CN101806905A - Navigation positioning method and device for agricultural machines - Google Patents
Navigation positioning method and device for agricultural machines Download PDFInfo
- Publication number
- CN101806905A CN101806905A CN201010112693A CN201010112693A CN101806905A CN 101806905 A CN101806905 A CN 101806905A CN 201010112693 A CN201010112693 A CN 201010112693A CN 201010112693 A CN201010112693 A CN 201010112693A CN 101806905 A CN101806905 A CN 101806905A
- Authority
- CN
- China
- Prior art keywords
- information
- neural network
- rbf neural
- agricultural machinery
- swarm optimization
- 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.)
- Pending
Links
Images
Landscapes
- Position Fixing By Use Of Radio Waves (AREA)
- Navigation (AREA)
Abstract
The invention discloses a navigation positioning method and a navigation positioning device for agricultural machines. The method comprises the following steps of: acquiring position information and attitude information of the agricultural machines; and acquiring positioning information corresponding to the position information and the attitude information of the agricultural machines based on an RBF neural network obtained by training through particle swarm optimization. The device comprises an information acquisition module and a positioning information acquisition module. In the technical scheme of the invention, the position information of the navigation positioning of the agricultural machines can be acquired based on the RBF neural network obtained by training through the particle swarm optimization, the navigation positioning information has high accuracy, and the cost of the navigation positioning can be reduced effectively.
Description
Technical field
The embodiment of the invention relates to the agricultural machines navigation technical field, relates in particular to a kind of agricultural machines navigation localization method and device.
Background technology
Navigator fix is the key in the agricultural machines navigation, the work quality of the accuracy affects agricultural machinery operation of navigator fix, it is the key factor of control agricultural machinery reliability service, therefore, how improving the agricultural machines navigation locating accuracy is to improve agricultural machinery running quality technical matters to be solved.
At present, the navigation locating method of agricultural machinery mainly comprises absolute fix method and relative positioning method.Wherein, commonly utilize the GPS navigation positioning system to carry out navigator fix in the absolute fix method, it can provide the absolute position for the GPS navigation location receiver round-the-clockly, information such as speed and direction, but, there is the satellite more than four could realize three-dimensional localization in this system requirements visual line of sight, satellite-signal is blocked or weakens and will have influence on bearing accuracy, cause the increase of navigation positioning error, and the bearing accuracy of GPS navigation positioning system is high more, price is just expensive more, thereby has limited the application of GPS navigation positioning system in the agricultural machines navigation location; Utilize machine vision and inertial navigation system etc. that the relative positioning method is common are carried out navigator fix, it mainly is to utilize sensor to obtain the attitude information of agricultural machinery, realization is to the navigator fix of agricultural machinery, but, the adaptability and the robustness of Vision Builder for Automated Inspection are relatively poor, can only be applicable to the navigator fix of the agricultural machinery of crop under specific environment, applicability is relatively poor, can't satisfy the accurate location of agricultural machinery.
In above-mentioned single navigate mode, costing an arm and a leg of absolute fix method, and the relatively poor problem of relative positioning method applicability, also proposed to utilize combined navigation locating method to realize the navigator fix of agricultural machinery in the prior art, it is mainly based on adopting RTK-DGPS and Inertial Measurement Unit IMU to obtain the preliminary locating information of agricultural machinery, by kalman filter method the Primary Location information that obtains is handled then, and the locating information of acquisition agricultural machinery, wherein, kalman filter method is based on mathematical model accurately, it is only applicable to linear system and system noise is the system of white Gaussian noise, in actual applications, owing to lack the priori of system works environmental baseline, be prone to the problem of filter divergence based on the calculating of Kalman filtering, filtering accuracy is poor, makes that the navigation and positioning accuracy of agricultural machinery is lower.
The inventor finds in realizing process of the present invention, in the existing navigation locating method, costs an arm and a leg based on the absolute fix method of single navigator fix, and the agricultural machines navigation positioning cost is higher, and relative positioning method low precision, applicability is relatively poor; In the integrated navigation location based on Kalman filtering, though can solve the expensive problem of absolute navigator fix, but, the kalman filter method condition of compatibility requires high, the feasible problem that is prone to filtering divergence based on the calculating of Kalman filtering, the precision of navigator fix is lower, can't satisfy the needs of the navigator fix of agricultural machinery.
Summary of the invention
The invention provides a kind of agricultural machines navigation localization method and device, when reducing the agricultural machines navigation positioning cost, can effectively improve the precision of navigator fix, improve the navigation effect of agricultural machinery.
The embodiment of the invention provides a kind of agricultural machines navigation localization method, comprising:
Obtain the positional information and the attitude information of agricultural machinery;
Based on the RBF neural network that obtains by the particle swarm optimization training, obtain positional information and the corresponding locating information of attitude information with described agricultural machinery.
Wherein, described positional information and the attitude information that obtains agricultural machinery comprises:
Utilize the DGPS receiver to obtain the GPS locator data of agricultural machinery, utilize attitude sensor to obtain the 3 d pose information of agricultural machinery, described 3 d pose information comprises roll, pitching and course information;
Described GPS locator data and 3 d pose information translation under the same coordinate system, are obtained the positional information and the attitude information of described agricultural machinery.
In addition, above-mentioned agricultural machines navigation localization method also can comprise: by the step of particle swarm optimization training RBF neural network, described step by particle swarm optimization training RBF neural network comprises:
Set up the RBF neural network, described RBF neural network comprises input layer, hidden layer and output layer, and wherein, the input information of described input layer comprises: the positional information of agricultural machinery, attitude information and velocity information;
Utilize particle swarm optimization that the RBF neural network is trained, determine the network parameter of RBF neural network, described network parameter comprise the hidden layer basis function the center, each is to width, network weight and hidden layer neuron number.
The described particle swarm optimization that utilizes is trained the RBF neural network, determines that the network parameter of RBF neural network comprises:
The center of RBF neural network hidden layer basis function, each is set to each dimensional vector of vector x to width and network weight;
Utilize the optimal location of particle swarm optimization search particulate group in the parameter vector space, then each dimensional vector of optimal location vector x promptly be RBF neural network hidden layer basis function the center, each is to width and network weight.
The described optimal location of particle swarm optimization search particulate group in the parameter vector space that utilize comprises:
Utilize the particle swarm optimization formula that the RBF neural network is trained, wherein, described particle swarm optimization formula is:
v
k+1=c
0v
k+c
1(pbest
k-x
k)+c
2(gbest
k-x
k)
x
k+1=x
k+v
k+1
Wherein, pbest
kIt is the position of the optimum solution that found of particle itself; Gbest
kThe position of representing the current optimum solution that finds of whole population; v
kIt is the particle's velocity vector; x
kIt is current particle position; c
0, c
1, c
2Expression group cognition coefficient, c
0Be the random number between (0,1), c
1, c
2Get the random number between (0,2), k is the sequence number of particle;
Utilize the fitness computing formula to calculate fitness, wherein, described fitness computing formula is:
Wherein, N is the training set sample number; The number of C network output neuron; Y is an idea output; I represents the sequence number of sample; J represents the neuron sequence number; t
J, iRepresent i sample j neuronic output valve.
Judge whether J reaches the square error ε of expection, if, then stop training to the RBF neural network, obtain the optimal location of particulate group in the parameter vector space; Otherwise, RBF neural network hidden layer basis function is increased 1, and utilize the particle swarm optimization formula to upgrade the speed and the position of particulate, continue RBF is trained.
In addition, described when utilizing particle swarm optimization that the RBF neural network is trained, RBF neural network hidden layer basis function number is set to 1 when training first.
Further, above-mentioned agricultural machines navigation localization method also can comprise:
According to the locating information of the described agricultural machinery that obtains, agricultural machinery is controlled.
The embodiment of the invention provides a kind of agricultural machines navigation locating device, comprising:
The information acquisition module is used to obtain the positional information and the attitude information of agricultural machinery;
The locating information acquisition module is used for obtaining positional information and the corresponding locating information of attitude information with described agricultural machinery based on training the RBF neural network that obtains by the particulate algorithm.
Described information acquisition module comprises:
Information acquisition unit is used to utilize the DGPS receiver to obtain the GPS locator data of agricultural machinery, utilizes attitude sensor to obtain the 3 d pose information of agricultural machinery, and described 3 d pose information comprises roll, pitching and course information;
The information translation unit is used for described GPS locator data and 3 d pose information translation obtaining the positional information and the attitude information of described agricultural machinery under the same coordinate system.
Above-mentioned agricultural machines navigation locating device also can comprise:
The RBF neural network is set up module, is used to set up the RBF neural network, and described RBF neural network comprises input layer, hidden layer and output layer, and wherein, the input information of described input layer comprises: the positional information of agricultural machinery, attitude information and velocity information;
RBF neural metwork training module, be used to utilize particle swarm optimization that the RBF neural network is trained, determine the network parameter of RBF neural network, described network parameter comprise the hidden layer basis function the center, each is to width, network weight and hidden layer neuron number.
Agricultural machines navigation localization method and device that the embodiment of the invention provides, can be by positional information and attitude information to the lower accuracy of the agricultural machinery that obtains, and obtain the RBF neural network by particle swarm optimization training, make that the training time of RBF neural network is shorter, the RBF neural network is approached real system more, the true character of reflection system, thus make that the navigator fix precision of information of output is higher, can effectively satisfy the needs of agricultural machines navigation location; In the technical solution of the present invention, only need to obtain the positional information and the attitude information of agricultural machinery lower accuracy, therefore, can effectively reduce the agricultural machines navigation positioning cost; In the technical solution of the present invention, during by particle swarm optimization training RBF neural network, can effectively improve the speed of convergence and the pace of learning of neural network, reduce the RBF neural metwork training time, improve training effectiveness.The embodiment of the invention can effectively improve the precision of navigator fix, thereby can realize the precision navigation of agricultural machinery in the cost that reduces navigator fix.
Description of drawings
Fig. 1 is the schematic flow sheet of agricultural machines navigation localization method embodiment of the present invention;
Fig. 2 obtains the positional information of agricultural machinery and the schematic flow sheet of attitude information among the agricultural machines navigation localization method embodiment of the present invention;
Fig. 3 is the schematic flow sheet of training RBF neural network among the agricultural machines navigation localization method embodiment of the present invention;
Fig. 4 is the structural representation of RBF neural network in the embodiment of the invention;
Fig. 5 is the structural representation of agricultural machines navigation locating device embodiment of the present invention;
Fig. 6 is the structural representation of information acquisition module among the agricultural machines navigation locating device embodiment of the present invention.
Embodiment
For the purpose, technical scheme and the advantage that make the embodiment of the invention clearer, below in conjunction with the accompanying drawing in the embodiment of the invention, technical scheme in the embodiment of the invention is clearly and completely described, obviously, described embodiment is the present invention's part embodiment, rather than whole embodiment.Based on the embodiment among the present invention, those of ordinary skills belong to the scope of protection of the invention not making the every other embodiment that is obtained under the creative work prerequisite.
Fig. 1 is the schematic flow sheet of agricultural machines navigation localization method embodiment of the present invention.Particularly, as shown in Figure 1, the present embodiment navigation locating method comprises:
Present embodiment can be applicable in the navigator fix of agricultural machinery, particularly, can pass through common sensor, gather and obtain initial lower positional information and the attitude information of precision of agricultural machinery, and based on (Particle Swarm Optimization PSO) trains the RBF neural network that obtains, and obtains the navigator fix information of agricultural machinery by particle swarm optimization, the precision height of the navigator fix information that obtains can be just provides foundation for the navigator fix and the control of agricultural machinery.
As can be seen, the RBF neural network that embodiment of the invention utilization obtains by the particle swarm optimization training, after the positional information of agricultural machinery and attitude information merged, output has the navigator fix information of degree of precision, simultaneously, utilizes particle swarm optimization to obtain optimum RBF neural network parameter, make the RBF neural network sample training time shorter, the RBF neural network that obtains is approached real system more, can reflect the truth of system, can effectively satisfy the needs of agricultural machines navigation location; Present embodiment only need can obtain to have the navigator fix information of degree of precision by common sensor, makes that the cost of navigator fix is lower, is convenient to promotion and application in the navigator fix of agricultural machinery.
Fig. 2 obtains the positional information of agricultural machinery and the schematic flow sheet of attitude information among the agricultural machines navigation localization method embodiment of the present invention.In the present embodiment, can be by differential Global Positioning System (the Difference Global Positioning System of low precision, DGPS) obtain the GPS locator data of agricultural machinery, to reduce the agricultural machines navigation positioning cost, particularly, in the step 101 shown in above-mentioned Fig. 1, described positional information and the attitude information that obtains agricultural machinery can comprise the steps:
In the present embodiment, can gather the GPS locator data of agricultural machinery, and can utilize attitude sensor to collect the attitude information of agricultural machinery by adopting low-cost low DGPS receiver with lower accuracy.Particularly, can lead the latitude and longitude information of gathering under the agricultural machinery geocentric coordinate system by the DGPS receiver, and can convert thereof into the vehicle axis system at agricultural machinery place,, be convenient to the calculating of information so that coordinate system is unified; Simultaneously, can utilize attitude sensor to obtain the 3 d pose information of agricultural machinery, comprise course heading information, roll and luffing angle information, the attitude information of acquisition is a vehicle axis system, thereby the coordinate system of the agricultural machinery that obtains is unified, be convenient to follow-up application.In addition, the frequency acquisition of the attitude sensor in the present embodiment is 10HZ, the employing frequency of DGPS is 1HZ, therefore, for improving the output frequency of navigation positional device, present embodiment can improve the sample frequency of DGPS by using the interpolate value method, thereby improve the precision of navigator fix, the locating information of high frequency is provided for navigator fix.
In addition, for improving the precision of navigator fix, the attitude compensation is carried out in rocking of also can causing jolting owing to the field road, causes the reduction of navigation and positioning accuracy to avoid the attitude information error.
In the present embodiment, after obtaining agricultural machinery preliminary GPS locator data and attitude information by DGPS receiver and attitude sensor, can utilize the mode of coordinate system conversion, GPS locator data and attitude information are changed in the vehicle coordinate at agricultural machinery place, make GPS locator data and attitude information be transformed in the same coordinate system, and be described as follows:
In the DGPS system, because employing is the WGS-84 coordinate system, belong to geocentric coordinate system, for the GPS locator data that the DGPS receiver is collected can be used for navigator fix control, need carry out coordinate conversion to the GPS locator data that collects.Since on the earth position of any point both can be expressed as (h), wherein, λ is a longitude for λ, φ,
Be dimension, h is an elevation, i.e. coordinate under the earth coordinates simultaneously, can also be expressed as that (ze), i.e. coordinate under the geocentric coordinate system, and described geocentric coordinate system promptly is the WGS-84 coordinate system for xe, ye.And the conversion formula between geocentric coordinate system and the earth coordinates is:
Wherein,
The radius of curvature in prime vertical of the earth, e are first excentricity of the earth.
In addition, the transition matrix between geocentric coordinate system and the navigation coordinate system is:
Transition matrix between navigation coordinate system and the vehicle axis system is:
P
n=C
e2nP
e;P
n=C
b2nP
b
Wherein, P
nBe navigation coordinate system, P
eBe earth coordinates, P
bBe vehicle axis system, C
E2nBe the transition matrix between earth coordinates and the navigation coordinate system, C
B2nBe the transition matrix between vehicle axis system and the navigation coordinate system;
Be respectively roll angle, luffing angle and course heading by the agricultural machinery of attitude sensor acquisition.
Present embodiment is by above-mentioned coordinate conversion matrix, and the attitude information that gps data that the DGPS receiver can be obtained and attitude sensor obtain all changes into vehicle coordinate.
Fig. 3 is the schematic flow sheet of training RBF neural network among the agricultural machines navigation localization method embodiment of the present invention; Fig. 4 is the structural representation of RBF neural network in the embodiment of the invention.In the present embodiment, the RBF neural network is to obtain by the particle swarm optimization training, particularly, as shown in Figure 3, can comprise the steps: by particle swarm optimization training RBF neural network
Step 201, set up the RBF neural network, described RBF neural network comprises input layer, hidden layer and output layer, and wherein, the input information of described input layer comprises: the positional information of agricultural machinery, attitude information and velocity information.
In the present embodiment, the RBF neural network of foundation can comprise three layers, is respectively: input layer, hidden layer and output layer, and concrete structure can be with reference to figure 4, and as can be seen, input layer is made up of some source points, and couples together with external environment.Particularly, in the present embodiment, attitude information that the GPS locator data that the DGPS receiver can be collected, attitude sensor obtain and velocity information etc. are as input layer, the effect of hidden layer is the nonlinear transformation of carrying out from the input space to the hidden layer space, action function in the hidden layer node, be that the hidden layer basis function will produce response to input signal in the part, output layer is linear, and its effect is that the signal to input layer provides response.Below embodiment of the invention RBF neural network is described:
In the RBF neural network, need carry out normalization to the pre-service of input information data.Therefore, the sample data of fan-in network and check data must be changed with same ratio, particularly, can adopt linear conversion method to change.Maximal value and the minimum value of supposing variable are respectively X
MaxAnd X
Min, and the physical constraints scope of network is A
MaxAnd A
Min, variable X can convert A to by following formula so:
A=r(X-X
min)+A
min
Wherein,
For the output valve A of network, then can be converted to variable X by following formula:
The hidden layer basis function generally uses Gaussian function as basis function, is specially:
R
i(x)=exp(-‖x-c
i‖
2/2σ
2 i),i=1,2,...,m
Wherein, x is a n dimension input vector; c
iBe the center of i basis function, have identical dimension with x; σ
iBe the normalized factor of i basis function, it has determined the width of basis function around central point, ‖ x-c
i‖ is vector x-c
iThe Euclidean norm, expression x and c
iBetween distance, m is the hidden layer neuron number, therefore, the center c of the hidden layer basis function of RBF neural network
i, width parameter σ
i, and network weight w
iDetermine it is very important.
The RBF neural network is output as:
Step 202, utilize particle swarm optimization that the RBF neural network is trained, determine the network parameter of RBF neural network, described network parameter comprise the hidden layer basis function the center, each is to width, network weight and hidden layer neuron number.
Present embodiment adopts particle swarm optimization that the RBF neural network is trained, and with the network parameter of optimization RBF neural network, and the last parameter that will determine is as the center of the hidden layer basis function of RBF neural network, respectively to width and network weight.
Particularly, in this step, utilize particle swarm optimization that the RBF neural network is trained, can comprise the steps:
Step 2021, utilize the particle swarm optimization formula that the RBF neural network is trained, wherein, described particle swarm optimization formula is:
v
k+1=c
0v
k+c
1(pbest
k-x
k)+c
2(gbest
k-x
k)
x
k+1=x
k+v
k+1
Wherein, v
kIt is the particle's velocity vector; x
kIt is current particle position; c
0, c
1, c
2Expression group cognition coefficient, c
0Be the random number between (0,1), c
1, c
2Get the random number between (0,2);
Step 2022, utilize the fitness computing formula to calculate fitness, wherein, described fitness computing formula is:
Wherein, N is the training set sample number; The number of C network output neuron; Y is an idea output; I represents the sequence number of sample; J represents the neuron sequence number; t
J, iRepresent i sample j neuronic output valve.
Whether step 2023, judgement J reach the square error ε of expection, if, then stop training to the RBF neural network, obtain the optimal location of particulate group in the parameter vector space; Otherwise, RBF neural network hidden layer basis function is increased 1, and utilize the particle swarm optimization formula to upgrade the speed and the position of particulate, continue RBF is trained.
Particularly, suppose to optimize in the space in the m dimension, by the population that n particulate formed, the current location of i particulate can be expressed as X
i=(x
I1, x
I2, x
I3..., x
Im), the position of each particulate is the problem of asking, it is a feasible solution of each parameter of neural network, therefore, the component of each particle individuality in the population can be mapped as the parameter of network, thereby constitute a RBF neural network, simultaneously, the training sample that the individual corresponding neural network of each particle is imported is trained, obtain corresponding fitness value, make the particulate of fitness value minimum be optimum solution by the fitness computing function.
At first, the initialized location vector x then, is utilized the optimal location of particle swarm optimization search particulate group in the parameter vector space, behind the position of determining particulate, can determine the structure and the parameter of RBF neural network, thereby finish the training of RBF neural network.
In the present embodiment, can train, till training reaches the expection square error by utilizing particle swarm optimization.Particularly, the particle swarm optimization formula can be expressed as:
v
k+1=c
0v
k+c
1(pbest
k-x
k)+c
2(gbest
k-x
k)
x
k+1=x
k+v
k+1
Wherein, v
kIt is the particle's velocity vector; x
kIt is current particle position; c
0, c
1, c
2Expression group cognition coefficient, c
0Be the random number between (0,1), c
1, c
2Get the random number between (0,2); And each dimension particle rapidity all is limited in a maximal rate v
MaxIf the speed after certain one dimension upgrades surpasses v
Max, the speed of this one dimension is restricted to v so
Max, i.e. v
k>v
MaxThe time, v
k=v
Max, perhaps v
k<-v
MaxThe time, v
k=-v
Max
Utilize the input and output data of RBF neural network, can error of calculation index, i.e. fitness, computing formula is as follows:
Wherein, N is the training set sample number; The number of C network output neuron; Y is an idea output; I represents the sequence number of sample; J represents the neuron sequence number; t
J, iRepresent i sample j neuronic output valve.
When J has reached the square error ε of expection, then can withdraw from training, at this moment the optimal location vector x respectively tie up parameter be the hidden layer basis function that the RBF neural metwork training obtains the center, each is to width and network weight, thereby finish the training of RBF neural network.
In addition, when utilizing particle swarm optimization that the RBF neural network is trained, RBF neural network hidden layer basis function number is set to 1 in the time of can training first.Particularly, in the present embodiment, when utilizing particle swarm optimization that the RBF neural network is trained, the number of hidden layer basis function is not predetermined, and its number is dynamic change and determining in whole training process, when initialization, the hidden layer number can be initialized as 1, be that particle swarm optimization is only trained the RBF neural network of a hidden layer,, then can withdraw from circulation if reached the square error of expection, expression is met the RBF neural network of condition, and network training is finished; Otherwise, if the RBF neural network occurring can not meet the desired the square error requirement in current hidden layer number, should increase the number of hidden layer basis function, and utilize particle swarm optimization to train once more, till training reaches the expection square error.Suppose that hidden layer includes k neuron, then unknown parameter is k basis function center c
Ik, k central point width parameter σ
Ik, k network weight parameter ω
Ik, in addition, also comprise an input block and an output unit, then initialization vector x is the 3*k dimensional vector, each group 3*k dimensional vector is represented the relevant parameter of one group of RBF neural network.
By the parameter after optimizing, can set up the RBF neural network, and can be with the input of the GPS locator data, attitude information and the velocity information that obtain as the RBF neural network, and after calculating through the RBF neural network, output corresponding navigation locating information.
In the embodiment of the invention, by particle swarm optimization the parameter in the RBF network is optimized, thereby determines optimum network parameter, make the RBF network performance approach actual conditions more, the feasible navigation positioning system that obtains can effectively satisfy the location needs of agricultural machines navigation accurately and reliably.In addition, when present embodiment is trained the RBF network by particle swarm optimization, can effectively avoid the defective of existing kalman filter method, not be subjected to the interference of system noise, reduce systematic error, can form continuously, stable, relative accuracy navigator fix information preferably.
In addition, the inventor verifies to embodiment of the invention technical scheme by the method for right software emulation that also concrete proof procedure is as follows:
(1) while is installed the high-precision GPS receiver on the vehicle of experiment, and low precision GPS receiver and attitude sensor, and vehicle is exercised along straight line.Wherein, the high-precision GPS receiver adopts GPS332+MS7500 equipment, and error range is in 2cm, and the signals collecting frequency is 10HZ; Low precision GPS adopts unit GPS132, and its positioning error is about 2m, and the signals collecting frequency is 1HZ, and the signals collecting frequency of attitude sensor is 10HZ.
(2) utilize Kalman filtering algorithm and respectively based on the RBF neural network of particle swarm optimization, information to low precision GPS and attitude sensor acquisition is handled, and the positional information that obtains with high-precision GPS is as the reference benchmark, and the position location deviation that obtains at last is as follows respectively:
Kalman filtering algorithm | RBF neural network based on particle swarm optimization | |
Position deviation after the calculating | ??1.13m | ?0.56m |
By emulation as can be seen, can approach real system more accurately based on the RBF neural network that particle swarm optimization is set up, the actual conditions that can reflect system are more exactly utilized the bearing accuracy of the output of the RBF neural network that particle swarm optimization training obtains to obtain bearing accuracy than the kalman filter method of routine and are exceeded about 50%.In addition, by particle swarm optimization the RBF neural network is trained, obtain the optimal value of network parameter in the RBF neural network, can under the situation of less hidden layer number, finish computation optimization function similarly, the domestication time of network is shorter, Chang Gui kalman filter method relatively, the time of its training network can improve about 30%.
Fig. 5 is the structural representation of agricultural machines navigation locating device embodiment of the present invention.Particularly, as shown in Figure 5, the present embodiment device comprises information acquisition module 1 and locating information acquisition module 2, wherein:
Locating information acquisition module 2 is used for obtaining positional information and the corresponding locating information of attitude information with described agricultural machinery based on training the RBF neural network that obtains by the particulate algorithm.
Fig. 6 is the structural representation of information acquisition module among the agricultural machines navigation locating device embodiment of the present invention.As shown in Figure 6, information acquisition module 1 specifically can comprise information acquisition unit 11 and information translation unit 12 in the present embodiment, wherein:
In addition, as shown in Figure 5, the present embodiment device can comprise that also the RBF neural network sets up module 3 and RBF neural metwork training module 4, wherein:
The RBF neural network is set up module 3, is used to set up the RBF neural network, and described RBF neural network comprises input layer, hidden layer and output layer, and wherein, the input information of described input layer bag comprises: positional information, attitude information and the velocity information of drawing together agricultural machinery;
RBF neural metwork training module 4, be used to utilize particle swarm optimization that the RBF neural network is trained, determine the network parameter of RBF neural network, described network parameter comprise the hidden layer basis function the center, each is to width, network weight and hidden layer neuron number.
Present embodiment can be applicable in the navigator fix of agricultural machinery, particularly, can pass through common sensor, gather and obtain initial lower positional information and the attitude information of precision of agricultural machinery, and based on the RBF neural network that obtains by the particle swarm optimization training, obtain the navigator fix information of agricultural machinery, the precision height of the navigator fix information of acquisition can be just provides foundation for the navigator fix and the control of agricultural machinery.Particularly, can not repeat them here with reference to the explanation among the invention described above method embodiment.
The embodiment of the invention is passed through positional information and the attitude information to the lower accuracy of the agricultural machinery that obtains, utilization obtains the RBF neural network by the particle swarm optimization training and merges, acquisition has the navigator fix information of degree of precision, simultaneously, utilize the time of particle swarm optimization training RBF neural network shorter, the RBF neural network that obtains is approached real system more, can reflect the truth of system, can effectively satisfy the needs of agricultural machines navigation location; In the technical solution of the present invention, only need to obtain the positional information and the attitude information of agricultural machinery lower accuracy, therefore, can effectively reduce the agricultural machines navigation positioning cost, be convenient to the promotion and application of navigator fix in agricultural machinery; In the technical solution of the present invention, during by particle swarm optimization training RBF neural network, can effectively improve the speed of convergence and the pace of learning of neural network, reduce the RBF neural metwork training time, improve training effectiveness.
One of ordinary skill in the art will appreciate that: all or part of step that realizes said method embodiment can be finished by the relevant hardware of programmed instruction, aforesaid program can be stored in the computer read/write memory medium, this program is carried out the step that comprises said method embodiment when carrying out; And aforesaid storage medium comprises: various media that can be program code stored such as ROM, RAM, magnetic disc or CD.
It should be noted that at last: above embodiment only in order to technical scheme of the present invention to be described, is not intended to limit; Although with reference to previous embodiment the present invention is had been described in detail, those of ordinary skill in the art is to be understood that: it still can be made amendment to the technical scheme that aforementioned each embodiment put down in writing, and perhaps part technical characterictic wherein is equal to replacement; And these modifications or replacement do not make the essence of appropriate technical solution break away from the spirit and scope of various embodiments of the present invention technical scheme.
Claims (10)
1. an agricultural machines navigation localization method is characterized in that, comprising:
Obtain the positional information and the attitude information of agricultural machinery;
Based on the RBF neural network that obtains by the particle swarm optimization training, obtain positional information and the corresponding locating information of attitude information with described agricultural machinery.
2. agricultural machines navigation localization method according to claim 1 is characterized in that, described positional information and the attitude information that obtains agricultural machinery comprises:
Utilize the DGPS receiver to obtain the GPS locator data of agricultural machinery, utilize attitude sensor to obtain the 3 d pose information of agricultural machinery, described 3 d pose information comprises roll, pitching and course information;
Described GPS locator data and 3 d pose information translation under the same coordinate system, are obtained the positional information and the attitude information of described agricultural machinery.
3. agricultural machines navigation localization method according to claim 1 and 2 is characterized in that, also comprises: by the step of particle swarm optimization training RBF neural network, described step by particle swarm optimization training RBF neural network comprises:
Set up the RBF neural network, described RBF neural network comprises input layer, hidden layer and output layer, and wherein, the input information of described input layer comprises: the positional information of agricultural machinery, attitude information and velocity information;
Utilize particle swarm optimization that the RBF neural network is trained, determine the network parameter of RBF neural network, described network parameter comprise the hidden layer basis function the center, each is to width, network weight and hidden layer neuron number.
4. agricultural machines navigation localization method according to claim 3 is characterized in that, the described particle swarm optimization that utilizes is trained the RBF neural network, determines that the network parameter of RBF neural network comprises:
The center of RBF neural network hidden layer basis function, each is set to each dimensional vector of vector x to width and network weight;
Utilize the optimal location of particle swarm optimization search particulate group in the parameter vector space, then each dimensional vector of optimal location vector x promptly be RBF neural network hidden layer basis function the center, each is to width and network weight.
5. agricultural machines navigation localization method according to claim 4 is characterized in that, the described optimal location of particle swarm optimization search particulate group in the parameter vector space that utilize comprises:
Utilize the particle swarm optimization formula that the RBF neural network is trained, wherein, described particle swarm optimization formula is:
v
k+1=c
0v
k+c
1(pbest
k-x
k)+c
2(gbest
k-x
k)
x
k+1=x
k+v
k+1
Wherein, pbest
kIt is the position of the optimum solution that found of particle itself; Gbest
kThe position of representing the current optimum solution that finds of whole population; v
kIt is the particle's velocity vector; x
kIt is current particle position; c
0, c
1, c
2Expression group cognition coefficient, c
0Be the random number between (0,1), c
1, c
2Get the random number between (0,2), k is the sequence number of particle;
Utilize the fitness computing formula to calculate fitness, wherein, described fitness computing formula is:
Wherein, N is the training set sample number; The number of C network output neuron; Y is an idea output; I represents the sequence number of sample; J represents the neuron sequence number; t
J, iRepresent i sample j neuronic output valve;
Judge whether J reaches the square error ε of expection, if, then stop training to the RBF neural network, obtain the optimal location of particulate group in the parameter vector space; Otherwise, RBF neural network hidden layer basis function is increased 1, and utilize the particle swarm optimization formula to upgrade the speed and the position of particulate, continue RBF is trained.
6. agricultural machines navigation localization method according to claim 5 is characterized in that, described when utilizing particle swarm optimization that the RBF neural network is trained, RBF neural network hidden layer basis function number is set to 1 when training first.
7. agricultural machines navigation localization method according to claim 1 is characterized in that, also comprises:
According to the locating information of the described agricultural machinery that obtains, agricultural machinery is controlled.
8. an agricultural machines navigation locating device is characterized in that, comprising:
The information acquisition module is used to obtain the positional information and the attitude information of agricultural machinery;
The locating information acquisition module is used for obtaining positional information and the corresponding locating information of attitude information with described agricultural machinery based on training the RBF neural network that obtains by the particulate algorithm.
9. agricultural machines navigation locating device according to claim 8 is characterized in that, described information acquisition module comprises:
Information acquisition unit is used to utilize the DGPS receiver to obtain the GPS locator data of agricultural machinery, utilizes attitude sensor to obtain the 3 d pose information of agricultural machinery, and described 3 d pose information comprises roll, pitching and course information;
The information translation unit is used for described GPS locator data and 3 d pose information translation obtaining the positional information and the attitude information of described agricultural machinery under the same coordinate system.
10. agricultural machines navigation locating device according to claim 8 is characterized in that, also comprises:
The RBF neural network is set up module, is used to set up the RBF neural network, and described RBF neural network comprises input layer, hidden layer and output layer, and wherein, the input information of described input layer comprises: the positional information of agricultural machinery, attitude information and velocity information;
RBF neural metwork training module, be used to utilize particle swarm optimization that the RBF neural network is trained, determine the network parameter of RBF neural network, described network parameter comprise the hidden layer basis function the center, each is to width, network weight and hidden layer neuron number.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201010112693A CN101806905A (en) | 2010-02-22 | 2010-02-22 | Navigation positioning method and device for agricultural machines |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201010112693A CN101806905A (en) | 2010-02-22 | 2010-02-22 | Navigation positioning method and device for agricultural machines |
Publications (1)
Publication Number | Publication Date |
---|---|
CN101806905A true CN101806905A (en) | 2010-08-18 |
Family
ID=42608782
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201010112693A Pending CN101806905A (en) | 2010-02-22 | 2010-02-22 | Navigation positioning method and device for agricultural machines |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN101806905A (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104330084A (en) * | 2014-11-13 | 2015-02-04 | 东南大学 | Neural network assisted integrated navigation method for underwater vehicle |
US9772625B2 (en) | 2014-05-12 | 2017-09-26 | Deere & Company | Model referenced management and control of a worksite |
CN108345021A (en) * | 2018-01-19 | 2018-07-31 | 东南大学 | A kind of Doppler radar assistant GPS/INS vehicle speed measuring methods |
US10114348B2 (en) | 2014-05-12 | 2018-10-30 | Deere & Company | Communication system for closed loop control of a worksite |
CN110632636A (en) * | 2019-09-11 | 2019-12-31 | 桂林电子科技大学 | Carrier attitude estimation method based on Elman neural network |
CN111674360A (en) * | 2019-01-31 | 2020-09-18 | 青岛科技大学 | Method for establishing distinguishing sample model in vehicle tracking system based on block chain |
CN112954585A (en) * | 2021-01-29 | 2021-06-11 | 华南农业大学 | UWB-based agricultural machine field positioning system and method |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1382997A (en) * | 2002-06-13 | 2002-12-04 | 上海交通大学 | Precise tracking method based on nerve network for moving target |
US6539304B1 (en) * | 2000-09-14 | 2003-03-25 | Sirf Technology, Inc. | GPS navigation system using neural networks |
CN101078935A (en) * | 2007-06-28 | 2007-11-28 | 华南农业大学 | Agricultural machine path tracking control method based on nerve network |
-
2010
- 2010-02-22 CN CN201010112693A patent/CN101806905A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6539304B1 (en) * | 2000-09-14 | 2003-03-25 | Sirf Technology, Inc. | GPS navigation system using neural networks |
CN1382997A (en) * | 2002-06-13 | 2002-12-04 | 上海交通大学 | Precise tracking method based on nerve network for moving target |
CN101078935A (en) * | 2007-06-28 | 2007-11-28 | 华南农业大学 | Agricultural machine path tracking control method based on nerve network |
Non-Patent Citations (12)
Title |
---|
《2007 仪表,自动化及先进集成技术大会论文集(一)》 20070831 付培众等 基于PSO-RBF的神经网络及其应用研究 《仪器仪表学报》杂志社 464~466 1~10 第28卷, * |
《中国博硕士学文论文数据库》 20050915 冯雷 基于GPS和传感技术的农用车辆自动导航***的研究 全文 1~10 , 第5期 * |
《传感技术学报》 20080515 薛晗等 基于无线传感器网络的移动机器人智能导航算法 834~840 1~10 第21卷, 第05期 * |
《农业机械学报》 20020930 周俊等 自主车辆导航***中的多传感器融合技术 113~116 1~10 第33卷, 第5期 * |
《农业机械学报》 20070531 陈军等 基于神经网络的农用车辆自动跟踪控制 131~133 1~10 第38卷, 第5期 * |
《数据采集与处理》 20060630 陈建勇等 基于RBF神经网络的组合导航融合算法 198~202 1~10 第21卷, 第2期 * |
《计算机与现代化》 20090415 贺文阳等 RBF神经网络的混合微粒群学习算法 35~38 1~10 , 第04期 * |
李康等: "GPS坐标系的转换及其在姿态求解中的应用 ", 《指挥控制与仿真》 * |
李康等: "GPS坐标系的转换及其在姿态求解中的应用", 《指挥控制与仿真》, vol. 30, no. 05, 15 October 2008 (2008-10-15), pages 113 - 118 * |
林雪原等: "GPS/罗兰C/SINS/AHRS组合导航***及试验 ", 《电子科技大学学报》 * |
林雪原等: "GPS/罗兰C/SINS/AHRS组合导航***及试验", 《电子科技大学学报》, vol. 37, no. 01, 30 January 2008 (2008-01-30), pages 4 - 7 * |
马云峰: "MSINS/GPS组合导航***及其数据融合技术研究", 《中国博士学位论文全文数据库 工程科技II辑》 * |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9772625B2 (en) | 2014-05-12 | 2017-09-26 | Deere & Company | Model referenced management and control of a worksite |
US10114348B2 (en) | 2014-05-12 | 2018-10-30 | Deere & Company | Communication system for closed loop control of a worksite |
US10705490B2 (en) | 2014-05-12 | 2020-07-07 | Deere & Company | Communication system for closed loop control of a worksite |
CN104330084A (en) * | 2014-11-13 | 2015-02-04 | 东南大学 | Neural network assisted integrated navigation method for underwater vehicle |
CN104330084B (en) * | 2014-11-13 | 2017-06-16 | 东南大学 | A kind of submarine navigation device neural network aiding Combinated navigation method |
CN108345021A (en) * | 2018-01-19 | 2018-07-31 | 东南大学 | A kind of Doppler radar assistant GPS/INS vehicle speed measuring methods |
CN111674360A (en) * | 2019-01-31 | 2020-09-18 | 青岛科技大学 | Method for establishing distinguishing sample model in vehicle tracking system based on block chain |
CN110632636A (en) * | 2019-09-11 | 2019-12-31 | 桂林电子科技大学 | Carrier attitude estimation method based on Elman neural network |
CN110632636B (en) * | 2019-09-11 | 2021-10-22 | 桂林电子科技大学 | Carrier attitude estimation method based on Elman neural network |
CN112954585A (en) * | 2021-01-29 | 2021-06-11 | 华南农业大学 | UWB-based agricultural machine field positioning system and method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN101806905A (en) | Navigation positioning method and device for agricultural machines | |
CN101285686B (en) | Agricultural machines navigation hierarchical positioning process and system | |
CN111596652A (en) | Pesticide spraying intelligent agricultural machinery path finding navigation control method and system | |
CN102819029B (en) | Supercompact combination satellite navigation receiver | |
CN102322862B (en) | Method for determining absolute orbit and relative orbit of formation flight satellite | |
Pfaff et al. | Towards mapping of cities | |
CN111982106A (en) | Navigation method, navigation device, storage medium and electronic device | |
EP4024328B1 (en) | Method and apparatus for map fusion, server and storage medium | |
CN104061932A (en) | Method for navigation positioning by using gravitation vector and gradient tensor | |
CN110187372B (en) | Combined navigation method and system in low-speed unmanned vehicle park | |
CN102928858A (en) | GNSS (Global Navigation Satellite System) single-point dynamic positioning method based on improved expanded Kalman filtering | |
CN106643489A (en) | Ground quick response (QR) code calibration method and device based on laser tracker | |
CN112904317A (en) | Calibration method for multi-laser radar and GNSS-INS system | |
CN105547300A (en) | All-source navigation system and method used for AUV (Autonomous Underwater Vehicle) | |
CN110186661A (en) | The forward kinematics solution method for solving of the parallel institution of the branch of pair containing UP | |
CN102607558A (en) | Relative deformation measuring method based on inertial measurement units | |
Guo et al. | Shallow-sea application of an intelligent fusion module for low-cost sensors in AUV | |
CN103292813B (en) | A kind of information filter method improving water surface ship formation navigation accuracy | |
CN102052925A (en) | Adaptive area scene matching method based on spatial relationship constraint | |
Guo et al. | Intelligent assistance positioning methodology based on modified iSAM for AUV using low-cost sensors | |
Qing et al. | Weighted total least squares for the visual localization of a planetary rover | |
CN106352880A (en) | Method for determining inertial navigation algorithm framework based on nonlinear network | |
Kitsukawa et al. | Robustness Evaluation of Vehicle Localization in 3D Map Using Convergence of Scan Matching | |
Cui et al. | Design and development of GPS navigation and positioning system based on AutoCAD platform | |
CN112747752A (en) | Vehicle positioning method, device, equipment and storage medium based on laser odometer |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C12 | Rejection of a patent application after its publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20100818 |