item, D1:kIndicating information cumulatively collected from an initial time to a current time, thetak iThe point estimate of the source term parameter for the ith particle representing the kth step,
Figure BDA0002765640510000072
is the weight value corresponding to the particle, and has
Figure BDA0002765640510000073
Obtaining N random samples with weight
Figure BDA0002765640510000074
Approximating the posterior probability density function P (θ) as estimated source term as N particles according tok|D1:k)：
Figure BDA0002765640510000075
In the above formula, δ (·) represents a dirichlet function;
3.2) updating the source term estimation based on particle filtering: due to the posterior probability density function P (theta)k|D1:k) Is unknown and cannot be directly sampled from it. Introducing an importance function q (theta) by using an importance sampling principlek|D1:k). Its probability distribution should be equal to P (theta)k|D1:k) Similarly, and easily calculated, there are:
Figure BDA0002765640510000076
in the above formula, the first and second carbon atoms are,
Figure BDA0002765640510000077
represents the normalized weight at time k, P (θ)k|D1:k) Representing a posterior probability density function, q (θ)k|D1:k) Representing the importance function introduced. In this embodiment, an importance function q (θ) is introducedk|D1:k)＝P(θk-1|D1:k-1) At the same time, as the source of the odor remains unchanged during the sourcing process, i.e.
Figure BDA0002765640510000078
The update formula of the particle filter is simplified as follows:
Figure BDA0002765640510000079
in the above formula, the first and second carbon atoms are,
Figure BDA00027656405100000710
representing the non-normalized weights at time k,
Figure BDA00027656405100000711
representing the non-normalized weights at time k-1,
Figure BDA00027656405100000712
the function of the importance is represented by,
Figure BDA00027656405100000713
representing the weight normalized at the moment k, wherein N is the number of particles;
3.3) resampling step: under the condition of keeping the total number of the particles unchanged, copying the particles with large weight according to the weight to replace invalid particles so as to achieve the purposes of deleting small-weight particles and copying large-weight particles;
in particle filtering, the problem of particle degradation generally exists, that is, after a plurality of iterations, the weight of most particles is negligibly small except for a small part of particles, and the variance of the weight of the particles increases with time. Particles with particle weights that can be negligibly small are called invalid particles, and as the number of invalid particles increases, a large amount of computation is spent on particles that hardly contribute to the estimated a posteriori probability density function, so that the particle filter estimation performance is degraded. Resampling is also called resampling and aims at reducing the number of invalid particles. Under the condition of keeping the total number of the particles unchanged, the particles with large weight are copied according to the weight to replace the invalid particles. In short, small-weight particles are deleted and large-weight particles are copied. The basic process of resampling is to resample the random discrete probability distribution column approximated by the existing particle set to obtain a new sample set, and the weight of each particle in the new sample set is equal.
3.4) Markov chain Monte Carlo step: the Metropolis-Hastings sampling algorithm based on the Markov chain Monte Carlo method is adopted to increase the diversity of the particles.
A new problem is caused after the resampling step. Resampling results in a loss of particle diversity, and after several iterations, all particles may become duplicates of some few particles. Furthermore, since all particles may initially have errors in the point estimates of the source term parameters, such errors cannot be reduced or eliminated by updating the particle filter. Therefore, it is necessary to solve this problem by increasing the particle diversity, which is increased by the Metropolis-Hastings sampling algorithm based on the Markov chain Monte Carlo method after resampling.
As shown in fig. 4, the step of clustering the particles in the particle filter projected on the two-dimensional plane by using the density-based clustering algorithm in step 4) of this embodiment includes: utilizing a density-based clustering algorithm to cluster and identify the particles in the particle filter projected on a two-dimensional plane as a cluster, wherein the cluster is a region with dense particle distribution in the particle filter, and the parameter of the cluster is an Eps neighborhood, wherein the definition function expression of the Eps neighborhood is as follows:
NEps(p)＝{q∈D|dist(p,q)≤Eps}
in the above formula, NEps(p) representing an Eps neighborhood of a sample p given a sample p e D, D being a sample set, dist (p, q) representing a distance between the sample p and q, and Eps being a set distance threshold; the arbitrary cluster C is a non-empty subset of the clustered sample set D, is a maximum density connected sample set derived from a density reachable relation, and satisfies the following conditions: condition 1: for all p, q ∈ D, if p ∈ C and q is reachable from p, the sample q ∈ C; condition 2: for all p, q ∈ C, sample p and sample q are density-connected.
The density-based clustering algorithm describes how closely a sample is distributed in terms of a set of parameters about the "neighborhood", and the main terms of the algorithm are defined as follows:
eps neighborhood: given a sample p e D, the Eps neighborhood N of pEps(p) is defined as the area centered at p and having a radius of Eps, i.e.: n is a radical ofEps(p) { q ∈ D | dist (p, q) ≦ Eps } where D is the sample set and dist (p, q) denotes the distance between samples p and q.
Core point, boundary point: for sample p, given an integer MinPts, if at least MinPts samples are contained within the Eps neighborhood of p, i.e. | NEpsIf (p) | is equal to or greater than MinPts, p is called the core point. Samples that are not core points but fall within the Eps neighborhood of a certain core point are called boundary points.
The density is up to: for samples p and q, if q is the core point and p falls into the Eps neighborhood of q, the sample p is called from sample q with direct density. The direct density is not symmetrical, i.e. the sample q is not necessarily direct density from the sample p.
The density can reach: when there is a sample chain p1,p2,…pnWherein p is1＝q,pnP, and sample pi+1Is from a sample piIf the starting density is direct, the sample p is called the starting density reachable from the sample q. The density can be transitive, but not symmetrical.
Density connection: samples p and q are said to be density-connected if there is a sample o e D such that both samples p and q are density-reachable from sample o. The density connection has symmetry.
Cluster and noise points: for sample set D, cluster C is a non-empty subset of D, and is the largest density-connected sample set derived from the density reachability relationship, i.e. the following condition is satisfied:
condition 1: for all p, q ∈ D, q ∈ C if p ∈ C and q is reachable from p with density.
Condition 2: for all p, q ∈ C, p and q are density-connected.
For a sample that does not belong to any cluster, the sample is defined as a noise point.
Fig. 4 is a flowchart of the density-based clustering algorithm in step 4) of the method according to the embodiment of the present invention. In the step 4), a clustering algorithm based on density is utilized, so that the region with more densely distributed particles in the particle filter can be identified as a cluster, and the action decision of the mobile robot can be supported based on the information of the cluster.
The detailed steps in step 5) of this embodiment include:
5.1) carrying out effective cluster judgment on the obtained cluster: if the proportion of particles in a certain cluster to the total number of particles exceeds a threshold value, judging the cluster as an effective cluster;
in each time step, the robot selects a target cluster for approaching and exploring, and therefore, the method is called a clustering tendency method. For a given set of particles representing the current source item estimate, clustering the set of particles using a density-based clustering algorithm, wherein the cluster containing the most particles among all clusters is called the largest cluster. In general, the maximum cluster region may be considered as a region where a leakage source is most likely to exist, and is a target cluster. In some cases, however, the maximum cluster contains a very small proportion of the total number of particles, and the information provided is insufficient to indicate that the leakage source is located in this region. For example, in the initial stage of sourcing, the collected information does not cause the particle filter to converge significantly, and the clusters may consist of a plurality of randomly dispersed particles. The concept of a valid cluster is thus defined: if the proportion of the particles contained in a certain cluster to the total number of particles exceeds a threshold value, the cluster is called a valid cluster. The concept of valid clusters is to reduce the possibility of misjudgment, i.e. to continue collecting information to increase the accuracy of source item estimation without blindly searching without certain confidence. The cluster containing the most particles among the effective clusters is called the most effective cluster, and a schematic diagram thereof is shown in fig. 5.
5.2) selecting a target cluster: if the effective cluster exists in the scene at a certain source searching stage, the area where the maximum effective cluster is located is the target cluster; if no effective cluster exists in the scene, the whole particle cluster is taken as a target cluster;
if an effective cluster exists in a certain source searching stage, the area where the maximum effective cluster is located is the target cluster; and if no effective cluster exists in the scene, taking the whole particle cluster as a target cluster.
5.3) the mobile robot moves to the mean point of the target cluster by a fixed step length, and once the mean point is reached, the whole target cluster is explored by adopting square search until the next target cluster appears or the source searching is terminated.
Once the target cluster is selected. The mobile robot moves to the mean point of the target cluster by adopting a fixed step length, and once the mean point is reached, the whole target cluster is explored by adopting square search until the next target cluster appears or the source searching is terminated. The square search model is shown in fig. 6, and the complete clustering tendency action strategy flow is shown in fig. 7.
In summary, in step 1) of the autonomous searching and locating method for an odor source based on the clustering tendency principle, in order to fully utilize the relevant information released into the air by the odor source and the environmental information such as the wind field, a forward diffusion mode of the gas needs to be modeled first. This step can be referred to some existing and more mature diffusion models. Such as commonly used semi-empirical gaussian models, convective diffusion models derived based on convective diffusion equations, and fluid mechanics models based on fluid mechanics analysis methods. Considering that a general source searching task has a very strict time requirement, when a model is established, a trade-off between the calculation accuracy and the calculation efficiency needs to be considered. In step 2), the process of autonomously searching for the odor source by adopting the source searching method is generally modeled as a sequential decision process: the autonomous homing algorithm will at each step guide the mobile robot to take an action, such as moving to a position adjacent to the current position. The mobile robot detects at a new position by using the sensor, extracts information in a detection value and assists a next movement decision. By repeating this process, the mobile robot can finally move to the smell source attachment and perform source confirmation. Since the actual position of the odor source is unknown in this process, an estimate of the source item information (source position) needs to be made simultaneously throughout the process. In step 3), based on a Bayesian framework, the collected information of each step can be integrated into the source item estimation. For a more flexible representation of the estimate, particle filtering may be employed to represent an iterative update of the source term estimate. Particle filtering employs a set of weighted particles to approximate a probability density function, i.e., a source term estimate. Each particle represents a point estimate of the source term parameter, whose weight represents the likelihood that the point estimate is a true value. And updates to the source term estimates may be translated into updates to the particle weights. In order to maintain particle diversity and avoid particle degradation, a resampling step and a markov chain monte carlo step can also be added in the updating process. In step 4), based on the point estimation of each particle in the particle filter, the particles can be projected to a two-dimensional plane where the search area is located, and the whole particle set presents a certain spatial distribution on the two-dimensional plane. Under the current source term estimation, the more dense the particles are, the more likely it is that the leakage source is located at that position. Using a density-based clustering algorithm, areas with sufficiently high density can be divided into clusters, and arbitrarily shaped clusters can be found in noisy data sets. In the step 5), the mobile robot can be assisted to perform action decision according to the cluster related information extracted in the step 4). For example, the density of the particles contained in a certain local region reflects the possibility that the actual source item parameter falls into the region, so that the position where the leakage source is most likely to exist can be determined by searching the region with the highest particle density. However, there may be problems with this moving method, and more efficient strategies for robot action decision making are needed. Therefore, the scent source autonomous searching and positioning method based on the clustering tendency principle extracts the information of particle filter particle distribution and a clustering tendency action strategy designed based on the information. The autonomous searching and positioning method for the odor source based on the clustering tendency principle can realize the fast and accurate searching and positioning of the unknown odor source. Compared with the conventional method, the autonomous searching and positioning method of the odor source based on the clustering tendency principle has the advantages of better searching performance and higher calculation efficiency, and therefore, the autonomous searching and positioning method of the odor source has higher practical value. The autonomous searching and positioning method for the odor source based on the clustering tendency principle has the advantages of strong practicability, wide application prospect, high efficiency, high searching speed and the like.
In addition, this embodiment also provides an autonomous searching and positioning system of odor source based on clustering tendency principle, including:
the model building program unit is used for building a gas forward diffusion model and a sensor detection model;
the search process modeling program unit is used for modeling the search process of the odor source and dividing the search process into a source item estimation part, an action decision part and an execution action part;
a particle filter updating program unit for approximating a posterior probability density function for representing the source item estimation by using the particle filter and converting the update of the source item estimation in the searching process of the odor source into the update of the particle filter;
the density clustering program unit is used for clustering the particles in the particle filter projected on the two-dimensional plane by using a density-based clustering algorithm to obtain clustering cluster information;
and the action strategy design program unit is used for designing a clustering tendency action strategy for the action of the mobile robot by using the obtained clustering cluster information.
In addition, the embodiment further provides a system for searching and locating scent sources based on the clustering tendency principle, which includes a computer device, the computer device includes a microprocessor and a memory connected to each other, and the microprocessor is programmed or configured to execute the steps of the method for searching and locating scent sources based on the clustering tendency principle.
In addition, the embodiment further provides a system for searching and locating scent sources based on the clustering tendency principle, which includes a computer device, the computer device includes a microprocessor and a memory connected to each other, and the memory stores therein a computer program programmed or configured to execute the method for searching and locating scent sources based on the clustering tendency principle.
In addition, the present embodiment also provides a computer-readable storage medium, in which a computer program programmed or configured to execute the aforementioned scent source autonomous search and location method based on the clustering tendency principle is stored.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-readable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein. The present application is directed to methods, apparatus (systems), and computer program products according to embodiments of the application wherein instructions, which execute via a flowchart and/or a processor of the computer program product, create means for implementing functions specified in the flowchart and/or block diagram block or blocks. These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks. These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The above description is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments, and all technical solutions belonging to the idea of the present invention belong to the protection scope of the present invention. It should be noted that modifications and embellishments within the scope of the invention may occur to those skilled in the art without departing from the principle of the invention, and are considered to be within the scope of the invention.

Claims (9)

1. A scent source autonomous searching and positioning method based on clustering tendency principle is characterized by comprising the following steps:
1) constructing a gas forward diffusion model and a sensor detection model;
2) modeling a search process of an odor source, and dividing the search process into a source item estimation part, an action decision part and an execution action part;
3) approximating a posterior probability density function for representing source item estimation by using particle filtering, and converting the update of the source item estimation in the searching process of the odor source into the update of the particle filtering;
4) clustering the particles in the particle filter projected on the two-dimensional plane by using a density-based clustering algorithm to obtain clustering cluster information;
5) the method for designing the clustering tendency action strategy for the action of the mobile robot by utilizing the obtained clustering cluster information comprises the following steps:
5.1) carrying out effective cluster judgment on the obtained cluster: if the proportion of particles in a certain cluster to the total number of particles exceeds a threshold value, judging the cluster as an effective cluster;
5.2) selecting a target cluster: if the effective cluster exists in the scene at a certain source searching stage, the area where the maximum effective cluster is located is the target cluster; if no effective cluster exists in the scene, the whole particle cluster is taken as a target cluster;
5.3) the mobile robot moves to the mean point of the target cluster by a fixed step length, and once the mean point is reached, the whole target cluster is explored by adopting square search until the next target cluster appears or the source searching is terminated.
2. The scent source autonomous searching and locating method based on clustering tendency principle according to claim 1, wherein the function expression of the gas forward diffusion model constructed in step 1) is:
Figure FDA0003580856400000011
in the above formula, R (R | θ)0) Watch (A)The sensor and coordinate at r ═ { x, y } position is shown as r0＝{x0,y0The average number of contacts of gas molecules emitted by the odor source, where a is the radius of the sensor, Q represents the rate at which the odor source leaks gas molecules, D is the diffusivity of the gas molecules, θ0＝{r0Q is a source term parameter to be estimated, V is the wind speed, lambda represents an intermediate parameter, and tau represents the average lifetime of the gas molecules;
the function expression of the sensor detection model constructed in the step 1) is as follows:
Figure FDA0003580856400000012
in the above formula, P (d (r) | θ0) The probability that the actual number of contacts of the sensor with the gas molecules per unit time is d, R (R | θ [ ])0) Representing the velocity Q and coordinate θ of a sensor with radius a at the r ═ { x, y } position0＝{x0,y0The average number of contacts of gas molecules emanating from the odor source of (d) is the number of contacts.
3. The scent source autonomous searching and locating method based on clustering tendency principle according to claim 1, wherein the step of modeling the scent source searching process in step 2) comprises:
2.1) source term estimation: constructing an initial source term estimate from prior information about the odor source, the initial source term estimate using a probability density function P (theta)0) Represents;
2.2) action decision: obtaining an action instruction according to a preset action strategy according to initial source item estimation;
2.3) performing an action: the mobile robot executes the action command and moves to a new position, and the mobile robot collects new information; updating the source item estimation based on the Bayesian posterior probability formula according to the new information to obtain a new source item estimation P (theta)k|D1:k) Before the next time step begins, the mobile robot judges whether the mobile robot can currently judgePerforming source confirmation or exhausting source searching resources, wherein the source searching resources refer to energy or time, if the source confirmation can be performed, the source searching is finished, and the position information of the odor source is output; if the scent source is not determined and the remaining source searching resources exist, skipping to execute the step 2.2) after the next time step; and if the source searching resources are exhausted, judging that the source searching fails, ending and exiting.
4. The scent source autonomous searching and locating method based on clustering tendency principle according to claim 1, wherein step 3) comprises:
3.1) Source term estimation representation based on particle Filtering: for the posterior probability density function P (theta)k|D1:k) Sampling to generate N random samples with weights
Figure FDA0003580856400000021
Wherein theta iskPoint estimates representing parameters of the source item, D1:kRepresenting information collected cumulatively from an initial time to a current time,
Figure FDA0003580856400000022
the point estimate of the source term parameter for the ith particle representing the kth step,
Figure FDA0003580856400000023
is the weight value corresponding to the particle, and has
Figure FDA0003580856400000024
Obtaining N random samples with weight
Figure FDA0003580856400000025
Approximating the posterior probability density function P (θ) as estimated source term as N particles according tok|D1:k)：
Figure FDA0003580856400000026
In the above formula, the first and second carbon atoms are,
Figure FDA0003580856400000027
representing a dirichlet function;
3.2) updating the source term estimation based on particle filtering: introducing an importance function q (theta)k|D1:k)＝P(θk-1|D1:k-1) The updating formula of the particle filter is simplified as follows:
Figure FDA0003580856400000028
in the above formula, the first and second carbon atoms are,
Figure FDA0003580856400000029
representing the non-normalized weights at time k,
Figure FDA00035808564000000210
representing the non-normalized weights at time k-1,
Figure FDA00035808564000000211
a function of the importance is represented by,
Figure FDA00035808564000000212
representing the weight normalized at the moment k, wherein N is the number of particles;
3.3) resampling step: under the condition of keeping the total number of the particles unchanged, copying the particles with large weight according to the weight to replace invalid particles so as to achieve the purposes of deleting small-weight particles and copying large-weight particles;
3.4) Markov chain Monte Carlo step: the Metropolis-Hastings sampling algorithm based on the Markov chain Monte Carlo method is adopted to increase the diversity of the particles.
5. The scent source autonomous searching and locating method based on clustering tendency principle according to claim 1, wherein the step of clustering the particles in the particle filter projected on the two-dimensional plane by using the density-based clustering algorithm in step 4) comprises: utilizing a density-based clustering algorithm to cluster and identify the particles in the particle filter projected on a two-dimensional plane as a cluster, wherein the cluster is a region with dense particle distribution in the particle filter, and the parameter of the cluster is an Eps neighborhood, wherein the definition function expression of the Eps neighborhood is as follows:
NEps(p)＝{q∈D|dist(p,q)≤Eps}
in the above formula, NEps(p) representing an Eps neighborhood of a sample p given a sample p e D, D being a sample set, dist (p, q) representing a distance between the sample p and q, and Eps being a set distance threshold; the arbitrary cluster C is a non-empty subset of the clustered sample set D, is a maximum density connected sample set derived from a density reachable relation, and satisfies the following conditions: condition 1: for all p, q ∈ D, if p ∈ C and q is reachable from p, the sample q ∈ C; condition 2: for all p, q ∈ C, sample p and sample q are density-connected.
6. An autonomous scent source searching and locating system based on clustering tendency principle is characterized by comprising:
the model building program unit is used for building a gas forward diffusion model and a sensor detection model;
the search process modeling program unit is used for modeling the search process of the odor source and dividing the search process into a source item estimation part, an action decision part and an execution action part;
a particle filter updating program unit for approximating a posterior probability density function for representing the source item estimation by using the particle filter and converting the update of the source item estimation in the searching process of the odor source into the update of the particle filter;
the density clustering program unit is used for clustering the particles in the particle filter projected on the two-dimensional plane by using a density-based clustering algorithm to obtain clustering cluster information;
an action strategy design program unit, which is used for designing a clustering tendency action strategy for the action of the mobile robot by using the obtained clustering cluster information, and comprises the following steps: 5.1) carrying out effective cluster judgment on the obtained cluster: if the proportion of particles in a certain cluster to the total number of particles exceeds a threshold value, judging the cluster as an effective cluster; 5.2) selecting a target cluster: if the effective cluster exists in the scene at a certain source searching stage, the area where the maximum effective cluster is located is the target cluster; if no effective cluster exists in the scene, the whole particle cluster is taken as a target cluster; 5.3) the mobile robot moves to the mean point of the target cluster by a fixed step length, and once the mean point is reached, the whole target cluster is explored by adopting square search until the next target cluster appears or the source searching is terminated.
7. An autonomous scent source searching and locating system based on clustering tendency principle, comprising a computer device, wherein the computer device comprises a microprocessor and a memory which are connected with each other, characterized in that the microprocessor is programmed or configured to execute the steps of the autonomous scent source searching and locating method based on clustering tendency principle according to any one of claims 1 to 5.
8. An autonomous scent source searching and locating system based on clustering tendency principle, comprising a computer device including a microprocessor and a memory connected to each other, wherein the memory stores therein a computer program programmed or configured to execute the autonomous scent source searching and locating method based on clustering tendency principle according to any one of claims 1 to 5.
9. A computer-readable storage medium, wherein a computer program is stored in the computer-readable storage medium, the computer program being programmed or configured to perform the cluster tropism principle based scent source autonomous searching and locating method according to any one of claims 1 to 5.
CN202011232403.0A 2020-11-06 2020-11-06 Autonomous searching and positioning method for odor sources based on clustering tendency principle Active CN112488153B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011232403.0A CN112488153B (en) 2020-11-06 2020-11-06 Autonomous searching and positioning method for odor sources based on clustering tendency principle

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011232403.0A CN112488153B (en) 2020-11-06 2020-11-06 Autonomous searching and positioning method for odor sources based on clustering tendency principle

Publications (2)

Publication Number Publication Date
CN112488153A CN112488153A (en) 2021-03-12
CN112488153B true CN112488153B (en) 2022-07-05

Family

ID=74928900

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011232403.0A Active CN112488153B (en) 2020-11-06 2020-11-06 Autonomous searching and positioning method for odor sources based on clustering tendency principle

Country Status (1)

Country Link
CN (1) CN112488153B (en)

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113313191A (en) * 2021-06-13 2021-08-27 西北工业大学 Distributed aperture interactive intelligent evaluation method based on unsupervised learning
CN113704690A (en) * 2021-08-26 2021-11-26 无锡赛弗安全装备有限公司 Real-time monitoring method and system for intelligent storage cabinet of hazardous chemical substances
CN114131600A (en) * 2021-12-01 2022-03-04 中国人民解放军国防科技大学 Method and system for generating robot source search scheme based on Gaussian mixture model
CN114154383A (en) * 2021-12-01 2022-03-08 中国人民解放军国防科技大学 Multi-robot-source search scheme generation method and system based on cognitive search strategy
CN114326755B (en) * 2022-03-15 2022-06-07 中国人民解放军国防科技大学 Robot gas source searching method based on local path planning

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2007057893A2 (en) * 2005-11-15 2007-05-24 Yissum Research Development Company Of The Hebrew University Of Jerusalem Method and system for producing a video synopsis
CN106197859A (en) * 2016-07-04 2016-12-07 重庆大学 A kind of gas source localization method considering that the confined space retrains
CN107705288A (en) * 2017-09-04 2018-02-16 武汉工程大学 Hazardous gas spillage infrared video detection method under pseudo- target fast-moving strong interferers
CN110514567A (en) * 2019-08-28 2019-11-29 哈尔滨工业大学 Gas source searching method based on comentropy
CN111563920A (en) * 2020-04-15 2020-08-21 西安工程大学 3D color point cloud registration method based on global optimization and multi-constraint condition iteration
CN111612033A (en) * 2020-04-15 2020-09-01 广东电网有限责任公司 Distribution transformer fault diagnosis method based on gravity search and density peak clustering

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR101541495B1 (en) * 2012-08-17 2015-08-05 네이버 주식회사 Apparatus, method and computer readable recording medium for analyzing a video using the image captured from the video

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2007057893A2 (en) * 2005-11-15 2007-05-24 Yissum Research Development Company Of The Hebrew University Of Jerusalem Method and system for producing a video synopsis
CN106197859A (en) * 2016-07-04 2016-12-07 重庆大学 A kind of gas source localization method considering that the confined space retrains
CN107705288A (en) * 2017-09-04 2018-02-16 武汉工程大学 Hazardous gas spillage infrared video detection method under pseudo- target fast-moving strong interferers
CN110514567A (en) * 2019-08-28 2019-11-29 哈尔滨工业大学 Gas source searching method based on comentropy
CN111563920A (en) * 2020-04-15 2020-08-21 西安工程大学 3D color point cloud registration method based on global optimization and multi-constraint condition iteration
CN111612033A (en) * 2020-04-15 2020-09-01 广东电网有限责任公司 Distribution transformer fault diagnosis method based on gravity search and density peak clustering

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
"Adaptive Methods for Determining DBSCAN Parameters";Kedar Sawant;《IJISET - International Journal of Innovative Science, Engineering & Technology》;20140630;第1卷(第4期);全文 *
"Entrotaxis-Jump as a hybrid search algorithm for seeking an unknown emission source in a large-scale area with road network constraint ";Yong Zhao.et al;《Expert Systems with Applications》;20200501;全文 *
"Searching the diffusive source in an unknown obstructed environment by cognitive strategies with forbidden areas";Yong Zhao.et al;《Building and Environment》;20200930;全文 *
"气体扩散数学模型在安全评价方面的应用";任建国等;《中国安全科学学报》;20060331;第16卷(第3期);全文 *

Also Published As

Publication number Publication date
CN112488153A (en) 2021-03-12

Similar Documents

Publication Publication Date Title
CN112488153B (en) Autonomous searching and positioning method for odor sources based on clustering tendency principle
Wang et al. Locating hazardous gas leaks in the atmosphere via modified genetic, MCMC and particle swarm optimization algorithms
CN109902374B (en) Sudden pollution tracing method based on flight sensor patrol track optimization
CN112231964B (en) Gas leakage source autonomous searching and positioning method based on deep reinforcement learning
CN113359849B (en) Multi-unmanned aerial vehicle collaborative rapid search method for moving target
JP6749282B2 (en) Human flow rate prediction device, human flow rate prediction method, and human flow rate prediction program
CN112214689A (en) Method and system for maximizing influence of group in social network
CN110060738B (en) Method and system for predicting bacterial protective antigen protein based on machine learning technology
CN114047770A (en) Mobile robot path planning method for multi-inner-center search and improvement of wolf algorithm
Rahbar et al. An algorithm for odor source localization based on source term estimation
CN110334449A (en) A kind of aerofoil profile Fast design method based on online agent model algorithm
CN114611801A (en) Traveler problem solving method based on improved whale optimization algorithm
CN109084777A (en) Particle filtering and map constructing method based on self-adapted genetic algorithm
CN111505205A (en) Improved search algorithm for strong back calculation of gas leakage source
Jacquemart et al. Adaptive interacting particle system algorithm for aircraft conflict probability estimation
Bufi et al. Robust optimization of supersonic ORC nozzle guide vanes
CN114329996A (en) Method for constructing pollutant point source diffusion model based on VBEM method
CN114154383A (en) Multi-robot-source search scheme generation method and system based on cognitive search strategy
CN116805015B (en) Bird migration route graph theory modeling method based on GPS tracking data
CN115907079B (en) Airspace traffic flow prediction method based on attention space-time diagram convolutional network
Ma et al. Improved centrifugal compressor impeller optimization with a radial basis function network and principle component analysis
Cruz et al. Machine Learning-Based Indoor Air Quality Baseline Study of the Offices and Laboratories of the Northwest and Southwest Building of Mapúa University–Manila
Ojeda et al. Robotic Gas Source Localization with Probabilistic Mapping and Online Dispersion Simulation
CN113188243A (en) Comprehensive prediction method and system for air conditioner energy consumption
Ye et al. CBSC: A Crowdsensing system for automatic calibrating of barometers

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant