CN108247630B - Mobile robot obstacle avoidance method based on Bayesian network model - Google Patents

Mobile robot obstacle avoidance method based on Bayesian network model Download PDF

Info

Publication number
CN108247630B
CN108247630B CN201711246586.XA CN201711246586A CN108247630B CN 108247630 B CN108247630 B CN 108247630B CN 201711246586 A CN201711246586 A CN 201711246586A CN 108247630 B CN108247630 B CN 108247630B
Authority
CN
China
Prior art keywords
obstacle
mobile robot
angle
node
value
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
CN201711246586.XA
Other languages
Chinese (zh)
Other versions
CN108247630A (en
Inventor
柴慧敏
吕少楠
方敏
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.)
Xidian University
Original Assignee
Xidian University
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 Xidian University filed Critical Xidian University
Priority to CN201711246586.XA priority Critical patent/CN108247630B/en
Publication of CN108247630A publication Critical patent/CN108247630A/en
Application granted granted Critical
Publication of CN108247630B publication Critical patent/CN108247630B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
    • B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
    • B25J9/00—Programme-controlled manipulators
    • B25J9/16—Programme controls
    • B25J9/1602—Programme controls characterised by the control system, structure, architecture
    • B25J9/161—Hardware, e.g. neural networks, fuzzy logic, interfaces, processor
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
    • B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
    • B25J9/00—Programme-controlled manipulators
    • B25J9/16—Programme controls
    • B25J9/1674—Programme controls characterised by safety, monitoring, diagnostic
    • B25J9/1676—Avoiding collision or forbidden zones
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/04—Inference or reasoning models
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N7/00—Computing arrangements based on specific mathematical models
    • G06N7/01—Probabilistic graphical models, e.g. probabilistic networks

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Mathematical Physics (AREA)
  • Evolutionary Computation (AREA)
  • Data Mining & Analysis (AREA)
  • Automation & Control Theory (AREA)
  • Robotics (AREA)
  • Mechanical Engineering (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Computational Mathematics (AREA)
  • Pure & Applied Mathematics (AREA)
  • Mathematical Optimization (AREA)
  • Mathematical Analysis (AREA)
  • Fuzzy Systems (AREA)
  • Algebra (AREA)
  • Probability & Statistics with Applications (AREA)
  • Computational Linguistics (AREA)
  • Control Of Position, Course, Altitude, Or Attitude Of Moving Bodies (AREA)

Abstract

The invention provides a mobile robot obstacle avoidance method based on a Bayesian network model, which solves the problem of insufficient consideration of the distance and the angle of an obstacle in the obstacle avoidance of the mobile robot in the prior art, and comprises the following steps: establishing a mobile robot obstacle avoidance judging rule; establishing a Bayesian network model for obstacle avoidance behavior processing; establishing fuzzy membership functions of distances and angles of obstacle targets around the mobile robot, and performing fuzzification processing on data of the targets; inputting the blurred distance and angle data of the obstacle target into a Bayesian network model for obstacle avoidance as an evidence; and selecting the deviation angle with the maximum posterior probability value as the rotation angle of the mobile robot to avoid the obstacle. According to the method, the Bayesian network model is used in the obstacle avoidance behavior processing of the mobile robot to obtain the angle required to rotate, so that the mobile robot can avoid obstacles in the unknown environment, the obstacle avoidance behavior is more accurate, and the method is used for mobile robot path selection.

Description

Mobile robot obstacle avoidance method based on Bayesian network model
Technical Field
The invention belongs to the technical field of computers, relates to robot obstacle avoidance, and particularly relates to a mobile robot obstacle avoidance method based on a Bayesian network model, which can be used for obstacle avoidance behavior processing of a mobile robot in an unknown environment local path planning process in practical application.
Background
Obstacle avoidance of the mobile robot is a key problem in local path planning. At present, in this aspect, the optimal direction of the mobile robot for avoiding the obstacle is obtained by calculating, mainly by extracting the obstacle target information obtained by the sensor, matching the obstacle target information with the obstacle avoidance rule or giving a cost function. In a real-world scene, it is difficult to accurately provide all predefined rules corresponding to obstacle avoidance behaviors, and uncertain information of an obstacle target is difficult to be completely described in the rules. And calculating an optimal value according to the cost function, wherein if the cost function is complex, the solving process is complex and only a local optimal solution can be obtained. The Bayesian network model is used as a main uncertainty reasoning model in artificial intelligence, has better processing capability on uncertainty information, and can be closely combined with rule knowledge to improve the intelligence of robot obstacle avoidance.
The patent of northern automation control equipment research institute for "a low-cost autonomous obstacle avoidance method of a mobile robot" (patent application number CN201410538970.7, publication number CN105487536A) discloses a low-cost autonomous obstacle avoidance method of a mobile robot. The method mainly comprises the steps that according to the fact that in different directions around the robot, a sensor detects the distance between an obstacle and the robot, whether the distance is within a safe threshold value is judged, the safe threshold value is divided into three levels of far, near and very near, and different avoidance strategies are adopted in different levels. The method disclosed in this patent application has the following disadvantages: obstacle avoidance is performed only according to the fact that the distance between the obstacles is in different safety threshold levels, when the distance between the obstacles and the robot is far, but the included angle between the obstacle and the traveling direction of the robot is small, obstacle avoidance processing is possibly not accurate, and the robot cannot effectively avoid the obstacles.
The patent of Shenyang Xinsong robot automation corporation for 'a mobile robot obstacle avoidance method based on a laser range finder' (patent application No. CN201010611255.3, publication No. CN102541057A) discloses a mobile robot obstacle avoidance method based on a laser range finder. Grouping laser ranging information, and mapping barrier points in each group into a robot coordinate system; the robot is expanded into a circle with the radius of R, two tangent lines of the circle are made through the obstacle points, the feasible direction range of the robot which can pass is obtained by utilizing the included angle between the tangent lines and the X axis of the robot coordinate system, and the optimal direction is further selected by defining a cost function. The method disclosed in this patent application has the disadvantages that: the distance factor of the obstacle from the robot is not fully considered in calculating the feasible direction range of the robot, and the feasible direction range has limitation and influences the selection of the optimal direction.
Zhangqi in its published paper, "path planning and positioning technology research of mobile robot" (university of harbinge industry, doctor's academic thesis, 2014.6), proposes a method for realizing obstacle avoidance of mobile robot based on fuzzy knowledge rule matching. The method establishes a fuzzy rule knowledge base for processing the obstacle avoidance behavior of the mobile robot, performs matching judgment according to the obstacle target information obtained by a sensor of the mobile robot and a fuzzy rule, and adopts a specific obstacle avoidance processing mode of the rule under the condition that the judgment meets a certain rule. The method has the following defects: the obstacle avoidance of the mobile robot is matched with all fuzzy rules every time, the intelligent level is low, and the mobile robot has an intelligent selection function in an unknown complex environment.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a mobile robot obstacle avoidance method based on a Bayesian network model, which is higher in intelligence and accuracy.
The invention relates to a mobile robot obstacle avoidance method based on a Bayesian network model, which is characterized by comprising the following steps of:
(1) establishing a judgment rule of the obstacle avoidance condition of the mobile robot:
the obstacle avoidance problem of the mobile robot is decomposed into 7 cases by taking the coordinate system of the mobile robot as a reference:
(1a) if an obstacle is arranged right in front of the mobile robot and no obstacle exists on the left side and the right side, the mobile robot deviates a certain angle to the right or left to avoid the obstacle;
(1b) if the mobile robot has an obstacle on the left and has no obstacle right in front of the mobile robot, the mobile robot only deviates a certain angle right to avoid the obstacle;
(1c) if the mobile robot has an obstacle on the right side and no obstacle on the left side and right front side, the mobile robot only deviates a certain angle on the left side to avoid the obstacle;
(1d) if the obstacle is positioned right ahead and on the left of the mobile robot and the obstacle is not positioned on the right, the mobile robot only deviates a certain angle right to avoid the obstacle;
(1e) if the obstacle is positioned right in front of the mobile robot and on the right side of the mobile robot, and the obstacle is not positioned on the left side of the mobile robot, the mobile robot only deviates a certain angle to the left to avoid the obstacle;
(1f) if the mobile robot has barriers on the left and right sides and no barrier is in the front of the mobile robot, the mobile robot deviates a certain angle to the right or left to avoid the barriers;
(1g) if the mobile robot has obstacles right in front, to the left and to the right, the mobile robot deviates a certain angle to the right or left to avoid the obstacles;
(2) establishing a Bayesian network model for obstacle avoidance of the mobile robot:
(2a) establishing a Bayesian network structure of the mobile robot for obstacle avoidance by adopting a diagnostic Bayesian network modeling mode according to the judgment rules of 7 conditions of the mobile robot for obstacle avoidance;
(2b) setting each node parameter in the established Bayesian network structure for obstacle avoidance of the mobile robot, wherein the node parameters are set according to the following rules: setting parameter values of the symptom nodes in three states as follows: 1/3 (0.33); setting the parameter value of the result node to be 1 when the condition of the judgment criterion is met, and setting the parameter value of the result node to be 0 when the condition of the judgment criterion is not met;
(3) respectively establishing fuzzy membership functions of the distance and the angle of the obstacle target around the mobile robot, and performing fuzzification processing on the distance and the angle data of the target:
extracting obstacle target information obtained by a sensor, setting a fuzzy set of the obstacle target information as follows according to the distance between the obstacle target and the mobile robot, wherein the unit is centimeter: (near, Gmean) using a decreasing half trapezoidal membership function to give a representation of the fuzzy set; according to the size of the positive included angle between the obstacle target and the X axis of the mobile robot coordinate system, the unit is degree, and the fuzzy set is set as follows: (small, Gsmall), and representing the fuzzy set by adopting a halving trapezoidal membership function;
and calculating the fuzzy membership degree of the distance value on a fuzzy set (near, Gmean) thereof and the fuzzy membership degree of the angle value on a fuzzy set (small, Gmean) thereof according to the established fuzzy membership function of the distance and the angle of the obstacle target.
(4) Inputting the blurred distance and angle data of the obstacle target into a Bayesian network model for obstacle avoidance behavior processing as an evidence, and carrying out Bayesian network reasoning:
(4a) inputting the fuzzy processing result of the distance and the angle of the obstacle target into a Bayesian network model: taking the symptom node as an evidence node, and carrying out fuzzy processing on the distance and the angle of the obstacle target: the membership degrees on the fuzzy set are used as evidence values of corresponding evidence nodes in different states and are respectively input into the Bayesian network model;
(4b) performing Bayesian network inference: adopting a Bayesian network accurate reasoning algorithm to carry out reasoning on the Bayesian network model after the evidence values are input, and calculating posterior probability values of each result node to obtain a Bayesian network reasoning result;
(5) and selecting a result node with the maximum posterior probability value in the Bayesian network inference results, wherein the state value of the result node is used as the rotation angle of the mobile robot to avoid the obstacle.
The invention comprehensively considers the distance and angle information of the barrier target, and the Bayesian network model has stronger processing capability on uncertain information and better intelligence.
Compared with the prior art, the invention has the following advantages:
firstly, the obstacle avoidance behavior of the robot is modeled and processed through a Bayesian network model, the establishment of the model structure and the setting of parameters are closely combined with the rule knowledge of the obstacle avoidance processing of the mobile robot, and uncertain information is processed through a Bayesian inference algorithm, so that the defect of low intelligentization level in the prior art is overcome, and the intelligence of the obstacle avoidance behavior processing of the mobile robot is improved in an unknown complex environment.
Secondly, when obstacle avoidance processing is carried out through the Bayesian network model, distance information and angle information of an obstacle target are comprehensively considered, a coordinate system is determined more appropriately and objectively, a traveling direction rule is specified, the distance information and the angle information can be accurately applied to route planning of robot obstacle avoidance, and therefore accuracy of the mobile robot obstacle avoidance is improved.
Drawings
FIG. 1 is a flow chart of an implementation of the present invention;
FIG. 2 is a diagram of a Bayesian network architecture for use in practicing the present invention.
Detailed Description
The invention is described in detail below with reference to the figures and the specific embodiments
Example 1
The mobile robot is also a very front-end research field, and various countries have fierce competition in this respect and invest in large manpower and material resources, but the existing research and development technology for obstacle avoidance of the mobile robot is not very perfect, and has the defects of low intellectualization level and low accuracy, so that the mobile robot has an accurate and intelligent obstacle avoidance function in an unknown complex environment. Therefore, the invention develops research, provides a method for modeling the obstacle avoidance of the mobile robot by adopting a Bayesian network model, and overcomes the defects.
The invention relates to a mobile robot obstacle avoidance method based on a Bayesian network model, which comprises the following steps of:
(1) establishing a judgment rule of the obstacle avoidance condition of the mobile robot:
the obstacle avoidance problem of the mobile robot is decomposed into 7 cases by taking the coordinate system of the mobile robot as a reference:
(1a) if an obstacle is in front of the mobile robot, and no obstacle exists on the left side and the right side, the mobile robot deviates a certain angle to the right or left to avoid the obstacle.
(1b) If the mobile robot has an obstacle to the left and no obstacle to the right and the front, the mobile robot only deviates to the right by a certain angle to avoid the obstacle.
(1c) If the mobile robot has an obstacle to the right and no obstacle to the left and right in front, the mobile robot only deviates a certain angle to the left to avoid the obstacle.
(1d) If the mobile robot has obstacles right in front of and on the left side and has no obstacles on the right side, the mobile robot deviates a certain angle to the right to avoid the obstacles.
(1e) If the mobile robot has obstacles right in front of and to the right and has no obstacles to the left, the mobile robot deviates a certain angle to the left to avoid the obstacles.
(1f) If the mobile robot has obstacles on the left and right and no obstacle is in the front, the mobile robot deviates a certain angle to the right or left to avoid the obstacle.
(1g) If the mobile robot has obstacles right in front, to the left and to the right, the mobile robot deviates a certain angle to the right or left to avoid the obstacles.
The invention basically covers all possible situations that the mobile robot encounters an obstacle in the traveling process, and divides all the possible situations.
(2) Establishing a Bayesian network model for obstacle avoidance of the mobile robot, referring to FIG. 2, wherein FIG. 2 is a Bayesian network structure diagram used in the embodiment of the invention, and the invention divides the rotation direction of the mobile robot into three types, one is only left-turning, one is only right-turning, and the other is right-turning or left-turning; and the father node of only the right turning angle node is the left obstacle deflection distance and the left obstacle deflection angle, the father node of only the left turning angle node is the right obstacle deflection distance and the right obstacle deflection angle, the father node of the right turning or left turning angle node is the dead ahead obstacle distance, and the father node of only the left turning angle node, only the right turning angle node and the right turning or left turning angle node are the father nodes of the rotation angle nodes, so that the accurate turning of the mobile robot is determined. The concrete description is as follows:
(2a) according to the judgment rule of 7 conditions of obstacle avoidance of the mobile robot, a Bayesian network structure of the obstacle avoidance of the mobile robot is established by adopting a diagnostic Bayesian network modeling mode, a symptom node and a result node are established, and the symptom node is used as a father node of the result node.
(2b) Setting each node parameter in the established Bayesian network structure for obstacle avoidance of the mobile robot, wherein the node parameters are set according to the following rules: setting parameter values of the symptom nodes in three states as follows: 1/3 (0.33); and setting the parameter value of the result node to be 1 when the condition of the judgment criterion is met, and setting the parameter value of the result node to be 0 when the condition of the judgment criterion is not met.
(3) Respectively establishing fuzzy membership functions of the distance and the angle of the obstacle target around the mobile robot, and performing fuzzification processing on the distance and the angle data of the target:
extracting obstacle target information obtained by a sensor, setting a fuzzy set of the obstacle target information as follows according to the distance between the obstacle target and the mobile robot, wherein the unit is centimeter: (near, Gmean) using a decreasing half trapezoidal membership function to give a representation of the fuzzy set; according to the size of the positive included angle between the obstacle target and the X axis of the mobile robot coordinate system, the unit is degree, and the fuzzy set is set as follows: (small, Gsmall), a reduced-half trapezoidal membership function is used to give a representation of the fuzzy set.
And calculating the fuzzy membership degree of the distance value on a fuzzy set (near, Gmean) thereof and the fuzzy membership degree of the angle value on a fuzzy set (small, Gmean) thereof according to the established fuzzy membership function of the distance and the angle of the obstacle target.
(4) Inputting the blurred distance and angle data of the obstacle target into a Bayesian network model for obstacle avoidance behavior processing as an evidence, and carrying out Bayesian network reasoning:
(4a) inputting the fuzzy processing result of the distance and the angle of the obstacle target into a Bayesian network model: taking the symptom node as an evidence node, and carrying out fuzzy processing on the distance and the angle of the obstacle target: and the membership degrees on the fuzzy set are used as evidence values of corresponding evidence nodes in different states and are respectively input into the Bayesian network model. The method only carries out fuzzy processing on the symptom nodes and the evidence nodes, grasps the main veins under the condition of not trapping in complicated data processing, and obtains the accurate angle information in the path of the mobile robot.
(4b) Performing Bayesian network inference: and adopting a Bayesian network accurate reasoning algorithm to reason the Bayesian network model after the evidence values are input, and calculating posterior probability values of all result nodes to obtain a Bayesian network reasoning result.
(5) And selecting a result node with the maximum posterior probability value in the Bayesian network inference results, wherein the state value of the result node is used as the rotation angle of the mobile robot to avoid the obstacle.
According to the invention, the problem of obstacle avoidance of the robot is solved by establishing the Bayesian network model, uncertain information of obstacles encountered in the process of robot traveling is processed by a Bayesian inference algorithm, and compared with the insufficient intelligence in the prior art, the intelligence of obstacle avoidance behavior processing of the mobile robot is greatly improved.
Example 2
The mobile robot obstacle avoidance method based on the bayesian network model is the same as that in embodiment 1, and the mobile robot coordinate system in the step (1) is to establish an XYZ three-dimensional coordinate system with the center of the mobile robot as a coordinate origin, where an X-axis positive direction represents a positive front direction of the mobile robot, an X-axis negative direction represents a positive rear direction of the mobile robot, a Y-axis positive direction represents a positive right direction of the mobile robot, a Y-axis negative direction represents a positive left direction of the mobile robot, a Z-axis positive direction represents a positive lower direction of the mobile robot, and a Z-axis negative direction represents a positive upper direction of the mobile robot. The invention relates to a technology for researching obstacle avoidance of a robot, wherein a three-dimensional coordinate system is established by taking the center of a mobile robot as a coordinate origin, so that the position of an obstacle in the coordinate system and the rotation angle of the robot can be described, and the technology is a basis for solving the problem of obstacle avoidance of the robot.
Example 3
The barrier avoiding method for the mobile robot based on the bayesian network model is the same as that in the embodiment 1-2, and the right front of the mobile robot in the step (1) is that the positive direction of the Y axis of the coordinate system of the mobile robot is taken as a reference, and an included angle between the positive direction of the Y axis and the positive direction of the Y axis is in a range: (60 DEG, 120 DEG).
The left side of the mobile robot in the step (1) refers to that the forward direction of the Y axis of the coordinate system of the mobile robot is taken as a reference, and an included angle between the forward direction of the Y axis and the included angle is in a range: [120 °,160 ° ].
The right-side deviation of the mobile robot in the step (1) refers to that the included angle between the forward direction of the Y axis and the forward direction of the Y axis of a coordinate system of the mobile robot is in a range by taking the forward direction of the Y axis as a reference: [20 °,60 ° ].
In a patent of northern automation control equipment research institute, a low-cost autonomous obstacle avoidance method for a mobile robot (patent application No. CN201410538970.7, publication No. CN105487536A), obstacle avoidance is performed only according to the level of a safety threshold where the distance of an obstacle is different, and when the obstacle is far from the robot but has a small included angle with the traveling direction of the robot, obstacle avoidance processing may not be very accurate, which may cause the robot to be unable to effectively avoid the obstacle. The invention defines the included angle range of a larger interval, considers the possibility of different included angle ranges, and considers both the static condition and the dynamic condition.
The invention not only establishes a coordinate system, but also defines the range of included angles in different directions, wherein the forward direction is the interval with the forward included angle of (60 degrees and 120 degrees) of the Y axis of the coordinate system of the mobile robot, the leftward direction is the interval with the forward included angle of [120 degrees and 160 degrees ] of the Y axis of the coordinate system of the mobile robot, and the rightward direction is the interval with the forward included angle of [20 degrees and 60 degrees ] of the Y axis of the coordinate system of the mobile robot, and the judgment of the directions defines a clear angle interval.
Example 4
The method for avoiding the obstacle of the mobile robot based on the bayesian network model is the same as that in the embodiment 1-3, and the establishing of the bayesian network structure for avoiding the obstacle of the mobile robot by adopting the diagnostic bayesian network modeling mode in the step (2a) specifically comprises the following steps:
2a1. establishing Bayesian network structure symptom nodes: extracting the following description of the obstacle avoidance condition judgment rule of the mobile robot about the obstacle, wherein the obstacle exists right ahead, the obstacle exists leftwards, and the obstacle exists rightwards, and according to the distance and angle attributes of the obstacle, the distance of the left obstacle, the angle of the left obstacle, the distance of the right obstacle, the angle of the right obstacle, and the distance of the obstacle right ahead are used as sign nodes.
2a2, establishing a Bayesian network structure result node: and extracting the following description of the mobile robot obstacle avoidance condition judgment rule about the deviation angle of the mobile robot, wherein the mobile robot deviates a certain angle only leftwards, the mobile robot deviates a certain angle only rightwards, the mobile robot deviates a certain angle rightwards or leftwards, and the deviation angle only leftwards, the deviation angle only rightwards, the deviation angle rightwards or leftwards is taken as a result node.
2a3. A Bayesian network model structure is formed by using the symptom nodes as the father nodes of the result nodes in a diagnostic Bayesian network modeling mode.
2a4. sets all symptom nodes and result nodes in the Bayesian network model as discrete nodes: each node has three different states, and the values of the distance node of the left obstacle, the distance node of the right obstacle and the distance node of the right obstacle are as follows: near, very near, Null; values of the angle node of the left obstacle and the angle node of the right obstacle are as follows: small deviation angle (small), small deviation angle (Gsmall), Null (Null); the right deviation angle node only takes values as: only 30 ° (RS) to the right, only 60 ° (RB) to the right, Null (Null); the left deviation angle node only takes values as: 30 ° (LS) to the left only, 60 ° (LB) to the left only, Null (Null); the right or left deviation angle node takes the values as: 30 ° to the right or Left (LSRS), 60 ° to the right or Left (LBRB), Null (Null).
The Bayesian network modeling method is adopted to establish a Bayesian network structure for obstacle avoidance of the mobile robot, the symptom nodes and the result nodes are established, the symptom nodes are used as father nodes of the result nodes, and therefore a Bayesian network model structure is formed, each node has three different parameter values, and the parameter values are set for the corresponding nodes.
Example 5
The barrier avoiding method for the mobile robot based on the bayesian network model is the same as that in embodiments 1 to 4, and the evidence values of the corresponding evidence nodes in different states in step (4a) refer to:
4a1., taking the fuzzy membership value (near, Gnear) of the distance of the obstacle object existing right ahead as the value of the evidence state near, Gnear corresponding to the obstacle distance node right ahead, and the value of the evidence state null is 0, otherwise, the evidence of the obstacle distance node right ahead is: null takes the value 1.
4a2, taking the fuzzy membership value (near, Gnear) of the distance of the obstacle target existing on the left as the value of the evidence state near, Gnear corresponding to the distance node of the left obstacle, and the value of the evidence state null is 0, otherwise, the evidence of the distance node of the left obstacle is: null takes the value 1.
4a3., using the fuzzy membership value (small, Gsmall) of the angle of the obstacle object existing on the left side as the value of the evidence state small, Gsmall corresponding to the left obstacle angle node, where the value of the evidence state null is 0, and the evidence of the left obstacle angle node under the other conditions is: null takes the value 1.
4a4., taking the fuzzy membership value (near, Gnear) of the distance of the obstacle object existing on the right as the value of the evidence state near, Gnear corresponding to the distance node of the obstacle on the right, and the value of the evidence state null is 0, otherwise, the evidence of the distance node of the obstacle on the right is: null takes the value 1.
4a5., taking the fuzzy membership value (small, Gsmall) of the angle of the obstacle object existing on the right side as the value of the evidence state small, Gsmall corresponding to the right obstacle angle node, wherein the value of the evidence state null is 0, and the evidence of the right obstacle angle node under the other conditions is as follows: null takes the value 1.
The invention carries out specific dereferencing on the evidence values of the corresponding evidence nodes in the Bayesian network model under different states so as to more accurately judge the advancing direction.
Example 6
The barrier avoiding method for the mobile robot based on the bayesian network model is the same as that in the embodiments 1 to 5, and the bayesian network accurate inference algorithm in the step (4b) is to accurately calculate the posterior probability value of the result node through probabilistic inference according to the set parameter value of the symptom node, the parameter value of the result node and the evidence value.
And selecting a result node with the maximum posterior probability value from the Bayesian inference results, and taking the state value of the result node as the angle of rotation of the mobile robot for obstacle avoidance.
The invention adopts the Bayesian network model with better processing capability on the uncertain information, and obtains a processing result with stronger intelligence.
The following detailed description of embodiments of the invention is provided in connection with the accompanying drawings and the examples.
Example 7
The method for avoiding the obstacle of the mobile robot based on the Bayesian network model is the same as the embodiments 1-6, and the method for avoiding the obstacle of the mobile robot based on the Bayesian network model, referring to FIG. 1, comprises the following steps:
step 1) establishing a judgment rule of the obstacle avoidance condition of the mobile robot:
the geometric center of the mobile robot is taken as the origin of coordinates, the positive front direction of the mobile robot is the positive X-axis direction, the positive back direction of the mobile robot is the negative X-axis direction, the positive right direction of the mobile robot is the positive Y-axis direction, the positive left direction of the mobile robot is the negative Y-axis direction, the positive down direction of the mobile robot is the positive Z-axis direction, and the positive up direction of the mobile robot is the negative Z-axis direction, so that a three-dimensional mobile robot coordinate system is established.
Taking the positive direction of the Y axis of the mobile robot coordinate system as a reference, and enabling an included angle between the positive direction of the Y axis and the mobile robot coordinate system to be in a range:
(60 °, 120 °) as being directly in front of the mobile robot; the finger takes the positive direction of the Y axis of the mobile robot coordinate system as a reference, and the included angle between the finger and the positive direction of the Y axis is in the range: [120 °,160 ° ], as the left side of the mobile robot; and taking the positive direction of the Y axis of the mobile robot coordinate system as a reference, wherein an included angle between the positive direction of the Y axis and the positive direction of the Y axis is in a range: [20 °,60 ° ], which is the right side of the mobile robot.
The obstacle avoidance problem of the mobile robot is decomposed into 7 cases by taking the coordinate system of the mobile robot as a reference:
1a, if an obstacle is positioned right in front of the mobile robot and no obstacle is positioned on the left and right, the mobile robot deviates a certain angle to the right or left to avoid the obstacle.
If the mobile robot has an obstacle to the left and no obstacle to the right and the front, the mobile robot deviates a certain angle to the right to avoid the obstacle.
And 1c, if the mobile robot has an obstacle at the right side and does not have an obstacle at the left side and right front side, the mobile robot only deviates a certain angle at the left side to avoid the obstacle.
And 1d, if the mobile robot has obstacles right in front and on the left and has no obstacles on the right, the mobile robot deviates a certain angle right to avoid the obstacles.
And 1e, if the mobile robot has obstacles right in front and at the right side and has no obstacles at the left side, the mobile robot only deviates a certain angle to the left to avoid the obstacles.
And 1f, if the mobile robot has obstacles to the left and the right and has no obstacle right in front, the mobile robot deviates a certain angle to the right or the left to avoid the obstacle.
And 1g, if the mobile robot has obstacles right in front, to the left and to the right, the mobile robot deviates a certain angle to the right or left to avoid the obstacles.
Step 2), establishing a Bayesian network model for obstacle avoidance behavior processing of the mobile robot:
2a) and establishing a Bayesian network structure of the mobile robot for obstacle avoidance by adopting a diagnostic Bayesian network modeling mode according to the judgment rules of 7 conditions of the mobile robot for obstacle avoidance.
2a1. extracting the following description of the obstacle avoidance condition judgment rule of the mobile robot about the obstacle: the right front part is provided with an obstacle, the left side is provided with an obstacle, the right side is provided with an obstacle, and the distance from the left obstacle, the angle from the left obstacle, the distance from the right obstacle, the angle from the right obstacle and the distance from the right obstacle are used as sign nodes according to the distance and angle attributes of the obstacles.
The symptom node comprises: the system comprises a left obstacle distance node, a left obstacle angle node, a right obstacle distance node, a right obstacle angle node and a front obstacle distance node.
2a2, extracting the following description of the mobile robot obstacle avoidance condition judgment rule about the deviation angle of the mobile robot: the mobile robot deviates only a certain angle to the left, the mobile robot deviates only a certain angle to the right, the mobile robot deviates a certain angle to the right or left, and the deviation of the angle to the right, the deviation of the angle to the left, the deviation of the angle to the right or the left are taken as result nodes.
The result node includes: right-turn angle only nodes, left-turn angle only nodes, right-turn or left-turn angle nodes.
2a3. A Bayesian network model structure is formed by using the symptom nodes as father nodes of the result nodes and the result nodes as child nodes of the symptom nodes in a diagnostic Bayesian network modeling mode.
Referring to fig. 2, the established bayesian network model structure is specifically: the parent nodes of the right-hand corner-only node (RTurn) are: a left obstacle distance node (LDistance), a left obstacle angle node (Langle); the parent of the left turn angle node (LTurn) only is: a right obstacle distance node (RDistance), a right obstacle angle node (RAMgle); the parent node of the right or left turn angle node (FTurn) is: frontal obstacle distance (FDistance); adding a rotation angle node (Turn), and connecting the result node: and a right-turn angle node (RTurn) only, a left-turn angle node (LTurn) only, and a right-turn or left-turn angle node (FTurn) as parent nodes of the rotation angle node.
2a4. sets all symptom nodes and result nodes in the Bayesian network model as discrete nodes: the values of the left obstacle distance node, the right obstacle distance node and the right front obstacle distance node are as follows: near, very near, Null; values of the angle node of the left obstacle and the angle node of the right obstacle are as follows: small deviation angle (small), small deviation angle (Gsmall), Null (Null); the right turn angle node takes the values as: 30 ° to the Right (RS), 60 ° to the Right (RB), Null (Null); the left-turn angle node values are as follows: 30 ° left offset (LS), 60 ° left offset (LB), Null (Null); the left-turn or right-turn deviation angle node values are as follows: 30 ° left or right (LSRS), 60 ° left or right (LBRB), Null (Null).
Setting the rotation angle node (Turn) as an eight-value discrete node, wherein the value state is as follows: 30 ° (RS) to the right, 60 ° (RB) to the right, 30 ° (LS) to the left, 60 ° (LB) to the left, 30 ° (LSRS) to the left or right, 60 ° (LBRB) to the left or right, 0 ° (Zero), Null (Null).
2b) Setting parameters of each node in the established Bayesian network structure for obstacle avoidance of the mobile robot:
setting parameter values of the symptom nodes in three states as follows: (0.33,0.33,0.34).
Setting the parameter value of the result node to be 1 under the condition that the judgment criterion of the angular rotation size of the mobile robot is met, setting the parameter value under the condition that the judgment criterion of the angular rotation size of the mobile robot is not met to be 0, wherein the judgment criterion of the angular rotation size of the mobile robot is specifically as follows:
and 2b1, if the distance node of the left-hand obstacle is very close (Gnear) or the angle of the left-hand obstacle is very small (Gsmall), the parameter of the right-hand rotation angle node RB is 1, otherwise, the parameter is 0.
2b2., if the left obstacle distance node is near (near) and the left obstacle angle is small (small), the right rotation angle node is RS, the parameter is 1, otherwise, it is 0.
2b3. if the left-hand obstacle distance node is Null or the left-hand obstacle angle is Null, the right-hand rotation angle node is Null with a parameter of 1, otherwise 0.
2b4. if the distance node of the right-hand obstacle is very close (Gnear) or the angle of the right-hand obstacle is very small (Gsmall), the parameter of the left-hand turning angle node LB is 1, otherwise it is 0.
2b5., if the distance node of the right obstacle is near (near) and the angle of the right obstacle is small (small), the parameter of the node of the left turn angle RS is 1, otherwise, it is 0.
2b6. if the distance node of the right-inclined obstacle is Null (Null) or the angle of the right-inclined obstacle is Null (Null), the parameter of the node of the left-turning angle is Null is 1, otherwise, it is 0.
2b7. if the distance node of the front obstacle is very close (Gnear), the parameter of the right turn or left turn angle node (FTurn) LBRB is 1, otherwise it is 0.
2b8. if the distance node of the front obstacle is near (near), the parameter of the right turn or left turn angle node (FTurn) is LSRS is 1, otherwise it is 0.
2b9. if the distance node of the front obstacle is Null (Null), the parameter of the right turn or left turn angle node (FTurn) is Null 1, otherwise it is 0.
The parameters of the rotation angle node (FTurn) are set as:
A. under the condition that the criterion of the angle rotation size of the mobile robot is met, a right-turn angle node value RS, a left-turn angle node value Null, and a right-turn or left-turn angle node value Null or LSRS are set to be 1.
B. Under the condition that the criterion of the angle rotation size of the mobile robot is met, a right turning angle node value RB, a left turning angle node value LS and a right turning or left turning angle node value Null are set to be 1.
C. And under the condition of meeting the criterion of the angular rotation of the mobile robot, taking the RB as a right-turn angle node, taking the Null as a left-turn angle node, and setting the RB parameter value of the rotation angle node as 1.
D. Under the condition that the criterion of the angle rotation size of the mobile robot is met, a right-turn angle node value RS, a left-turn angle node value Null, and a right-turn or left-turn angle node value LBRB are set to be 1.
E. Under the condition that the criterion of the angle rotation size of the mobile robot is met, a right turning angle node value Null, a left turning angle node value LS, a right turning or left turning angle node value Null or LSRS is set to be 1.
F. Under the condition that the criterion of the angle rotation size of the mobile robot is met, a right turning angle node value RS, a left turning angle node value LB and a right or left turning angle node value Null are obtained, and an LS parameter value of the rotation angle node is set to be 1.
G. And under the condition of meeting the criterion of the angle rotation size of the mobile robot, taking a value Null of a right-turn angle node and a value LB of a left-turn angle node, and setting the LB parameter value of the rotation angle node to be 1.
H. Under the condition that the criterion of the angle rotation size of the mobile robot is met, a right turning angle node value Null, a left turning angle node value LS and a right turning or left turning angle node value LBRB set the LB parameter value of the turning angle node to be 1.
I. Under the condition that the criterion of the angle rotation size of the mobile robot is met, a right-turn angle node value RS, a left-turn angle node value LB and a right-turn or left-turn angle node value LBRB are set to be 1.
J. Under the condition that the criterion of the angle rotation size of the mobile robot is met, a right turning angle node value RS, a left turning angle node value LS and a right turning or left turning angle node value LSRS are set to be 1.
K. And under the condition of meeting the criterion of the angle rotation size of the mobile robot, setting the LSRS parameter value of the rotation angle node to be 1, wherein the right rotation angle node takes the value Null, the left rotation angle node takes the value Null, and the right rotation or left rotation angle node takes the value LSRS.
And L, under the condition of meeting the criterion of the angle rotation of the mobile robot, setting the LBRB parameter value of the rotation angle node to be 1, wherein the right rotation angle node takes a value RS, the left rotation angle node takes a value LS, and the right rotation or left rotation angle node takes a value LBRB.
And M, under the condition of meeting the criterion of the angle rotation of the mobile robot, setting the LBRB parameter value of the rotation angle node to be 1, wherein the right rotation angle node takes RB, the left rotation angle node takes LB, and the right rotation or left rotation angle node takes LBRB or LSRS.
And N, under the condition of meeting the criterion of the angle rotation of the mobile robot, taking a value Null of a right-turn angle node, taking a value Null of a left-turn angle node, taking a value LBRB of the right-turn or left-turn angle node, and setting the LBRB parameter value of the rotation angle node to be 1.
And O, under the condition of meeting the criterion of the angle rotation of the mobile robot, setting the Zero parameter value of the rotation angle node to be 1, wherein the right rotation angle node takes the value RS, the left rotation angle node takes the value LS, and the right rotation or left rotation angle node takes the value Null.
And P, under the condition of meeting the criterion of the angle rotation of the mobile robot, setting the Zero parameter value of the rotation angle node to be 1, wherein the right rotation angle node takes the value RB, the left rotation angle node takes the value LB, and the right rotation or left rotation angle node takes the value Null.
And Q, under the condition of meeting the criterion of the angle rotation size of the mobile robot, taking a Null value of a right-turn angle node, taking a Null value of a left-turn angle node, taking a Null value of a right-turn or left-turn angle node, and setting the Null parameter value of the rotation angle node to be 1.
Parameters of rotation angle node (FTurn): 30 ° (RS) to the right, 60 ° (RB) to the right, 30 ° (LS) to the left, 60 ° (LB) to the left, 30 ° (LSRS) to the right or to the left, 60 ° (LBRB) to the right or to the left, 0 ° (Zero), Null (Null), and set to 0 if the 17 criteria are not met.
(3) Respectively establishing fuzzy membership functions of the distance and the angle of the obstacle target around the mobile robot, and performing fuzzification processing on the distance and the angle data of the target:
according to the distance between the obstacle target and the mobile robot, the unit is centimeter, and the fuzzy set is set as follows: (near, Gmean), using a decreasing half trapezoidal membership function to give a representation of the fuzzy set:
Figure BDA0001490843080000151
Figure BDA0001490843080000152
according to the size of the positive included angle between the obstacle target and the X axis of the mobile robot coordinate system, the unit is degree, and the fuzzy set is set as follows: (small, Gsmall), using a reduced-half trapezoidal membership function to give a representation of the fuzzy set:
Figure BDA0001490843080000153
Figure BDA0001490843080000154
calculating fuzzy membership of the distance value on a fuzzy set (near, Gmean) of the obstacle target according to the established fuzzy membership function of the distance and the angle of the obstacle target; the fuzzy membership of the angle value on its fuzzy set (small, Gsmall).
(4) Inputting the blurred distance and angle data of the obstacle target into a Bayesian network model for obstacle avoidance behavior processing as an evidence, and carrying out Bayesian network reasoning:
(4a) taking the symptom node as an evidence node, and carrying out fuzzy processing on the distance and the angle of the obstacle target: and the membership degrees on the fuzzy set are used as evidence values of corresponding evidence nodes in different states and are respectively input into the Bayesian network model. The evidence values of the evidence nodes in different states are specifically as follows:
4a1., taking the fuzzy membership value (near, Gnear) of the distance of the obstacle object existing right ahead as the value of the evidence state near, Gnear corresponding to the distance node of the obstacle right ahead, and taking the value of the evidence state null as 0; the evidence of the distance node between the front obstacle and the node under the other conditions is as follows: null takes the value 1.
4a2, taking the fuzzy membership value (near, Gnear) of the distance of the obstacle target existing on the left side as the value of the evidence state near, Gnear corresponding to the distance node of the left obstacle, and taking the value of the evidence state null as 0; the evidence of the distance node of the left obstacle under the other conditions is as follows: null takes the value 1.
4a3., taking the fuzzy membership value (small, Gsmall) of the angle of the obstacle target existing on the left side as the value of the evidence state small, Gsmall corresponding to the angle node of the left obstacle, and taking the value of the evidence state null as 0; the evidence of the left obstacle angle node under other conditions is as follows: null takes the value 1.
4a4., taking the fuzzy membership value (near, Gnear) of the distance of the obstacle object existing on the right as the value of the evidence state near, Gnear corresponding to the distance node of the obstacle on the right, and taking the value of the evidence state null as 0; the evidence of the distance node of the right obstacle under the other conditions is as follows: null takes the value 1.
4a5., taking fuzzy membership value (small, Gsmall) of the angle of the obstacle target existing on the right side as the value of the evidence state small and Gsmall corresponding to the angle node of the obstacle on the right side, wherein the value of the evidence state null is 0; the evidence of the right obstacle angle node under other conditions is as follows: null takes the value 1.
(4b) Adopting a Bayesian network accurate reasoning algorithm to carry out reasoning on the Bayesian network model after the evidence value is input, and calculating the posterior probability value of the rotation angle node:
adopting a Bayesian network accurate inference algorithm, respectively inputting 5 sign nodes of evidence values of a left obstacle distance node (LDistance), a left obstacle angle node (Langle), a right obstacle distance node (RDistance), a right obstacle angle node (RAgle) and a front obstacle distance node (FDistance) on the basis of a determined Bayesian network model for obstacle avoidance processing of the mobile robot, wherein the evidence values respectively use e1,e2,e3,e4,e5Representing, then forming an evidence set: e ═ E1,e2,...,e5And accurately calculating the posterior probability value of the rotation angle node (Turn) by taking the input evidence set E as a posterior condition: p (Turn ═ RS/E), P (Turn ═ RB/E), P (Turn ═ LS/E), P (Turn ═ LB/E), P (Turn ═ LSRS/E), P (Turn ═ LBRB/E), P (Turn ═ Zero/E), and P (Turn ═ Null/E), one maximum a posteriori probability is selected, and the value state of the rotation angle node represented by the a posteriori probability is used as the angle at which the mobile robot needs to rotate to avoid the obstacle.
When obstacle avoidance processing is carried out through the Bayesian network model, distance information and angle information of an obstacle target are comprehensively considered, a coordinate system is determined more appropriately and objectively, a traveling direction rule is specified, the distance information and the angle information can be accurately applied to route planning of robot obstacle avoidance, and therefore the accuracy of the mobile robot obstacle avoidance is improved.
Example 8
The mobile robot obstacle avoidance method based on the bayesian network model is the same as that in embodiments 1 to 7, and the following takes the calculation process of P (Turn ═ RS/E) as an example, to describe the specific calculation steps of the posterior probability values of the values of each state of the rotation angle node in the present invention as follows:
Figure BDA0001490843080000171
wherein α represents a scale factor, P (e)1),P(e2),P(e3),P(e4),P(e5) Can be directly obtained from the parameters of the bayesian network,
Figure BDA0001490843080000172
can be given by the sum calculation of the parameters of the Bayesian network. To eliminate the scale factor α, the same calculation is done according to the procedure above: p (Turn ═ RB/E), P (Turn ═ LS/E), P (Turn ═ LB/E), P (Turn ═ LSRS/E), P (Turn ═ LBRB/E), P (Turn ═ Zero/E), P (Turn ═ Null/E), because the calculation is a posteriori probability for the same node, the scaling factors α are the same. Then
P(Turn/E)＝α＜P(Turn＝RB/E),P(Turn＝RS/E),P(Turn＝LB/E),P(Turn＝LS/E),
P (Turn ═ LBRB/E), P (Turn ═ LSRS/E), P (Turn ═ Zero/E), P (Turn ═ Null/E) >, and further normalization processing is performed:
Figure BDA0001490843080000173
i.e. to obtain an accurate value of P (Turn ═ RS/E).
(5) Selecting a result state with the maximum posterior probability value in the Bayesian network inference results, wherein the state value is taken as the rotation angle of the mobile robot to avoid the obstacle:
selecting a maximum value state from the posterior probability values of the rotation angle nodes (Turn) of the Bayesian network model, and taking the rotation angle represented by the state as the angle of the mobile robot which needs to rotate to avoid the obstacle. The specific operation mode is as follows:
selecting a value state with the maximum posterior probability value from the rotation angle nodes (Turn): if the maximum value state of the posterior probability value is RS, the mobile robot avoids the obstacle by the required rotation angle: off 30 ° to the right; if the maximum value state of the posterior probability value is RB, the mobile robot avoids the obstacle by the required rotating angle: 60 degrees off to the right; if the maximum value state of the posterior probability value is LS, the mobile robot avoids the obstacle by the required rotating angle: left offset by 30 °; if the maximum value state of the posterior probability value is LB, the mobile robot avoids the obstacle by the required rotating angle: left offset by 60 °; if the maximum value state of the posterior probability value is LSRS, the mobile robot avoids the obstacle by the required rotating angle: 30 ° left or right; if the maximum value state of the posterior probability value is LBRB, the mobile robot avoids the obstacle by the required rotating angle: 60 degrees left or right; if the maximum value state of the posterior probability value is Zero, the mobile robot avoids the obstacle by the required rotating angle: 0 degree deviation; and if the maximum value state of the posterior probability value is Null, the mobile robot does not need to avoid the obstacle.
In summary, the mobile robot obstacle avoidance method based on the bayesian network model provided by the invention aims to enable the mobile robot to avoid the obstacle in the process of traveling from the starting point to the target point in the unknown environment by establishing the bayesian network model, and solve the problem of insufficient consideration of the distance and angle of the obstacle in the obstacle avoidance of the mobile robot in the prior art. The method comprises the following implementation steps: (1) establishing a judgment rule of the obstacle avoidance condition of the mobile robot; (2) establishing a Bayesian network model for processing obstacle avoidance behaviors of the mobile robot; (3) establishing fuzzy membership functions of the distance and the angle of the obstacle target around the mobile robot, and performing fuzzification processing on the distance and the angle data of the target; (4) inputting the blurred distance and angle data of the obstacle target into a Bayesian network model for obstacle avoidance behavior processing as an evidence; (5) and selecting the deviation angle with the maximum posterior probability value in the Bayesian network inference result as the rotation angle of the mobile robot to avoid the obstacle. The Bayesian network model is applied to obstacle avoidance behavior processing of the mobile robot, and the fuzzy theory method is adopted to fuzzify obstacle target data detected by the sensor to obtain the angle of the mobile robot which needs to rotate, so that the mobile robot can avoid obstacles in the unknown environment, and the obstacle avoidance behavior is more accurate and is used for path selection of the mobile robot.

Claims (4)

1. A mobile robot obstacle avoidance method based on a Bayesian network model is characterized by comprising the following steps:
(1) establishing a judgment rule of the obstacle avoidance condition of the mobile robot:
the obstacle avoidance problem of the mobile robot is decomposed into 7 cases by taking the coordinate system of the mobile robot as a reference:
(1a) if an obstacle is arranged right in front of the mobile robot and no obstacle exists on the left side and the right side, the mobile robot deviates a certain angle to the right or left to avoid the obstacle;
(1b) if the mobile robot has an obstacle on the left and has no obstacle right in front of the mobile robot, the mobile robot only deviates a certain angle right to avoid the obstacle;
(1c) if the mobile robot has an obstacle on the right side and no obstacle on the left side and right front side, the mobile robot only deviates a certain angle on the left side to avoid the obstacle;
(1d) if the obstacle is positioned right ahead and on the left of the mobile robot and the obstacle is not positioned on the right, the mobile robot only deviates a certain angle right to avoid the obstacle;
(1e) if the obstacle is positioned right in front of the mobile robot and on the right side of the mobile robot, and the obstacle is not positioned on the left side of the mobile robot, the mobile robot only deviates a certain angle to the left to avoid the obstacle;
(1f) if the mobile robot has barriers on the left and right sides and no barrier is in the front of the mobile robot, the mobile robot deviates a certain angle to the right or left to avoid the barriers;
(1g) if the mobile robot has obstacles right in front, to the left and to the right, the mobile robot deviates a certain angle to the right or left to avoid the obstacles;
(2) establishing a Bayesian network model for obstacle avoidance of the mobile robot:
(2a) according to the judgment rule of 7 conditions of obstacle avoidance of the mobile robot, a diagnostic Bayesian network modeling mode is adopted to establish a Bayesian network structure of the obstacle avoidance of the mobile robot, and the method specifically comprises the following steps:
2a1. establishing Bayesian network structure symptom nodes: extracting the following description of a mobile robot obstacle avoidance condition judgment rule about an obstacle, wherein the obstacle is positioned right ahead, the obstacle is positioned on the left side, the obstacle is positioned on the right side, and according to the distance and angle attributes of the obstacle, the distance of the left obstacle, the angle of the left obstacle, the distance of the right obstacle, the angle of the right obstacle and the distance of the obstacle right ahead are used as sign nodes;
2a2, establishing a Bayesian network structure result node: extracting the following description of a mobile robot obstacle avoidance condition judgment rule about a mobile robot deviation angle, wherein the mobile robot deviates a certain angle leftwards only, the mobile robot deviates a certain angle rightwards or leftwards, and the deviation angle leftwards only, the deviation angle rightwards or leftwards is taken as a result node;
2a3., forming a Bayesian network model structure by using the symptom nodes as father nodes of result nodes in a diagnostic Bayesian network modeling mode;
2a4. sets all symptom nodes and result nodes in the Bayesian network model as discrete nodes: each node has three different states, and the values of the distance node of the left obstacle, the distance node of the right obstacle and the distance node of the right obstacle are as follows: near, Gnear, empty Null; values of the angle node of the left obstacle and the angle node of the right obstacle are as follows: small deviation angle small, small deviation angle Gsmall, empty Null; the right deviation angle node only takes values as: only 30 ° RS, 60 ° RB, Null, right; the left deviation angle node only takes values as: left offset only 30 ° LS, left offset only 60 ° LB, Null; the right or left deviation angle node takes the values as: 30 ° LSRS right or left, 60 ° LBRB right or left, Null;
(2b) setting each node parameter in the established Bayesian network structure for obstacle avoidance of the mobile robot, wherein the node parameters are set according to the following rules: setting parameter values of the symptom nodes in three states as follows: 1/3 (0.33); setting the parameter value of the result node to be 1 when the condition of the judgment criterion is met, and setting the parameter value of the result node to be 0 when the condition of the judgment criterion is not met;
(3) respectively establishing fuzzy membership functions of the distance and the angle of the obstacle target around the mobile robot, and performing fuzzification processing on the distance and the angle data of the target:
extracting obstacle target information obtained by a sensor, setting a fuzzy set of the obstacle target information as follows according to the distance between the obstacle target and the mobile robot, wherein the unit is centimeter: (near, Gmean) using a decreasing half trapezoidal membership function to give a representation of the fuzzy set; according to the size of the positive included angle between the obstacle target and the X axis of the mobile robot coordinate system, the unit is degree, and the fuzzy set is set as follows: (small, Gsmall), and representing the fuzzy set by adopting a halving trapezoidal membership function;
according to the established fuzzy membership functions of the distance and the angle of the obstacle target, calculating the fuzzy membership of the distance value on a fuzzy set (near, Gmean) of the distance value and the fuzzy membership of the angle value on a fuzzy set (small, Gmean) of the angle value;
(4) inputting the blurred distance and angle data of the obstacle target into a Bayesian network model for obstacle avoidance behavior processing as an evidence, and carrying out Bayesian network reasoning:
(4a) inputting the fuzzy processing result of the distance and the angle of the obstacle target into a Bayesian network model: taking the symptom node as an evidence node, and carrying out fuzzy processing on the distance and the angle of the obstacle target: and the membership degrees on the fuzzy set are used as evidence values of corresponding evidence nodes in different states and are respectively input into the Bayesian network model, and the evidence values of the corresponding evidence nodes in different states refer to:
4a1., taking the fuzzy membership value (near, Gnear) of the distance of the obstacle object existing right ahead as the value of the evidence state near, Gnear corresponding to the obstacle distance node right ahead, and the value of the evidence state null is 0, otherwise, the evidence of the obstacle distance node right ahead is: null takes a value of 1;
4a2, taking the fuzzy membership value (near, Gnear) of the distance of the obstacle target existing on the left as the value of the evidence state near, Gnear corresponding to the distance node of the left obstacle, and the value of the evidence state null is 0, otherwise, the evidence of the distance node of the left obstacle is: null takes a value of 1;
4a3., using the fuzzy membership value (small, Gsmall) of the angle of the obstacle object existing on the left side as the value of the evidence state small, Gsmall corresponding to the left obstacle angle node, where the value of the evidence state null is 0, and the evidence of the left obstacle angle node under the other conditions is: null takes a value of 1;
4a4., taking the fuzzy membership value (near, Gnear) of the distance of the obstacle object existing on the right as the value of the evidence state near, Gnear corresponding to the distance node of the obstacle on the right, and the value of the evidence state null is 0, otherwise, the evidence of the distance node of the obstacle on the right is: null takes a value of 1;
4a5., taking the fuzzy membership value (small, Gsmall) of the angle of the obstacle object existing on the right side as the value of the evidence state small, Gsmall corresponding to the right obstacle angle node, wherein the value of the evidence state null is 0, and the evidence of the right obstacle angle node under the other conditions is as follows: null takes a value of 1;
(4b) performing Bayesian network inference: adopting a Bayesian network accurate reasoning algorithm to carry out reasoning on the Bayesian network model after the evidence values are input, and calculating posterior probability values of each result node to obtain a Bayesian network reasoning result;
(5) and selecting a result node with the maximum posterior probability value in the Bayesian network inference results, wherein the state value of the result node is used as the rotation angle of the mobile robot to avoid the obstacle.
2. The barrier avoiding method for the mobile robot based on the bayesian network model as claimed in claim 1, wherein the mobile robot coordinate system in step (1) is established by taking a center of the mobile robot as a coordinate origin, and an XYZ three-dimensional coordinate system is established, wherein an X-axis positive direction represents a positive front direction of the mobile robot, an X-axis negative direction represents a positive rear direction of the mobile robot, a Y-axis positive direction represents a positive right direction of the mobile robot, a Y-axis negative direction represents a positive left direction of the mobile robot, a Z-axis positive direction represents a positive lower direction of the mobile robot, and a Z-axis negative direction represents a positive upper direction of the mobile robot.
3. The barrier avoiding method for the mobile robot based on the bayesian network model as recited in claim 1, wherein the position right in front of the mobile robot in step (1) is determined by taking a Y-axis forward direction of a coordinate system of the mobile robot as a reference, and an included angle between the Y-axis forward direction and the coordinate system of the mobile robot is within a range: (60 °, 120 °);
the left side of the mobile robot in the step (1) refers to that the forward direction of the Y axis of the coordinate system of the mobile robot is taken as a reference, and an included angle between the forward direction of the Y axis and the included angle is in a range: [120 °,160 ° ];
the right-side deviation of the mobile robot in the step (1) refers to that the included angle between the forward direction of the Y axis and the forward direction of the Y axis of a coordinate system of the mobile robot is in a range by taking the forward direction of the Y axis as a reference: [20 °,60 ° ].
4. The Bayesian network model-based mobile robot obstacle avoidance method according to claim 1, wherein the Bayesian network accurate inference algorithm in the step (4b) is used for accurately calculating posterior probability values of result nodes through probabilistic inference according to set parameter values of symptom nodes, parameter values of result nodes and evidence values.
CN201711246586.XA 2017-12-01 2017-12-01 Mobile robot obstacle avoidance method based on Bayesian network model Active CN108247630B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201711246586.XA CN108247630B (en) 2017-12-01 2017-12-01 Mobile robot obstacle avoidance method based on Bayesian network model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201711246586.XA CN108247630B (en) 2017-12-01 2017-12-01 Mobile robot obstacle avoidance method based on Bayesian network model

Publications (2)

Publication Number Publication Date
CN108247630A CN108247630A (en) 2018-07-06
CN108247630B true CN108247630B (en) 2021-01-05

Family

ID=62720846

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201711246586.XA Active CN108247630B (en) 2017-12-01 2017-12-01 Mobile robot obstacle avoidance method based on Bayesian network model

Country Status (1)

Country Link
CN (1) CN108247630B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111568432B (en) * 2020-04-06 2023-04-07 沈阳工业大学 Fall detection method of omnibearing walking training robot based on fuzzy inference

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN100568144C (en) * 2008-09-04 2009-12-09 湖南大学 Mobile robot's multirow is for merging automatic navigation method under a kind of circumstances not known
CN102103663B (en) * 2011-02-26 2012-07-25 山东大学 Ward visit service robot system and target searching method thereof
US9984332B2 (en) * 2013-11-05 2018-05-29 Npc Robotics Corporation Bayesian-centric autonomous robotic learning
CN104157160B (en) * 2014-08-08 2016-08-17 中国联合网络通信集团有限公司 Vehicle travel control method, device and vehicle
CN105116902A (en) * 2015-09-09 2015-12-02 北京进化者机器人科技有限公司 Mobile robot obstacle avoidance navigation method and system
CN105631414B (en) * 2015-12-23 2019-04-05 上海理工大学 A kind of vehicle-mounted multi-obstacle avoidance sorter and method based on Bayes classifier
CN106584472A (en) * 2016-11-30 2017-04-26 北京贝虎机器人技术有限公司 Method and device for controlling autonomous mobile equipment

Also Published As

Publication number Publication date
CN108247630A (en) 2018-07-06

Similar Documents

Publication Publication Date Title
Li et al. Multivehicle cooperative local mapping: A methodology based on occupancy grid map merging
Ayawli et al. Mobile robot path planning in dynamic environment using Voronoi diagram and computation geometry technique
Miki et al. Elevation mapping for locomotion and navigation using gpu
Weng et al. Pole-based real-time localization for autonomous driving in congested urban scenarios
Kolhatkar et al. Review of SLAM algorithms for indoor mobile robot with LIDAR and RGB-D camera technology
CN110908377B (en) Robot navigation space reduction method
US20180174038A1 (en) Simultaneous localization and mapping with reinforcement learning
Wyffels et al. Negative information for occlusion reasoning in dynamic extended multiobject tracking
CN111830985A (en) Multi-robot positioning method, system and centralized communication system
CN110764518A (en) Underwater dredging robot path planning method and device, robot and storage medium
CN115143970A (en) Obstacle avoidance method and system of underwater vehicle based on threat degree evaluation
CN108247630B (en) Mobile robot obstacle avoidance method based on Bayesian network model
Dashora et al. Hybrid imitative planning with geometric and predictive costs in off-road environments
Heidenreich et al. LaneSLAM--Simultaneous Pose and Lane Estimation Using Maps with Lane-Level Accuracy
Short et al. Abio-inspiredalgorithminimage-based pathplanning and localization using visual features and maps
Lu et al. Lane marking-based vehicle localization using low-cost GPS and open source map
Kaur et al. Sensor fusion algorithm for software based advanced driver-assistance intelligent systems
CN116872915A (en) Genetic algorithm-based automatic parking multi-target self-adaptive track optimization method
Almeida et al. Real-time tracking of moving objects using particle filters
Dhiman et al. A review of path planning and mapping technologies for autonomous mobile robot systems
Almeida et al. Real-time tracking of multiple moving objects using particle filters and probabilistic data association
Parr et al. Agent-aware state estimation: Effective traffic light classification for autonomous vehicles
Cruz-Bernal Meta-heuristic optimization techniques and its applications in robotics
Hu et al. A modified particle filter for simultaneous robot localization and landmark tracking in an indoor environment
Xu et al. TrafficEKF: A learning based traffic aware extended Kalman filter

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