CN108802741B - Mobile robot sonar data fusion method based on DSmT theory - Google Patents
Mobile robot sonar data fusion method based on DSmT theory Download PDFInfo
- Publication number
- CN108802741B CN108802741B CN201810649501.0A CN201810649501A CN108802741B CN 108802741 B CN108802741 B CN 108802741B CN 201810649501 A CN201810649501 A CN 201810649501A CN 108802741 B CN108802741 B CN 108802741B
- Authority
- CN
- China
- Prior art keywords
- sonar
- mobile robot
- dsmt
- fusion
- measurement
- 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
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S15/00—Systems using the reflection or reradiation of acoustic waves, e.g. sonar systems
- G01S15/88—Sonar systems specially adapted for specific applications
- G01S15/93—Sonar systems specially adapted for specific applications for anti-collision purposes
- G01S15/931—Sonar systems specially adapted for specific applications for anti-collision purposes of land vehicles
Landscapes
- Engineering & Computer Science (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Computer Networks & Wireless Communication (AREA)
- General Physics & Mathematics (AREA)
- Measurement Of Velocity Or Position Using Acoustic Or Ultrasonic Waves (AREA)
- Control Of Position, Course, Altitude, Or Attitude Of Moving Bodies (AREA)
Abstract
The invention provides a mobile robot sonar data fusion method based on a DSmT theory, which solves the fusion problem of acquired sonar data under the condition of conflict and comprises the following implementation steps: acquiring information through a sonar sensor, and establishing a sonar sensor measurement model; dividing a sonar detection sector area into: open and occupiable areas; rasterizing a two-dimensional plane environment of the mobile robot, providing a grid discrimination framework, and establishing a reliability assignment calculation model under the discrimination framework; and constructing DSmT mixed combination rules of sonar data at different moments under the constraint condition, and completing fusion of the sonar data of the mobile robot according to the DSmT mixed combination rules of the sonar data at different moments. The DSmT theory is applied to sonar data fusion of the mobile robot, the problem of fusion of conflict data is solved, the grid state around the robot is judged more accurately, and the method can be used for map creation of the mobile robot in an unknown environment in practical application.
Description
Technical Field
The invention belongs to the technical field of computers, relates to data fusion, and particularly relates to a mobile robot sonar data fusion method based on a DSmT theory, which can be used for map creation of a mobile robot in an unknown environment in practical application.
Background
The detection of unknown complex environments by using intelligent mobile robots has been a hot and difficult subject of expert research of robots at home and abroad. The map creation is an expression form of the mobile robot for sensing the environment, in an unknown environment, the mobile robot acquires information of the surrounding environment through sensors, such as sonar, laser, infrared, vision and the like, which are loaded on a body, recombines and fuses the information, outlines or images of the surrounding environment are outlined, and the mobile robot is positioned. The sonar sensor is often used as an important sensor of a mobile robot due to its advantages of low price, simple usage mode, convenient data processing and the like. Due to the limitation of the sensor, data provided by the sensor usually contains a large amount of uncertain information, the information is often incomplete, inaccurate and fuzzy, sometimes even contradictory and wrong, and it is difficult to obtain an accurate environment model by directly using the sensing information to perform map creation, so that the sensing information usually needs to be reprocessed, and more accurate environment information is obtained by fusing multiple sensing information.
Zhang in a published paper, "three-dimensional environment modeling technical research of mobile robot based on information fusion" (Beijing post and telecommunication university, doctor academic paper, 2013.1), adopts a D-S evidence theory method to fuse sonar sensor data of the mobile robot, and establishes a two-dimensional plane grid map in an unknown environment. The method divides the trellis states into: the method comprises the following steps of (1) fusing sonar data through a synthetic rule of a D-S evidence theory by using an inaccessible area, a determined area (determined to have an obstacle or not) and an unknown area. The method has the following disadvantages: the synthesis rule of the D-S evidence theory is adopted to distribute the credibility of the conflict focal elements under the fusion framework to all propositional elements on average, but according to the D-S synthesis rule, when two data completely conflict, the two data cannot be synthesized by the rule, and when the two evidences conflict highly, the result which is contrary to the actual paradox can be caused by the synthesis by the rule.
A sonar data fusion method adopting a neural network and a Bayesian theory is disclosed in a patent of Hunan university (patent application No. CN200810143537.8, publication No. CN101413806) of 'a mobile robot grid map creation method of real-time data fusion'. The method extracts the measured values of three sonar sensors closest to the current calculation grid unit at the same time as the input of a neural network, wherein the input of the neural network is the grid state: idle, occupied and uncertain states, and finally updating the state of the grid by adopting a Bayesian rule. The method disclosed in this patent application has the following disadvantages: the determination of the neural network parameters needs the support of a large amount of sample data, the precision of the parameters has a large influence on the output result, and the reliability of the fusion result is also influenced.
In the process of creating a map by a mobile robot, in the fusion processing of sonar data in the prior art, the problem of fusion of conflict data cannot be solved, the calculation amount is large and the highest efficiency is not achieved, the precision of parameters also influences the output result and influences the reliability of the map creation result.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a more accurate and reliable mobile robot sonar data fusion method based on the DSmT theory.
The invention relates to a mobile robot sonar data fusion method based on a DSmT theory, which is characterized by comprising the following steps:
(1) obtaining information through a sonar sensor, and establishing a measurement model of the sonar sensor:
the unknown environment that mobile robot place, for two-dimensional plane environment, the sonar sensor of a plurality of different directions through mobile robot body loading obtains the information of surrounding environment in two-dimensional plane environment to establish sonar sensor's measurement model: the method comprises the following steps of forming a sonar detection sector area range by using a sound wave emitting angle and a maximum measurement distance of sonar, and setting a measurement value returned by the sonar as a distance measurement value of a target closest to the sonar distance in the sector area;
(2) according to a measurement model of the sonar sensor, dividing a sonar detection sector area range into: open area and occupiable area:
(2a) an open area: [0, R-epsilon), [0, R-epsilon ], in the area, the probability of the existence of the obstacle is 0, wherein R refers to the measurement distance returned by the sonar, and epsilon refers to the measurement error of the sonar;
(2b) the occupiable area: [ R- ε, R + ε ], the probability of no obstacle in this region is 0;
(3) rasterizing a two-dimensional plane environment where the mobile robot is located, giving a discrimination framework of a grid state, and establishing a calculation model of reliability assignment under the discrimination framework:
(3a) rasterizing a two-dimensional plane environment where the mobile robot is located, wherein each grid represents the space size of 80 × 80cm, and judging the states of the grids to be three types: empty, with obstacles, unknown;
(3b) according to the DSmT theory, a decision frame Ω of the grid state is given: q ═ { E, O }, where E represents null and O represents an obstacle;
(3c) according to the division of a measurement model and a measurement area of a sonar sensor, establishing a confidence assignment calculation model of grids around a mobile robot under a judgment frame omega { E, O }, according to the area where the grids are located, dividing the confidence assignment calculation model into a confidence assignment calculation model of an open area of sonar measurement, a confidence assignment calculation model of an occupiable area of sonar measurement and a calculation model of the occupiable area exceeding the sonar measurement, wherein a DSmT theory is established on the basis of an overpowering set, the confidence assignment calculation model specifically comprises the assignment of each element in the overpowering set formed by the judgment frame, the value range of the confidence assignment of each element is [0,1], the value of the confidence assignment of the element in the overpowering set is required to be 0, and the sum of the confidence assignments of all the elements in the overpowering set is 1;
(4) and establishing DSmT mixed combination rules of sonar data at different moments under the constraint condition, and completing the fusion of the sonar data according to the DSmT mixed combination rules of the sonar data at different moments.
The invention fuses the measured data of the sonar sensors at different moments through the DSmT theory, and judges the state of the grid around the mobile robot: the method has no obstacle, has obstacles and is unknown, and the problem of fusion of sonar data under the conflict condition is successfully solved.
Compared with the prior art, the invention has the following advantages:
firstly, the invention fuses the measured data of the sonar sensors at different moments through the DSmT theory, judges the state of the grid around the mobile robot, sets three conditions of no obstacle, obstacle and unknown, and distributes the combination reliability of the conflict focus elements to the fusion reliability of the unknown state by keeping the evidence conflict items as the focus elements of data fusion, thereby well solving the information fusion problem under the condition of evidence high conflict to achieve the effective identification of the grid state.
Secondly, the method performs fusion processing on the sonar data through the DSmT theory, reduces the calculated amount in the fusion process, and does not influence the reliability of the fusion result.
Drawings
FIG. 1 is a flow chart of an implementation of the present invention;
FIG. 2 is a sonar transducer measurement model used in the present invention;
FIG. 3 is a measurement area division of a sonar used in the present invention;
fig. 4 is a sonar sensor-based grid state reliability assignment calculation model used in the present invention;
FIG. 5 is a combination of multiple evidences for use with the present invention;
fig. 6 shows the angle setting of the sonar sensor in front of the mobile robot in the simulation experiment of the present invention.
Detailed Description
The invention is described in detail below with reference to the figures and the specific embodiments.
Example 1
In the sonar-based map creation, due to the fact that uncertainty of sonar information is strong, a plurality of sonar sensors are required to acquire information at different time intervals, and the sonar information is fused. At the present stage, the application of information fusion technology is mainly in the military field. Modern wars are widely applied to various high technologies, a multi-sensor and multi-source information system is required to be used for obtaining more battlefield information, the data processing amount and the processing capacity of the system far exceed those of a single sensor, and the multi-sensor information fusion technology is widely applied along with the continuous increase of the types and the number of sensors such as radar, infrared, photoelectric and the like. In addition, information fusion techniques have gradually penetrated into multiple domains. In the field of mobile robots, the information fusion technology has been increasingly applied, but there are still many problems, such as the inability to process collision data, for example, when there is a high degree of conflict, even if the data obtained by two different sensors are exactly opposite, this is considered conflicting data, in the process of creating the mobile robot map, due to the factors of the sonar sensors, conflict data exist in the information acquired by the sonar sensors, therefore, the accuracy of the obtained map data needs to be improved, so that research and innovation are developed, the invention provides a mobile robot sonar data fusion method based on the DSmT theory, and also can be said to be a sonar data fusion method based on the DSmT theory in mobile robot map creation, the problem of merging conflict data is successfully solved, and referring to fig. 1, the invention comprises the following steps:
(1) information is acquired through the sonar sensor, and a measurement model of the sonar sensor is established according to the working principle of the sonar sensor:
the unknown environment where the mobile robot is located is a two-dimensional plane environment, the sonar sensors in a plurality of different directions loaded by the mobile robot body in the two-dimensional plane environment acquire information of the surrounding environment, and according to the basic working principle of the sonar sensors, a measurement model of the sonar sensors is established, and the two-dimensional plane environment is shown in fig. 2: the sound wave emitting angle and the maximum measuring distance of the sonar form a sonar detection sector area range, and the measured value returned by the sonar is the measured value of the distance from the sonar to the nearest target in the sector area.
(2) According to a measurement model of the sonar sensor, dividing a sonar detection sector area range into: an open area and an occupiable area. Referring to fig. 2, psi is the sensor beam width, i.e. the emission angle of the acoustic wave, epsilon is the sensor measurement error, P is the obstacle target, rho is the true distance from the target P to the sonar, and alpha is the angle
The included angle between the central axis (x axis) of the sonar, namely the incident angle, R is the measured value of the sonar sensor on the P-point obstacle, and the measured distance R for the sonar return is shown in figure 3, epsilon is the measurement error of the sonar, and psi is the emission angle of the sonar. Because errors exist between the measured data and the actual data acquired by the sonar sensor, the errors are taken into account when the model is designed, and the actual situation is more comprehensively taken into account.
(2a) Defining an open area: 0, R-epsilon) in which the probability of the existence of an obstacle is 0, see fig. 3.
(2b) Defining an occupiable area: [ R- ε, R + ε ], see FIG. 3, the probability of no obstacle being present in this region is 0.
(3) Rasterizing a two-dimensional plane environment where the mobile robot is located, giving a discrimination framework of a grid state, and establishing a calculation model of reliability assignment under the discrimination framework. Referring to fig. 4, the central axis of the sonar is taken as an x axis, the position of the sonar is taken as an origin o, and psi is taken as a sonar emission angle; the OP1 is the measured distance value returned by the sonar, OP2 is OP1+ epsilon, OP3 is OP 1-epsilon, and epsilon is the sensor measurement error. After the two-dimensional plane environment is rasterized, the searching efficiency of the state space is improved, and the data precision is kept.
(3a) The method comprises the following steps of rasterizing a two-dimensional plane environment where the mobile robot is located, wherein each grid represents the space size of 80 × 80cm, and judging the states of the grids to be three according to data measured by a robot sonar sensor: empty, with obstacle, unknown, wherein the empty state is the state without obstacle.
(3b) According to the DSmT theory, a decision frame Ω of the grid state is given: in DSmT theory, a complete set of mutually incompatible basic propositions is called the discriminative framework, representing all possible answers to a question, but only one of them is correct.
(3c) According to the measurement model and the measurement area of the sonar sensor, a confidence assignment calculation model of grids around the mobile robot under a judgment frame omega { E, O } is established, according to the area where the grids are located, the confidence assignment calculation model is divided into a confidence assignment calculation model of an open area of sonar measurement, a confidence assignment calculation model of an occupied area of sonar measurement and a calculation model of the occupied area exceeding the sonar measurement, according to the DSmT theory, the confidence assignment calculation model specifically comprises the assignment of each element in a super power set formed by the judgment frame, the value range of the confidence assignment of each element is [0,1], the value of the confidence assignment of the element in the super power set is required to be 0, and the sum of the confidence assignments of all the elements in the super power set is required to be 1.
That is to say, the confidence value assignment calculation model is determined according to three conditions that grids around the mobile robot fall in an open area measured by sonar, fall in an occupied area measured by sonar and exceed the occupied area measured by sonar, the confidence value assignment is a value of each element of a super-power set formed by a discrimination frame in a [0,1] interval, the value of the empty set is required to be 0, and the sum of the values of all the elements is 1.
(4) The invention establishes the DSmT mixed combination rule of sonar data at different times under the constraint condition, and directly classifies the situation that a certain grid part is occupied by an obstacle and the part is empty, namely E n O, as the situation with the obstacle, so the E n O is set as the constraint condition and the fusion of the sonar data is completed according to the DSmT mixed combination rule of the sonar data at different times.
According to the invention, the measured data of the sonar sensors at different moments are fused through the DSmT theory, the DSmT theory solves the information fusion problem under the condition that evidence is in high conflict by reserving an evidence conflict item as a data fusion focal element, and the calculation amount in the fusion process is reduced by using the mixed fusion rule of the DSmT theory under a certain constraint condition without influencing the reliability of the fusion result.
Example 2
The mobile robot sonar data fusion method based on the DSmT theory and the DSmT mixed combination rule for establishing sonar data at different times under the constraint condition, which is described in the embodiment 1 and the step 4, means that the combination credibility of the grid states obtained by sonar sensors in different directions loaded by a mobile robot body under the conditions of being empty, having obstacles and being unknown forms a new combination credibility together, and the new combination credibility are combined with the combination credibility under the condition of being not empty of the feature function
And multiplying to obtain the reliability assignment of the grid.
The classical combination rule of the DSmT theory does not have any constraint condition, but mostly has the constraint condition in the actual evidence information fusion process, the DSmT theory mixed combination rule solves the problem, has a constraint condition set and is more suitable for the actual fusion requirement. The map information is a set of all grid state information, the reliability of the grid state around the mobile robot is assigned, the construction of a map state space and the formulation of a search strategy are simplified, the problem of search combination explosion existing in the traditional search algorithm can be avoided, complex operation is not required, and the satisfactory effect can be achieved from the criteria of completeness, time complexity, space complexity and optimality.
Example 3
The mobile robot sonar data fusion method based on the DSmT theory is the same as that in embodiment 1-2, and the specific confidence fusion result and the calculation formulas of the combination rules are respectively as follows, wherein the DSmT mixed combination rules of sonar data at different times under the constraint condition are established in step 4:
wherein m is1(. and m)2(. for basic trust assignments from two different data sources, mμ(Ω) (C) represents the confidence fusion result with constraint condition, μ is constraint condition E &, O ═ Φ, where Φ is the absolute null set, C ∈ DΩ，
Is a non-null feature function if
Then the
The value is '0', otherwise '1',
for the set of constraints, Φ represents the absolute null set,
indicating a relatively empty set condition. In the above formula, S1(C) For the combination rule of DSmT without any constraints, S2(C) Assigning the combined credibility of all absolute empty sets and relative empty sets to a combined rule under the condition of distributing a total unknown set and a relative unknown set, wherein the total unknown set is a union of all propositions in a judgment frame, S3(C) The combined credibility of the conflict focal elements is assigned to the combined rule under the condition of the union of the conflict focal elements. Three of the three jointly form a DSmT mixed combination rule to obtain a credibility fusion result mμ(Ω)(C)。
The DSmT mixed combination rule of the invention assigns basic credibility from different data sources, and performs multiple fusion under the constraint condition, thereby obtaining a real and effective result which better accords with the actual situation.
Example 4
The mobile robot sonar data fusion method based on the DSmT theory, as in embodiments 1-3, completes the fusion of sonar data according to the DSmT mixed combination rule of sonar data at different times described in step 4, and the DSmT theory is established on the basis of an overpowering set, and includes the following steps:
(4a) overpowering set D under grid state discrimination frame omegaΩComprises the following steps: dΩSetting E.n.O.O as constraint condition, where phi is absolute empty set, DΩAdding a pass-and operation to propositional elements in a discriminant framework omega: and U and traffic: andd. the set of composite propositions.
(4b) And under the constraint condition E n O phi, fusing the fusion result of the grid state reliability at the previous t-1 moment and the grid reliability assignment calculated according to the sonar measurement data at the current t moment by adopting a DSmT mixed combination rule, and completing the fusion of the sonar data of the surrounding environment information, which is obtained by the mobile robot through the sonar sensor, in the two-dimensional plane of the unknown environment.
And calculating the reliability assignment of the grid state around the robot in each simulation step according to the measured data of the robot sonar sensor obtained in each simulation step, and fusing the reliability assignments of different simulation steps to effectively identify the grid state.
Example 5
The mobile robot sonar data fusion method based on the DSmT theory is as in embodiments 1-4, (4b) that the fusion result of the grid state confidence at the previous t-1 moment and the grid confidence value calculated according to the sonar measurement data at the current t moment are fused by using the DSmT mixed combination rule, and specifically includes:
wherein m isij(omega) (C) represents the confidence fusion result of the ith row and jth column grids, and C belongs to DΩ，
For a non-empty feature function, C is the state of the grid, and the fusion constraint is: e n O is phi; c values are respectively a set of over powers DΩΦ, O, E, and EU O of 4 states of the grid; s1(C) As a combination rule of the free DSmT model, the free DSmT model refers to the direct transformation from the hyper-power set space DΩThe formed fusion model does not have any constraint condition; s2(C) And assigning the combined credibility assignment of all absolute empty sets and relative empty sets to a total unknown set and a relative unknown set, wherein the total unknown set refers to a union set of all propositions in a discrimination frame, and comprises the following steps: and (3) judging a frame omega, { E, O }, wherein the total unknown set is as follows: e, U.O, the relative unknown set is: judging a union set of partial propositions in the frame; s3(C) And combining the conflicting focal elements with confidence, such as: m (E.andgate O), assigned to the union of conflicting coke elements: m (E.U.O); blending the calculation Process and results mij(omega) (C) is as follows：
(1) C is Φ, then mij(Φ)＝0。
(2) C ═ O, i.e., there is an obstacle state:
non-null feature function
Then: m isij(O)＝S1(O)+S2(O)+S3(O), wherein:
S2(O)＝0，S3(O)＝0。
(3) c ═ E, i.e., null state:
non-null feature function
Then: m isij(E)＝S1(E)+S2(E)+S3(E) Wherein:
S2(E)＝0，S3(E)＝0。
(4) c ═ E ═ coo, i.e. the unknown state:
non-null feature function
Then: m isij(E∪O)＝S1(E∪O)+S2(E∪O)+S3(E.U.O), wherein:
S2(E)＝0，
it can be seen that, in (4), the combined confidence of the conflicting focal elements, namely:
In the calculation process of the 4 state-confidence fusion of the grid,
representing the fused result of the grid state confidence at time t-1,
and representing the grid state reliability assignment calculated according to the sonar measurement data at the time t.
The detailed classification calculation description is carried out on the process of calculating the reliability assignment of the grid state around the robot in each simulation step through the DSmT mixed combination rule, and all practical situations are covered.
Example 6
The mobile robot sonar data fusion method based on the DSmT theory is the same as that in embodiments 1 to 5, and the invention is a mobile robot sonar data fusion method based on the DSmT theory, and the method is shown in fig. 1 and comprises the following steps:
(1) according to the working principle of the sonar sensor, a measurement model of the sonar sensor is established, and referring to fig. 2, the sonar sensor is the most commonly used sensor in the mobile robot. The working principle of the device is that the transmitter transmits an ultrasonic detection signal, the receiver receives a signal reflected by the barrier, and the distance between the sensor and the barrier is calculated according to the time difference between transmission and reception. Sonar has been widely used in mobile robots due to its low cost and convenient use. The sonar transducer detects an object by emitting a cone-shaped sound wave and receives the reflected wave to calculate the distance of the object. Because the sound wave has a certain emission angle, the detection range of the sonar is a fan-shaped area, and the detected object is the object which is closest to the sonar in the fan-shaped area. In fig. 2, the coordinate system uses the central axis of the sonar as the x-axis, and the position of the sonar is the origin, and the measurement model includes:
● psi is the sensor beam width, which is the emission angle of the sound wave;
● epsilon is the sensor measurement error;
● P point is an obstacle target, and rho is the real distance from the target P point to the sonar;
● Angle α is
The included angle between the central axis (x axis) of the sonar and the central axis (x axis) of the sonar is the incident angle;
● R is the measured value of the point P obstacle by the sonar transducer.
(2) According to the measurement model of the sonar sensor, the measurement area range of the sonar is further divided, as shown in fig. 3, R is the measurement distance returned by the sonar, is the measurement error of the sonar, and ψ is the emission angle of the sonar. The measuring range of the sonar is as follows: psi emission angle and is within the maximum measurement range of the sonar. And aiming at the measurement distance R returned by the sonar, the measurement range of the sonar is further divided into an open area and an occupied area:
(2a) an open area: [0, R-. epsilon.), and the probability of the existence of an obstacle in this region is 0.
(2b) The occupiable area: [ R- ε, R + ε ], and the probability that no obstacle exists in this region is 0.
(3) Rasterizing a two-dimensional plane environment where the mobile robot is located, giving a discrimination framework of a grid state, and establishing a calculation model of reliability assignment under the discrimination framework:
(3a) the rasterization of the two-dimensional plane environment where the mobile robot is located and the rasterization of the two-dimensional plane environment where the mobile robot is located refer to the division of the environment to be described into a plurality of grids with a certain size, and the possibility that an obstacle exists in the environment is represented by the grid state. The size of the grid is generally the same, but a variable-scale grid may be used depending on the actual application, and the size of the grid determines the accuracy of the map. The grid map is not only strong in intuition, but also easy to create and maintain, and can better describe the environment compared with other maps. And thus it is more applied in mobile robot map creation. Each grid represents the space size of 80 × 80cm, as shown in fig. 4, the central axis of the sonar is taken as the x axis, the position where the sonar is located is taken as the origin o, and ψ is taken as the emission angle of the sonar; the OP1 is the measured distance value of the sonar return, OP2 is OP1+ epsilon, OP3 is OP 1-epsilon, and epsilon is the sensor measurement error. Judging the state of the grid into three types: empty (no obstacle), with obstacle, unknown.
(3b) According to the DSmT theory, a decision framework for the grid state is given: and E, O, where E represents null and O represents an obstacle.
(3c) According to the division of a measurement model and a measurement area of a sonar sensor, establishing a confidence assignment calculation model of grids around a mobile robot under a judgment frame omega { E, O }, according to the area where the grids are located, dividing the confidence assignment calculation model into a confidence assignment calculation model of an open area of sonar measurement, a confidence assignment calculation model of an occupiable area of sonar measurement and a calculation model of the occupiable area exceeding the sonar measurement, wherein a DSmT theory is established on the basis of an overpowering set, the confidence assignment calculation model specifically comprises the assignment of each element in the overpowering set formed by the judgment frame, the value range of the confidence assignment of each element is [0,1], the value of the confidence assignment of the element in the overpowering set is required to be 0, and the sum of the confidence assignments of all the elements in the overpowering set is 1; calculating the reliability assignment of grids around the mobile robot under a judgment frame omega { E, O }, wherein the central axis of the sonar is used as an x axis, the position of the sonar is used as an origin O, and psi is used as the emission angle of the sonar; the OP1 is a measurement distance value returned by the sonar, OP2 is OP1+ epsilon, OP3 is OP 1-epsilon, and epsilon is a sensor measurement error, and specifically comprises the following steps:
(3c1) for grids in the sonar emission angle psi range, the state reliability assignment is specifically calculated by adopting d to represent the distance (unit is cm) from the sonar from the center of a certain grid:
when d is less than OP 1-epsilon, namely the grid falls in the open area of sonar measurement, the confidence value assignment calculation model of the open area of sonar measurement is as follows:
when OP 1-epsilon < d < OP1+ epsilon, namely the grid falls into the occupied area of the sonar measurement, the confidence value assignment calculation model of the occupied area of the sonar measurement is as follows:
③ when d > OP1+ ε, i.e., the grid has exceeded the footprint of the sonar survey, no confidence scores need to be calculated outside our scope of discussion.
(3c2) Since the grid represents a space size of 80 x 80cm, d is the distance from the center of the grid to the sonar, when d > OP1+ epsilon, the center point of the grid is already beyond the occupiable area measured by the sonar. However, the center point of the grid is not in the occupiable area, and it cannot be said that the grid is entirely absent from the occupiable area, and perhaps that a portion of the grid is in the occupiable area. Therefore, to determine whether the grid portion is in the occupiable region, the distance is determined to be 50 cm, and the relationship of d-50 to OP1+ ε is further compared:
when d-50 is less than OP 1-epsilon, and the grid part is in the occupied area of sonar measurement, the confidence value assignment calculation model of the empty area of sonar measurement is as follows:
when OP 1-epsilon < d-50 < OP1+ epsilon, the grid part is in the occupied area of the sonar measurement, and the confidence value assignment calculation model of the occupied area of the sonar measurement is as follows:
③ when d-50 > OP1+ ε, the grid is not in the occupiable region of the sonar measurement, then the computational model of the occupiable region beyond the sonar measurement is:
(4) establishing a combination rule of sonar data at different moments under a constraint condition:
(4a) the overpowering set under the grid state discrimination framework is as follows: dΩAnd (2) setting E & ltn & gt O & ltn & gt to be a constraint condition, wherein E & ltn & gt O & ltn & gt is directly classified as an obstacle condition. And calculating the reliability assignment of grids around the mobile robot under a judgment frame omega { E, O } according to the sonar sensor measurement data obtained in each simulation step.
(4b) And under the constraint condition E n O phi, fusing the fusion result of the grid state reliability at the previous t-1 moment with the grid reliability assignment calculated according to the sonar measurement data at the current t moment by adopting a DSmT (dynamic range measurement) mixed combination rule. Under the DSmT theoretical framework, the combined calculation of multiple evidences can be obtained by calculation recursion of a combination of two evidences, specifically as shown in fig. 5, where m1 and m2.. mn respectively represent state reliability assignments of the same grid at times t1, t 2.
The invention provides a sonar measurement modeling method under a DSmT theoretical framework, fuses information acquired by a plurality of sonars on a mobile robot body by applying a DSmT fusion algorithm, realizes the implementation and creation of a two-dimensional environment map, overcomes the limitation of the prior art, and provides a new method and thought for the map creation and navigation of the robot.
The technical effects of the present invention will be explained below by experiments and results of simulation.
Example 7
The mobile robot sonar data fusion method based on the DSmT theory is the same as that in the embodiment 1-6, the method takes the Pioneer 2 mobile robot as reference, and the measurement data of 16 sonar sensors are generated in a simulation mode and are respectively the measurement data of the front 8 sensors and the measurement data of the back 8 sensors of the mobile robot. The angular arrangement of the sonar sensors on the mobile robot is based on the arrangement of 8 sonar sensors in front of the mobile robot, as shown in fig. 6, a coordinate system (Xt Ot Zt) in fig. 6 is a mobile robot coordinate system, the sensors are arranged in 8 directions of 90 °, 50 °, 30 ° and 10 ° clockwise and counterclockwise of the mobile robot, the total of 8 sensors are arranged in front of the mobile robot, all the angular ranges can be covered, and the angular arrangement of 8 sonar sensors behind the robot is symmetrical to the front arrangement.
Sonar sensor parameters are shown in the following table:
TABLE 1 Sonar sensor principal parameters
The measurement data of 16 sonar sensors are obtained in simulation, the number of obtained samples is large, the measurement range of the sonar sensors is 0-300cm, target object data in a large distance range are obtained, the emission angle of the sonar sensors is 150, and each sonar sensor obtains the target object data in a large angle range.
According to simulation experiments, in the sonar sensor measurement model, the measurement error is 10cm and is small, the influence of the precision of parameters on output results is reduced, the data precision of the output results is improved, the results obtained after fusion are more correct and reliable, the judgment on the unknown environment where the mobile robot is located is more accurate, and a map which is more in line with actual conditions is created.
In summary, the mobile robot sonar data fusion method based on the DSmT theory provided by the invention aims to calculate the confidence level assignment of the surrounding grids of the mobile robot under the judgment frame Ω ═ E, O } through the measured data of the plurality of sonar sensors loaded on the robot body obtained in each simulation step, and fuse the confidence level assignments of different simulation steps under the DSmT theory frame, so that the fusion problem of the conflict data is solved, and the judgment of the state of the surrounding grids of the robot is more accurate. The method comprises the following implementation steps: (1) acquiring information through a sonar sensor, and establishing a measurement model of the sonar sensor; (2) according to a measurement model of the sonar sensor, dividing a sonar detection sector area range into: open and occupiable areas; (3) rasterizing a two-dimensional plane environment where the mobile robot is located, giving a discrimination framework of a grid state, and establishing a calculation model of reliability assignment under the discrimination framework; (4) and establishing DSmT mixed combination rules of sonar data at different moments under the constraint condition, and completing fusion of the sonar data of the mobile robot according to the DSmT mixed combination rules of the sonar data at different moments. The invention uses the DSmT fusion algorithm to fuse the information acquired by the plurality of sonars on the mobile robot body, realizes the implementation and the creation of a two-dimensional environment map, overcomes the limitations of the prior art on data applicability and data precision, judges the grid state around the robot more accurately, can be used for the map creation of the mobile robot in an unknown environment in practical application, and has wide application prospect.
Claims (5)
1. A mobile robot sonar data fusion method based on a DSmT theory is characterized by comprising the following steps:
(1) obtaining information through a sonar sensor, and establishing a measurement model of the sonar sensor:
the unknown environment that mobile robot place, for two-dimensional plane environment, the sonar sensor through a plurality of different directions that mobile robot body loaded obtains the information of surrounding environment in two-dimensional plane environment, establishes sonar sensor's measurement model: the method comprises the following steps of forming a sonar detection sector area range by using a sound wave emitting angle and a maximum measurement distance of sonar, and setting a measurement value returned by the sonar as a distance measurement value of a target closest to the sonar distance in the sector area;
(2) according to a measurement model of the sonar sensor, dividing a sonar detection sector area range into: open area and occupiable area:
(2a) an open area: [0, R-epsilon), [0, R-epsilon ], in the area, the probability of the existence of the obstacle is 0, wherein R refers to the measurement distance returned by the sonar, and epsilon refers to the measurement error of the sonar;
(2b) the occupiable area: [ R- ε, R + ε ], the probability of no obstacle in this region is 0;
(3) rasterizing a two-dimensional plane environment where the mobile robot is located, providing a discrimination framework in a grid state, and establishing a calculation model of reliability assignment under the discrimination framework:
(3a) rasterizing a two-dimensional plane environment where the mobile robot is located, wherein each grid represents the space size of 80 × 80cm, and judging the states of the grids to be three types: empty, with obstacles, unknown;
(3b) according to the DSmT theory, a decision frame Ω of the grid state is given: q ═ { E, O }, where E represents null and O represents an obstacle;
(3c) according to the division of a measurement model and a measurement area of a sonar sensor, establishing a reliability assignment calculation model of grids around a mobile robot under a judgment frame omega { E, O }, according to the area where the grids are located, dividing the reliability assignment calculation model into a reliability assignment calculation model of an open area of sonar measurement, a reliability assignment calculation model of an occupiable area of sonar measurement and a calculation model of the occupiable area exceeding the occupiable area of sonar measurement, wherein a DSmT theory is established on the basis of a super power set, the reliability assignment calculation model specifically comprises the reliability assignment of each element in a super power set formed by the judgment frame, the value range of the reliability assignment of each element is [0,1], and the reliability assignment of the element in the empty set in the super power set is required to be 0, and the sum of the reliability assignments of all the elements in the super power set is 1;
(4) and establishing DSmT mixed combination rules of sonar data at different moments under the constraint condition, and completing fusion of the sonar data of the mobile robot according to the DSmT mixed combination rules of the sonar data at different moments.
2. The DSmT theory-based mobile robot sonar data fusion method of claim 1, wherein the sonar numbers at different times established under the constraint condition in step 4According to the DSmT mixed combination rule, the combination credibility of the grid states obtained by a plurality of sonar sensors in different directions loaded by a mobile robot body under the conditions of being empty, having obstacles and unknown forms new combination credibility together, and the new combination credibility and the non-empty characteristic function
And multiplying to obtain the reliability assignment of the grid.
3. The mobile robot sonar data fusion method based on the DSmT theory according to claim 2, wherein the DSmT mixed combination rule of the sonar data at different times established under the constraint condition in step 4 is specifically calculated by the following formula:
wherein m is1(. and m)2(. for basic trust assignments from two different data sources, mμ(Ω) represents a confidence fusion result with a constraint condition, μ is the constraint condition E ═ O ═ Φ, where Φ is the absolute null set; c is the state of the grid, C is the element of DΩ，
Is a non-null featureFunction if
Then
The value is '0', otherwise '1',
for the set of constraints, Φ represents the absolute null set,
indicating a relative null condition, S1(C) For the combination rule of DSmT without any constraints, S2(C) Assigning the combined credibility of all absolute empty sets and relative empty sets to a combined rule under the condition of distributing a total unknown set and a relative unknown set, wherein the total unknown set is a union of all propositions in a judgment frame, S3(C) The combined credibility of the conflict focal elements is assigned to the combined rule under the condition of the union of the conflict focal elements.
4. The mobile robot sonar data fusion method based on the DSmT theory according to claim 1, 2, or 3, wherein the step 4 of completing sonar data fusion according to the DSmT mixed combination rule of sonar data at different times includes the steps of:
(4a) overpowering set D under grid state discrimination frame omegaΩComprises the following steps: dΩSetting E.n.O.O as constraint condition, where phi is absolute empty set, DΩAdd a pass-and operation to the propositional elements in the decision frame Ω: and U and traffic: n is a set of composite propositions;
(4b) and under the constraint condition E n O phi, fusing the fusion result of the grid state reliability at the previous t-1 moment and the grid reliability assignment calculated according to the sonar measurement data at the current t moment by adopting a DSmT mixed combination rule, and completing the fusion of the sonar data of the surrounding environment information, which is obtained by the mobile robot through the sonar sensor, in the two-dimensional plane of the unknown environment.
5. The mobile robot sonar data fusion method based on the DSmT theory as recited in claim 4, wherein the fusion result of the grid state confidence at the previous t-1 time and the grid confidence value calculated according to the sonar measurement data at the current t time are fused by using the DSmT hybrid combination rule in (4b), and the method specifically includes:
wherein m isij(omega) (C) represents the confidence fusion result of the ith row and jth column grids, and C belongs to DΩ，
For non-null feature functions, the fusion constraints are: e n O is phi; c values are respectively a set of over powers DΩ4 states of grids at phi, O, E and EU O in the process of fusion calculation and result mij(Ω) (C) is as follows:
(1) when C is phi, then mij(Φ)＝0；
(2) C ═ O, i.e., there is an obstacle state:
non-null feature function
Then: m is a unit ofij(O)＝S1(O)+S2(O)+S3(O), wherein:
S2(O)＝0，S3(O)＝0；
(3) c ═ E, i.e., null state:
non-null feature function
Then: m isij(E)＝S1(E)+S2(E)+S3(E) Wherein:
S2(E)＝0，S3(E)＝0；
(4) c ═ E ═ O, i.e. the unknown state:
non-null feature function
Then: m isij(E∪O)＝S1(E∪O)+S2(E∪O)+S3(E.U.O), wherein:
S2(E)＝0，
in the calculation process of the 4 state-confidence fusion of the grid,
represents the fused result of the grid state confidence at time t-1,
and representing the grid state reliability assignment calculated according to the sonar measurement data at the time t.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810649501.0A CN108802741B (en) | 2018-06-22 | 2018-06-22 | Mobile robot sonar data fusion method based on DSmT theory |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810649501.0A CN108802741B (en) | 2018-06-22 | 2018-06-22 | Mobile robot sonar data fusion method based on DSmT theory |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108802741A CN108802741A (en) | 2018-11-13 |
| CN108802741B true CN108802741B (en) | 2022-05-17 |
Family
ID=64084479
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810649501.0A Active CN108802741B (en) | 2018-06-22 | 2018-06-22 | Mobile robot sonar data fusion method based on DSmT theory |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108802741B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111160447B (en) * | 2019-12-25 | 2023-11-14 | 中国汽车技术研究中心有限公司 | Multi-sensor perception fusion method of autonomous parking positioning system based on DSmT theory |
| CN111208521B (en) * | 2020-01-14 | 2021-12-07 | 武汉理工大学 | Multi-beam forward-looking sonar underwater obstacle robust detection method |
| CN111931833B (en) * | 2020-07-30 | 2022-08-12 | 上海卫星工程研究所 | Multi-source data driving-based space-based multi-dimensional information fusion method and system |
| CN112612037B (en) * | 2020-12-01 | 2023-10-24 | 珠海一微半导体股份有限公司 | Fusion positioning method and mobile robot |
| CN112734878B (en) * | 2020-12-31 | 2023-06-20 | 南昌工学院 | Method, equipment and storage medium for detecting connectivity between two points of large grid map |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101413806A (en) * | 2008-11-07 | 2009-04-22 | 湖南大学 | Mobile robot grating map creating method of real-time data fusion |
| CN103778441A (en) * | 2014-02-26 | 2014-05-07 | 东南大学 | Dezert-Smaradache Theory (DSmT) and Hidden Markov Model (HMM) aircraft sequence target recognition method |
| CN107462892A (en) * | 2017-07-28 | 2017-12-12 | 深圳普思英察科技有限公司 | Mobile robot synchronous superposition method based on more sonacs |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7337086B2 (en) * | 2005-10-18 | 2008-02-26 | Honeywell International, Inc. | System and method for combining diagnostic evidences for turbine engine fault detection |
-
2018
- 2018-06-22 CN CN201810649501.0A patent/CN108802741B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101413806A (en) * | 2008-11-07 | 2009-04-22 | 湖南大学 | Mobile robot grating map creating method of real-time data fusion |
| CN103778441A (en) * | 2014-02-26 | 2014-05-07 | 东南大学 | Dezert-Smaradache Theory (DSmT) and Hidden Markov Model (HMM) aircraft sequence target recognition method |
| CN107462892A (en) * | 2017-07-28 | 2017-12-12 | 深圳普思英察科技有限公司 | Mobile robot synchronous superposition method based on more sonacs |
Non-Patent Citations (3)
| Title |
|---|
| A NOVEL APPROACH TO EVIDENCE COMBINATION IN BATTLEFIELD SITUATION ASSESSMENT USING DEZERT-SMARANDACHE THEORY;HUIMIN CHAI;《Proceedings of the 2013 International Conference on Machine Learning and Cybernetics》;20130717;第720-727页 * |
| 基于混合DSm模型的移动机器人动态环境地图构建;李鹏 等;《机器人》;20090131;第40-52页 * |
| 基于经典DSmT的Sonar栅格地图创建;李新德 等;《计算机应用研究》;20070331;第209-212页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108802741A (en) | 2018-11-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108802741B (en) | Mobile robot sonar data fusion method based on DSmT theory | |
| CN1940591B (en) | System and method of target tracking using sensor fusion | |
| CN111201448B (en) | Method and device for generating an inverted sensor model and method for identifying obstacles | |
| Park et al. | Radar localization and mapping for indoor disaster environments via multi-modal registration to prior LiDAR map | |
| CN110082747A (en) | Increase the method for the reliability of determining vehicle location based on multiple test points | |
| CN108780149B (en) | Method for improving the detection of at least one object in the surroundings of a motor vehicle by indirect measurement of a sensor, control unit, driver assistance system and motor vehicle | |
| JPH07116981A (en) | Guiding device for mobile robot | |
| WO2019204800A1 (en) | Method and system for generating high definition map | |
| JP2019512699A5 (en) | ||
| Percival et al. | Multihypothesis fusion of multipath over-the-horizon radar tracks | |
| Baskent et al. | Surface profile determination from multiple sonar data using morphological processing | |
| Han et al. | Localization and classification of target surfaces using two pairs of ultrasonic sensors | |
| US20210033706A1 (en) | Methods and systems for automatically labeling point cloud data | |
| Ayrulu et al. | Identification of target primitives with multiple decision-making sonars using evidential reasoning | |
| Lee et al. | Robust 3-dimension point cloud mapping in dynamic environment using point-wise static probability-based NDT scan-matching | |
| Ayrulu et al. | Reliability measure assignment to sonar for robust target differentiation | |
| Gasós et al. | A fuzzy approach to build sonar maps for mobile robots | |
| Blackman et al. | Multiple sensor data association and fusion in aerospace applications | |
| Yuan et al. | DSmT-based ultrasonic detection model for estimating indoor environment contour | |
| Hebel et al. | Change detection in urban areas by direct comparison of multi-view and multi-temporal ALS data | |
| Ayrulu et al. | Target identification with multiple logical sonars using evidential reasoning and simple majority voting | |
| Li et al. | Robot map building from sonar sensors and DSmT | |
| Dia et al. | General inverse sensor model for probabilistic occupancy grid maps using multi-target sensors | |
| Moita et al. | Multi-echo technique for feature detection and identification using simple sonar configurations | |
| Berthold et al. | Deriving spatial occupancy evidence from radar detection data |
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 |