CN110598804B - Improved FastSLAM method based on clustering and membrane calculation - Google Patents
Improved FastSLAM method based on clustering and membrane calculation Download PDFInfo
- Publication number
- CN110598804B CN110598804B CN201910973115.1A CN201910973115A CN110598804B CN 110598804 B CN110598804 B CN 110598804B CN 201910973115 A CN201910973115 A CN 201910973115A CN 110598804 B CN110598804 B CN 110598804B
- Authority
- CN
- China
- Prior art keywords
- particle
- particles
- weight
- pose
- robot
- 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
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
- G01C21/20—Instruments for performing navigational calculations
-
- 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
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Automation & Control Theory (AREA)
- Control Of Position, Course, Altitude, Or Attitude Of Moving Bodies (AREA)
Abstract
The invention discloses an improved FastSLAM method based on clustering and membrane calculation, which comprises the following steps: the functions of the robot for quick self-positioning and environment map construction are realized through a series of processes of initialization, sampling, clustering, membrane calculation optimization, weight calculation, pose calculation, map updating and resampling. The method uses a clustering method to preprocess the particle set and uses the high parallelism of a film calculation optimization method, so that the particle searching speed is increased, the searching range is enlarged, the particle degradation condition is slowed down, the diversity of particles is ensured, the particles are promoted to be distributed near the real pose, and the accuracy and the speed of robot positioning and mapping are effectively improved.
Description
Technical Field
The invention relates to an improved method of a FastSLAM method for simultaneously positioning and mapping robots based on clustering and film calculation.
Background
Meanwhile, positioning and mapping (SLAM) refers to positioning a mobile robot through a sensor carried by the mobile robot in an unknown environment, and mapping surrounding environments at the same time, and is a key technology for performing tasks such as autonomous navigation and obstacle avoidance. At present, the most common method for solving the SLAM problem is a probability-based method, wherein the most prominent expression is a fast simultaneous localization and mapping (FastSLAM) method, and the FastSLAM has low computational complexity and robust data association capability, breaks through the limit of a Gaussian environment, and is widely used. However, frequent resampling by fastsslam can lead to a loss of diversity in the particle set of the robot estimated pose, and positioning accuracy can be degraded over time.
In recent years, researchers have continuously proposed improved methods to try to solve this problem. The Unscented FastSLAM method utilizes unscented transforms for particle filtering, feature initialization, and feature estimation. The filtering precision is greatly improved by using Unscented Particle Filtering (UPF) to replace an extended Kalman filtering method (EKF) in FastSLAM. However, this method is computationally intensive and time consuming, resulting in inefficiency. The PSOFastSLAM method proposed by Chang et al utilizes particle swarm to optimize FastSLAM, but the particle swarm method is easy to fall into local optimum and particle degradation, so that the precision of the FastSLAM method is not greatly improved, particle sets are only attracted by one optimum particle, the particle sets are possibly gathered in a small range after optimization, the search range is narrowed, and therefore, the robot has insufficient stability under the condition of state mutation. Cugliari et al use unscented transforms to improve the accuracy of the particle filter and feature estimator and propose an adaptive selective resampling technique. These techniques have shown better performance than conventional fastsslam. However, these methods are not effective in alleviating the problem of particle degradation.
Disclosure of Invention
In view of the above problems, it is an object of the present invention to provide an improved fastsslam method based on clustering and membrane computation. According to the method, the particles are effectively pre-classified by using a clustering method, the parallelism and the distribution type characteristics of film calculation are used, under the condition of the same number of particles, the searching speed of a faster global optimal value can be obtained, more particles can be distributed in the same running time, the local searching precision of the method is ensured, the searching range is enlarged, the diversity of global searching is met, the predicted particles are more rapidly distributed near the real position, the particle degradation is slowed down, and the positioning precision and speed of the robot are effectively improved.
In order to achieve the above purpose, the technical scheme adopted by the invention is as follows:
an improved fastsslam method based on clustering and membrane computation, the method comprising the steps of:
(1) Initializing, and acquiring the pose x of the robot at the current moment t t Map of environment map feature t From time t to time t+1, the robot motion amount u t Observed quantity z t Observed quantity noise covariance matrix H t ；
(2) Sampling at x t L particles are randomly arranged around the robot, each particle represents a possible pose of the robot, and then according to the proposed probability distribution
Wherein (k) represents the kth pose, sampling the possible poses of the robot at the time t+1, and obtaining Q particles to form a particle set phi, wherein the particle weight of the particle set phi is larger when the particle weight is closer to the proposed distribution t Let the weight of each particle be w (k) ，k∈[1,2,…,Q]：
(3) Clustering, according to the weight of each particle, the particle group phi t All particles in the group are clustered into M groups;
(3a) Cluster initialization, particle set phi t Q particles are added, M particles are randomly selected as initial group centers of M groups, and marked as
(m) represents the mth group, N represents the number of iterations, the initial value of the number of iterations is set to n=0, and the maximum number of iterations is set to N k-means ；
(3b) Clustering and calculating phi t Respectively to other particles in
Weight distance dis of (2) (km) ＝||w (k) -w (m) || 2 Wherein dis (km) Represents the weight distance, w, of particle k to particle m (m) Weight of center particle of mth group, w (k) Is phi t The kth particle weight except the center particle of the group, each particle selects the group with the smallest weight distance from the center of the group to become the group member;
(3c) Ending the clustering, if the particle members in the M groups are not changed, ending the clustering, and turning to the step (4), otherwise, if n= N k-means Ending the clustering, turning to the step (4), otherwise, respectively calculating the average value of all particles in the M groups, wherein the iteration times are n=n+1
As new population center->
Returning to the step (3 b);
(4) Film calculation optimization, for clustered particle set phi t Adopting a film calculation optimization method to enable particle distribution in the particle group to be closer to probability distribution of the actual robot pose;
(4a) Optimization initialization, construction of cell type membrane system
∏＝(V,T,μ,Q,W 0 ,W 2 ,…,W M ,R 0 ,R 1 ,…,R M ,i o )
Wherein V represents all particle objects within the cell, i.e. particle set φ t T represents the output object set, i.e. all possible optimal particles obtained by membrane calculation optimization, μ represents the membrane structure of the cell membrane as [ [ 0 [ 1 ] 1 ,[ 2 ] 2 ,···,[ M ] M ] 0 ，W 0 ～W M Each cell membrane is represented by a surface layer membrane W 0 M basic films W m ，(m∈[1,2,…,M])，R m (m∈[0,1,…,M]) For each membrane evolution rule, i o Surface film W representing completion of model operation 0 Final output result, set particle set phi t The M groups after clustering are respectively put into M basic films, and D is arranged in the M basic films m ,m∈[1,2,...,M]Individual particles, set learning factor c 1 、c 2 Taking the random number between (0, 1) as r 1 And r 2 Setting the initial iteration number n=0, the maximum iteration number N and the velocity v of each particle (md) Pose is x (md) Wherein D is [1,2, …, D m ]，m∈[1,2,…,M](md) represents the d-th particle in the m-th basic film, and the diversity threshold delta is set div Fitness threshold delta f ；
(4b) All basic films W m ，(m∈[1,2,…,M]) All particles in the respective membranes are searched simultaneously,searching for weight optimal particles, and marking as pbest (m) ,(m∈[1,2,…,M]) And deliver it to the surface film W 0 In (a) and (b);
(4c) Surface film to all pbest (m) ,(m∈[1,2,…,M]) Searching for weight optimal particles, marking the searched optimal particles as gbest, and returning the gbest to each basic film W m ,(m∈[1,2,…,M]) In (a) and (b);
(4d) Each basic film W m ，(m∈[1,2,…,M]) According to the formula
Calculating the diversity div of the member particles m ，(m∈[1,2,…,M]) According to pbest (m) ，(m∈[1,2,…,M]) And the value of gbest is calculated by the formula:
speed v of progress (md) And pose x (md) Is updated according to the update of (a);
(4e) According to the formula
And->
Calculating fitness value, wherein +.>
Representing the estimated value of the robot pose iterated to n+1 times>
Indicating that iteration to n+1 times the robot is in +.>
Estimated value of observed quantity on pose, its size is defined by +.>
Map information map at time t t Determining f fitness Representing the fitness value of the optimized particle, if f fitness ≥δ f The optimization is ended, go to step (5), if f fitness ＜δ f Judging whether the iteration number N reaches the maximum iteration number N, if n= N, finishing optimization, turning to step (5), if N<N, then n=n+1, return (4 b), enter the optimization iteration of the next time;
(5) Calculating the weight, wherein the weight of each particle at the time t+1 is calculated as follows:
the larger the weight value is, the higher the probability that the particle is close to the real value is;
(6) Calculating pose and updating a map, and selecting particles with the largest weight as pose x closest to a true value at moment t+1 of the robot t+1 According to p (map t+1 |x t+1 ,z t ) Updating the map feature map of the environment at time t+1 by the calculated value t+1 ；
(7) Resampling according to
Calculating the degree of degradation of the particles, if N eff Not to resample the particle set if Q/2 is not exceeded, the particle set phi t Copying the particles with larger medium weight to a new particle set to form a particle set phi at the time t+1 t+1 The method comprises the steps of carrying out a first treatment on the surface of the If N eff If Q/2 is less than that of the step (2), resampling is needed, and the sampling process is the same as that of the step (2) to obtain a particle set phi at the time t+1 t+1 In the obtained phi t+1 And (3) carrying out subsequent pose estimation and environment map updating of the robot.
The beneficial effects of the invention are as follows:
the autonomous navigation of the robot firstly needs accurate position information and environment map information, the FastSLAM method can realize the self-positioning of the robot and the construction of the environment map, but the robot uses the preset probability distribution, and the pose and the map information are acquired by the particle filtering method, so that larger errors are easy to occur, the running speed is low, and the efficiency is low. The method combines the clustering method and the film computing method with the FastSLAM method, utilizes strong parallelism and flexible evolution rules of film computing, can distribute more particles under the same running time, expands the searching range, increases the diversity of global searching, and ensures the searching result of global optimal particles. The particle set is preprocessed by using the clustering method, so that particle groups can be reasonably divided, the film calculation process is optimized, the particle degradation is slowed down, the search of global optimal particles is accelerated, predicted particles are promoted to be distributed near the real pose, and the accuracy and the speed of robot positioning and mapping are effectively improved.
Drawings
FIG. 1 is a block diagram of the overall flow of the method of the present invention.
FIG. 2 is a flow chart of a clustering method of the present invention.
FIG. 3 is a schematic diagram of a membrane computing system of the method of the present invention.
FIG. 4 is a flow chart of a method for optimizing membrane calculations in the method of the present invention.
Detailed Description
As shown in fig. 1, a process of the improved fastsslam method based on clustering and membrane computation is:
(1) Initializing, and acquiring the pose x of the robot at the current moment t t Map of environment map feature t From time t to time t+1, the robot motion amount u t Observed quantity z t Observed quantity noise covariance matrix H t ；
(2) Sampling at x t L particles are randomly arranged around the robot, each particle represents a possible pose of the robot, and then according to the proposed probability distribution x t (+k 1 )～p(x t+1 |x t (k) ,z t ,u t ) Wherein (k) represents the kth pose, sampling the possible poses of the robot at the time t+1, and obtaining Q particles by increasing the particle weight of the proposed distribution when the robot is closer to the proposed distributionForm particle set phi t Let the weight of each particle be w (k) ，k∈[1,2,…,Q]：
(3) Clustering, according to the weight of each particle, the particle group phi t All particles in the group are clustered into M groups;
(4) Film calculation optimization, for clustered particle set phi t Adopting a film calculation optimization method to enable particle distribution in the particle group to be closer to probability distribution of the actual robot pose;
(5) Calculating the weight, wherein the weight of each particle at the time t+1 is calculated as follows: w (w) t (+m 1 d)＝w t (md) ·p(z t |map t ,x t (md) ,u t ) The larger the weight value is, the higher the probability that the particle is close to the true value is;
(6) Calculating pose and updating a map, and selecting particles with the largest weight as pose x closest to a true value at moment t+1 of the robot t+1 According to p (map t+1 |x t+1 ,z t ) Updating the map feature map of the environment at time t+1 by the calculated value t+1 ；
(7) Resampling according to
Calculating the degree of degradation of the particles, if N eff Not to resample the particle set if Q/2 is not exceeded, the particle set phi t Copying the particles with larger medium weight to a new particle set to form a particle set phi at the time t+1 t+1 The method comprises the steps of carrying out a first treatment on the surface of the If N eff If Q/2 is less than that of the step (2), resampling is needed, and the sampling process is the same as that of the step (2) to obtain a particle set phi at the time t+1 t+1 In the obtained phi t+1 And (3) carrying out subsequent pose estimation and environment map updating of the robot.
As shown in fig. 2, the clustering method of the present invention specifically comprises the following steps:
(3a) Cluster initialization, particle set phi t Q particles are added, M particles are randomly selected as initial group centers of M groups, and marked as
(m) represents the mth group, N represents the number of iterations, the initial value of the number of iterations is set to n=0, and the maximum number of iterations is set to N k-means ；
(3b) Clustering and calculating phi t Respectively to other particles in
Weight distance dis of (2) (km) ＝||w (k) -w (m) || 2 Wherein dis (km) Represents the weight distance, w, of particle k to particle m (m) Weight of center particle of mth group, w (k) Is phi t The kth particle weight except the center particle of the group, each particle selects the group with the smallest weight distance from the center of the group to become the group member;
(3c) Ending the clustering, if the particle members in the M groups are not changed, ending the clustering, and turning to the step (4), otherwise, if n= N k-means Ending the clustering, turning to the step (4), otherwise, respectively calculating the average value of all particles in the M groups, wherein the iteration times are n=n+1
As new population center->
Returning to the step (3 b).
As shown in fig. 3, the film calculation specific structure of the method is as follows: surface film W 0 The surface layer film contains M basic films W 1 ,W 2 ,W 3 ,…,W M The M groups divided after particle aggregation are respectively placed into M basic films, and the M basic films comprise D m ，m∈[1,2,…,M]Particles, X (md) The d particle represented in the m-th basic film, R m ，m∈[1,2,…,M]Representing the particle evolution rules in the mth basic film, searching the optimal particle pbest in the respective film for each basic film (m) ，m∈[1,2,…,M]And outputs it to the surface layer film W 0 Wherein R is 0 Represents the surface layer film W 0 Is search for evolutionary rule ofAll pbest in the S-layer film (m) ，m∈[1,2,…,M]Obtain the optimal value gbest and pass the pbest (m) ，m∈[1,2,…,M]And gbest is returned to each of the base films, respectively;
as shown in fig. 4, the membrane calculation process of the method of the invention is as follows:
(4a) Optimization initialization, construction of cell type membrane system
Π＝(V,T,μ,Q,W 0 ,W 2 ,…,W M ,R 0 ,R 1 ,…,R M ,i o )
Wherein V represents all particle objects within the cell, i.e. particle set φ t T represents the output object set, i.e. all possible optimal particles obtained by membrane calculation optimization, μ represents the membrane structure of the cell membrane as [ [ 0 [ 1 ] 1 ,[ 2 ] 2 ,···,[ M ] M ] 0 ，W 0 ～W M Each cell membrane is represented by a surface layer membrane W 0 M basic films W m ，(m∈[1,2,…,M])，R m (m∈[0,1,…,M]) For each membrane evolution rule, i o Surface film W representing completion of model operation 0 Final output result, set particle set phi t The M groups after clustering are respectively put into M basic films, and D is arranged in the M basic films m ,m∈[1,2,...,M]Individual particles, set learning factor c 1 、c 2 Taking the random number between (0, 1) as r 1 And r 2 Setting the initial iteration number n=0, the maximum iteration number N and the velocity v of each particle (md) Pose is x (md) Wherein D is [1,2, …, D m ]，m∈[1,2,…,M](md) represents the d-th particle in the m-th basic film, and the diversity threshold delta is set div Fitness threshold delta f ；
(4b) All basic films W m ，(m∈[1,2,…,M]) Searching all particles in the respective membranes simultaneously, searching for particles with optimal weight, and marking as pbest (m) ,(m∈[1,2,…,M]) And deliver it to the surface film W 0 In (a) and (b);
(4c) Surface film to all pbest (m) ,(m∈[1,2,…,M]) Searching for weight optimal particles, marking the searched optimal particles as gbest, and returning the gbest to each basic film W m ,(m∈[1,2,…,M]) In (a) and (b);
(4d) Each basic film W m ，(m∈[1,2,…,M]) According to the formula
Calculating the diversity div of the member particles m ，(m∈[1,2,…,M]) According to pbest (m) ，(m∈[1,2,…,M]) And the value of gbest is calculated by the formula:
speed v of progress (md) And pose x (md) Is updated according to the update of (a);
(4e) According to the formula
And->
Calculating fitness value, wherein +.>
Representing the estimated value of the robot pose iterated to n+1 times>
Indicating that iteration to n+1 times the robot is in +.>
Estimated value of observed quantity on pose, its size is defined by +.>
Map information map at time t t Determining f fitness Representing the fitness value of the optimized particle, if f fitness ≥δ f The optimization is ended, go to step (5), if f fitness ＜δ f Judging whether the iteration number N reaches the maximum iteration number N, if n= N, finishing optimization, turning to step (5), if N<N, then n=n+1, return to (4 b), enter the next optimization iteration. />
Claims (1)
1. An improved fastsslam method based on clustering and membrane computation, the method comprising the steps of:
(1) Initializing, and acquiring the pose x of the robot at the current moment t t Map of environment map feature t From time t to time t+1, the robot motion amount u t Observed quantity z t Observed quantity noise covariance matrix H t ；
(2) Sampling at x t L particles are randomly arranged around the robot, each particle represents a possible pose of the robot, and then according to the proposed probability distribution
Wherein (k) represents the kth pose, sampling the possible poses of the robot at the time t+1, and obtaining Q particles to form a particle set phi, wherein the particle weight of the particle set phi is larger when the particle weight is closer to the proposed distribution t Let the weight of each particle be w (k) ，k∈[1,2,…,Q]：
(3) Clustering, according to the weight of each particle, the particle group phi t All particles in the group are clustered into M groups;
(3a) Cluster initialization, particle set phi t Q particles are added, M particles are randomly selected as initial group centers of M groups, and marked as
(m) represents the mth group, N represents the number of iterations, the initial value of the number of iterations is set to n=0, and the maximum number of iterations is set to N k-means ；
(3b) Clustering and calculating phi t Respectively to other particles in
Weight distance dis of (2) (km) ＝||w (k) -w (m) || 2 Wherein dis (km) Represents the weight distance, w, of particle k to particle m (m) Weight of center particle of mth group, w (k) Is phi t The kth particle weight except the center particle of the group, each particle selects the group with the smallest weight distance from the center of the group to become the group member;
(3c) Ending the clustering, if the particle members in the M groups are not changed, ending the clustering, and turning to the step (4), otherwise, if n= N k-means Ending the clustering, turning to the step (4), otherwise, respectively calculating the average value of all particles in the M groups, wherein the iteration times are n=n+1
As new population center->
Returning to the step (3 b);
(4) Film calculation optimization, for clustered particle set phi t Adopting a film calculation optimization algorithm to enable particle distribution in the particle group to be closer to probability distribution of the actual robot pose;
(4a) Optimization initialization, construction of cell type membrane system
Π＝(V,T,μ,Q,W 0 ,W 2 ,…,W M ,R 0 ,R 1 ,…,R M ,i o )
Wherein V represents all particle objects within the cell, i.e. particle set φ t T represents the output object set, i.e. all possible optimal particles obtained by membrane calculation optimization, μ represents the membrane structure of the cell membrane as [ [ 0 [ 1 ] 1 ,[ 2 ] 2 ,···,[ M ] M ] 0 ，W 0 ～W M Each cell membrane is represented by a surface layer membrane W 0 M basic films W m ，(m∈[1,2,…,M])，R m (m∈[0,1,…,M]) For each membrane evolution rule, i o Surface film W representing completion of model operation 0 Final output result, set particle set phi t The M groups after clustering are respectively put into M basic films, and D is arranged in the M basic films m ,m∈[1,2,...,M]Individual particles, set learning factor c 1 、c 2 Taking the random number between (0, 1) as r 1 And r 2 Setting the initial iteration number n=0, the maximum iteration number N and the velocity v of each particle (md) Pose is x (md) Wherein D is [1,2, …, D m ]，m∈[1,2,…,M](md) represents the d-th particle in the m-th basic film, and the diversity threshold delta is set div Fitness threshold delta f ；
(4b) All basic films W m ，(m∈[1,2,…,M]) Searching all particles in the respective membranes simultaneously, searching for particles with optimal weight, and marking as pbest (m) ,(m∈[1,2,…,M]) And deliver it to the surface film W 0 In (a) and (b);
(4c) Surface film to all pbest (m) ,(m∈[1,2,…,M]) Searching for weight optimal particles, marking the searched optimal particles as gbest, and returning the gbest to each basic film W m ,(m∈[1,2,…,M]) In (a) and (b);
(4d) Each basic film W m ，(m∈[1,2,…,M]) According to the formula
Calculating the diversity div of the member particles m ，(m∈[1,2,…,M]) According to pbest (m) ，(m∈[1,2,…,M]) And the value of gbest is calculated by the formula:
speed v of progress (md) And pose x (md) Is updated according to the update of (a);
(4e) According to the formula
And->
Calculating fitness value, wherein +.>
Representing the estimated value of the robot pose iterated to n+1 times>
Indicating that iteration to n+1 times the robot is in +.>
Estimated value of observed quantity on pose, its size is defined by +.>
Map information map at time t t Determining f fitness Representing the fitness value of the optimized particle, if f fitness ≥δ f The optimization is ended, go to step (5), if f fitness ＜δ f Judging whether the iteration number N reaches the maximum iteration number N, if n= N, finishing optimization, turning to step (5), if N<N, then n=n+1, return (4 b), enter the optimization iteration of the next time;
(5) Calculating the weight, wherein the weight of each particle at the time t+1 is calculated as follows:
the larger the weight value is, the higher the probability that the particle is close to the real value is;
(6) Calculating pose and updating a map, and selecting particles with the largest weight as pose x closest to a true value at moment t+1 of the robot t+1 According to p (map t+1 |x t+1 ,z t ) Updating the map feature map of the environment at time t+1 by the calculated value t+1 ；
(7) Resampling according to
Calculating the degree of degradation of the particles, if N eff Not to resample the particle set if Q/2 is not exceeded, the particle set phi t Copying the particles with larger medium weight to a new particle set to form a particle set phi at the time t+1 t+1 The method comprises the steps of carrying out a first treatment on the surface of the If N eff If Q/2 is less than that of the step (2), resampling is needed, and the sampling process is the same as that of the step (2) to obtain a particle set phi at the time t+1 t+1 In the obtained phi t+1 And (3) carrying out subsequent pose estimation and environment map updating of the robot. />
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910973115.1A CN110598804B (en) | 2019-10-14 | 2019-10-14 | Improved FastSLAM method based on clustering and membrane calculation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910973115.1A CN110598804B (en) | 2019-10-14 | 2019-10-14 | Improved FastSLAM method based on clustering and membrane calculation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110598804A CN110598804A (en) | 2019-12-20 |
| CN110598804B true CN110598804B (en) | 2023-05-09 |
Family
ID=68867109
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910973115.1A Active CN110598804B (en) | 2019-10-14 | 2019-10-14 | Improved FastSLAM method based on clustering and membrane calculation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110598804B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112560351B (en) * | 2020-12-22 | 2023-08-29 | 合肥工业大学 | Vehicle transverse stable state judging method with higher accuracy |
| CN113485363B (en) * | 2021-08-02 | 2024-02-20 | 安徽理工大学 | Coal mine underground robot multi-step long path planning method based on membrane calculation and RRT |
| CN114779467A (en) * | 2022-04-27 | 2022-07-22 | 吉林大学 | Novel spectrometer membrane system combination selection method based on detector characteristics |
| CN114912610B (en) * | 2022-05-26 | 2024-04-09 | 安徽理工大学 | Single cell membrane optimal solution calculation optimization method based on membrane calculation |
| CN116160444B (en) * | 2022-12-31 | 2024-01-30 | 中国科学院长春光学精密机械与物理研究所 | Mechanical arm kinematics inverse solution optimization method and device based on clustering algorithm |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107246873A (en) * | 2017-07-03 | 2017-10-13 | 哈尔滨工程大学 | A kind of method of the mobile robot simultaneous localization and mapping based on improved particle filter |
| WO2018072351A1 (en) * | 2016-10-20 | 2018-04-26 | 北京工业大学 | Method for optimizing support vector machine on basis of particle swarm optimization algorithm |
| CN109459033A (en) * | 2018-12-21 | 2019-03-12 | 哈尔滨工程大学 | A kind of robot of the Multiple fading factor positions without mark Fast synchronization and builds drawing method |
| CN109556611A (en) * | 2018-11-30 | 2019-04-02 | 广州高新兴机器人有限公司 | A kind of fusion and positioning method based on figure optimization and particle filter |
-
2019
- 2019-10-14 CN CN201910973115.1A patent/CN110598804B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018072351A1 (en) * | 2016-10-20 | 2018-04-26 | 北京工业大学 | Method for optimizing support vector machine on basis of particle swarm optimization algorithm |
| CN107246873A (en) * | 2017-07-03 | 2017-10-13 | 哈尔滨工程大学 | A kind of method of the mobile robot simultaneous localization and mapping based on improved particle filter |
| CN109556611A (en) * | 2018-11-30 | 2019-04-02 | 广州高新兴机器人有限公司 | A kind of fusion and positioning method based on figure optimization and particle filter |
| CN109459033A (en) * | 2018-12-21 | 2019-03-12 | 哈尔滨工程大学 | A kind of robot of the Multiple fading factor positions without mark Fast synchronization and builds drawing method |
Non-Patent Citations (1)
| Title |
|---|
| 基于粒子群优化的同步定位与地图构建;曲丽萍等;《沈阳工业大学学报》;20130621(第04期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110598804A (en) | 2019-12-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110598804B (en) | Improved FastSLAM method based on clustering and membrane calculation | |
| CN109798896B (en) | Indoor robot positioning and mapping method and device | |
| Xia et al. | Loop closure detection for visual SLAM using PCANet features | |
| CN113345008B (en) | Laser radar dynamic obstacle detection method considering wheel type robot position and posture estimation | |
| CN111376273B (en) | Brain-like inspired robot cognitive map construction method | |
| CN110632932A (en) | Local path planning algorithm based on membrane calculation and particle swarm optimization | |
| US11704572B1 (en) | Offloading for gradient computation | |
| CN113591839A (en) | Feature extraction model construction method, target detection method and device | |
| Burgard et al. | Mobile robot map learning from range data in dynamic environments | |
| CN113420648A (en) | Target detection method and system with rotation adaptability | |
| CN114842260B (en) | Point cloud classification method based on blueprint separation convolution | |
| KR102556765B1 (en) | Apparatus and method for visual localization | |
| CN107742295A (en) | A kind of cube star docking reconstructing method of view-based access control model | |
| Islam et al. | A semantically aware place recognition system for loop closure of a visual SLAM system | |
| CN116611468A (en) | Particle swarm search method based on communication-free robots | |
| Hata et al. | Particle filter localization on continuous occupancy maps | |
| CN115187614A (en) | Real-time simultaneous positioning and mapping method based on STDC semantic segmentation network | |
| CN115345932A (en) | Laser SLAM loop detection method based on semantic information | |
| CN114639238A (en) | Urban expressway traffic state estimation method based on normalized spectral clustering algorithm | |
| Lv et al. | An improved FastSLAM 2.0 algorithm based on FC&ASD-PSO | |
| CN114943766A (en) | Relocation method, relocation device, electronic equipment and computer-readable storage medium | |
| Shen et al. | Incremental learning-based land mark recognition for mirco-UAV autonomous landing | |
| Zhang et al. | Vision-based UAV obstacle avoidance algorithm on the embedded platform | |
| CN117058401B (en) | High-precision point cloud classification self-adaptive downsampling method and device for complex environment perception | |
| Song et al. | Recalling direct 2d-3d matches for large-scale visual localization |
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 |