CN111983926B - Maximum co-entropy extended ellipsoid collective filtering method - Google Patents
Maximum co-entropy extended ellipsoid collective filtering method Download PDFInfo
- Publication number
- CN111983926B CN111983926B CN202010893973.8A CN202010893973A CN111983926B CN 111983926 B CN111983926 B CN 111983926B CN 202010893973 A CN202010893973 A CN 202010893973A CN 111983926 B CN111983926 B CN 111983926B
- Authority
- CN
- China
- Prior art keywords
- ellipsoid
- matrix
- representing
- error
- system state
- 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
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
- G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
- G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
- G05B13/042—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators in which a parameter or coefficient is automatically adjusted to optimise the performance
Abstract
The invention provides a maximum co-entropy extended ellipsoid collective filtering method, belongs to the system information processing technology in navigation positioning time service in the field of robots, and particularly relates to a nonlinear state space model of a robot motion carrier constructed by using navigation positioning sensing data. The invention discloses an ellipsoid collective member filtering algorithm based on a minimum mean square error criterion, wherein a maximum covex MCC criterion is introduced, a new system noise expression is obtained by carrying out system model expansion operation by combining a nonlinear system prediction noise error and observation noise, and an error cost function is constructed based on the maximum covex criterion between a system state variable prediction vector and an observation vector, so that an observation updating calculation process of the ellipsoid collective member filtering algorithm is designed, and an ellipsoid collective member filtering algorithm calculation frame based on the maximum covex is constructed; the problem of the calculation stability of the traditional centralized filtering algorithm is effectively solved, and therefore the robot carrier combined navigation and positioning service function is completed.
Description
Technical Field
The invention relates to the technical field of robot navigation and positioning, in particular to a maximum covariance extended ellipsoid collective filtering method.
Background
The objective of ensemble filtering is to update a Feasible Parameter Set (FPS) containing true values of system parameters in real time according to the proposed model, noise bound and observed values, and this method replaces the traditional estimation vector description system with an ensemble, and any Parameter in the Feasible Parameter Set can be considered Feasible. The membership filtering algorithm can be divided into an ellipsoid algorithm, a polyhedron algorithm, a box algorithm, a full-symmetry multi-cell algorithm, a super-parallel acceleration method, an interval algorithm and the like according to different feasible sets of description parameters, and the feasible sets of the parameters are approximately described by adopting corresponding geometric shapes respectively.
The ellipsoid collective-member filtering algorithm is firstly proposed by Schweppe, an ellipsoid approximate description parameter feasible set is adopted for the first time, an algorithm for calculating the vector sum of two ellipsoids is given in the process of prediction updating, an outer-wrapped ellipsoid containing all regions of the two ellipsoids can be obtained, in the process of measurement updating, Schweppe gives an outer-wrapped ellipsoid calculation method for calculating the intersection of the two ellipsoids, the two calculation rules determine the most fundamental theoretical basis of the ellipsoid collective-member filtering algorithm, but the optimization calculation rule of the ellipsoid is not given, then Chernusko provides a conditional formula for predicting the vector sum of the update and calculating the minimum volume of the outer-wrapped ellipsoid, and Maksarov and Norton provide an optimization calculation rule based on the volume of the minimized ellipsoid and the half-axis long square sum of the ellipsoid. The box algorithm is proposed by Milanese and Belforte, the core algorithm of the box algorithm is to determine the outer delimitation of the box, the problem can be converted into a linear programming problem in a linear system, the polyhedron algorithm aims at solving the problem of incompatible elements existing in an ellipsoid and box algorithm, a feasible set of parameters can be accurately described, but the calculation load is large, the calculation amount of the hyper-parallel body algorithm is small, and high estimation precision can be achieved; the fully symmetric multi-cell algorithm can accurately and approximately represent the time-varying characteristics of system parameters. In the ellipsoid set membership filtering algorithm, Fogel and Huang propose a minimum ellipsoid volume rule and a minimum ellipsoid trajectory rule.
Disclosure of Invention
Aiming at the technical problem of computational instability of the existing central difference filtering algorithm, the invention designs the maximum co-entropy cost function of bounded noise items of system noise prediction errors and observation noise errors aiming at the theoretical thought of the information entropy in the machine learning theory on the basis of the assumption that the noise in the ellipsoid collective filtering algorithm is unknown but bounded, thereby constructing a maximum co-entropy extended ellipsoid collective filtering method, applying the maximum co-entropy extended ellipsoid collective filtering method to the calculation of state parameters of a state space model of an extended robot system in a pose model of an autonomous mobile robot, and achieving the purpose of improving the computational stability of the nonlinear collective filtering algorithm.
The technical scheme of the invention is realized as follows:
a maximum covariance extended ellipsoid collective filtering method comprises the following steps:
the method comprises the following steps that firstly, a nonlinear discrete system state space model of the ground-based autonomous mobile robot is constructed, system state variables of a nonlinear discrete system are initialized, and an ellipsoid set of the system state variables is given; the nonlinear discrete system state space model comprises a dynamic motion model equation and a discretization observation model equation;
step two, obtaining the estimated value of the system state variable at the k-1 moment according to the initialization result of the nonlinear discrete system state space modelAnd a system state variable ellipsoidal shape matrix Pk-1And for the system state variable ellipsoid shape matrix Pk-1Performing Cholesky decomposition operation;
step three, adopting Stirling interpolation polynomial to estimate the system state variable at the k-1 th momentCarrying out linear approximation calculation, and taking the linear second-order polynomial as a Lagrange remainder formula of the nonlinear discrete system;
step four, utilizing an ellipsoid to outsource Lagrange remainder sub-type of the nonlinear discrete system to obtain a linearization error, and calculating an outsourcing ellipsoid of the linearization error;
adding the linearization error and the process noise of the nonlinear discrete system to obtain a virtual process noise error, calculating a virtual process noise error ellipsoid, and performing Cholesky decomposition operation on the virtual process noise error ellipsoid;
step six, calculating the state parameter ellipsoid boundary of the system state variable and the predicted value of the system state variable by utilizing a linear ellipsoid collective filtering algorithmAn ellipsoid-shaped matrix of the system state variable is predicted according to the virtual process noise error ellipsoid;
seventhly, estimating the system state variable value based on the discretization observation model equation and the k-1 momentPredicting and updating the state variables of the nonlinear dispersion system to obtain a predicted value of the observation vector and a predicted ellipsoidal shape matrix, and calculating a predicted covariance matrix according to the square root of the predicted ellipsoidal shape matrix and the square root of the ellipsoidal shape matrix in the step six;
step eight, predicting values according to the system state variables in the step sixCalculating the prediction error of the state variable, and predicting the system state variable in the sixth stepCombining the system error expansion equation with an observation equation to obtain a system error expansion equation; obtaining a conversion expansion state model according to a variance matrix constructed by utilizing an expansion noise item of a system error expansion equation;
step nine, constructing a cost function by using an MCC (Motor control center) criterion according to the conversion expansion state model in the step eight, and obtaining the optimal estimation value of the system state variable at the kth moment by calculating the optimal solution of the cost function
And step ten, updating the state parameter ellipsoid boundary of the system state variable by using a linear ellipsoid collective filtering algorithm, and calculating a predicted value and a predicted variance matrix of the system state vector at the kth moment according to the state parameter ellipsoid boundary.
The nonlinear discrete system state space model is as follows:
wherein x isk∈RnState variable, x, representing time kk-1Representing the state variable at time k-1, f (-) and h (-) each representing a non-linear second order derivative function, qk-1Representing the process noise term at time k-1, rk∈RmRepresenting the observed noise term, y, over timekRepresenting an observation vector;
the ellipsoid set of the system state variables is:
E(a,P)={x∈Rn|(x-a)TP-1(x-a)≤1},
wherein a represents the center of the ellipsoid set, P is a positive qualitative ellipsoid envelope matrix, and the system initial state estimation ellipsoid set is
The system state variable estimation value at the k-1 time is subjected to a Stirling interpolation polynomialThe expression for the calculation of the linearized approximation is:
wherein x iskState variable representing time k, DΔxThe term is called a difference operator;
wherein, Δ xpRepresenting the estimated deviation, μ, of the decoupled state variable of the systempAs deviation operator, δpIs an averageThe operator(s) is (are) selected,representing the estimation deviation of the system state variable at the k-1 moment, wherein s is an interpolation step length;
said deviation operator mupComprises the following steps:
the average operator deltapComprises the following steps:
the Lagrange remainder formula is:
wherein the content of the first and second substances,indicating Lagrange remainder terms centered on the state estimate at time k-1.
The method for utilizing the ellipsoid to outsource the Lagrange remainder formula of the nonlinear discrete system to obtain the linearization error comprises the following steps:
wherein the content of the first and second substances,represents the diagonal elements of the linearized error shape matrix determined by Lagrange's remainder terms at time k-1,representing the non-diagonal elements of the linearization error shape matrix determined by Lagrange remainder items at the k-1 moment, wherein i and j represent the ith row and the jth column of the linearization error shape matrix, the value is i more than or equal to 1, j more than or equal to n, and n is the dimension of the system state vector;
the outer ellipsoid of the linearization error is:wherein the content of the first and second substances,representing a linearized error shape matrix determined by Lagrange's remainder terms.
The virtual process noise error is:
wherein the content of the first and second substances,representing the virtual process noise at time k-1,representing a virtual process noise variance matrix, Qk-1A process noise matrix is represented that represents the process noise matrix,a set of add operations is represented and,representing a process noise error optimization factor;
the virtual process noise error ellipsoid is:wherein the content of the first and second substances,representing a virtual process noise at time k;
performing Cholesky decomposition operation on the virtual process noise error ellipsoid to obtain:wherein the content of the first and second substances,representing the virtual process noise variance matrix square root.
the ellipsoid-shaped matrix of the system state variables is:
wherein the content of the first and second substances,a matrix of predicted shapes representing the state variables of the system,is represented by P'k,k-1The square root of (a);
prediction value of the observation vector
Wherein the content of the first and second substances,representing a system observation vector predicted value;
prediction covariance matrix Pxy,k,k-1Comprises the following steps:
wherein the content of the first and second substances,expressing a system observation equation prediction error matrix;
the state variable prediction error is:
wherein the content of the first and second substances,representing a state variable prediction error;
the system error expansion equation is:
wherein the content of the first and second substances,first order difference operator matrix representing a non-linear observation equation, BkA matrix of first order difference operators, R, representing the k-th momentkRepresenting the observation noise term, I representing the identity matrix,is the spreading noise term;
the variance matrix constructed by using the extended noise term of the system error extension equation is as follows:
wherein, thetakRepresenting the spread noise variance matrix, SkRepresenting the spread noise variance matrix ΘkCholesky decomposition matrix of (S)r,kRepresents the square root of the observed noise error matrix,representing a spread noise square root matrix;
the transition extended state model is:
Dk=Wkxk+ek,
wherein the content of the first and second substances,indicating the expanded state at the time of the k-th instant,represents the extended state transition matrix at time k,and is
The cost function is:
wherein d isi,kIs a matrix DkThe ith row vector ofi,kIs a matrix WkL ═ n + m denotes the matrix DkDimension of, JL(xk) Representing the MCC cost function, GσRepresenting an error cost function;
the optimal estimated value of the system state variable at the k-th momentComprises the following steps:
the cost function is subjected to partial derivation and sorting to obtain:
wherein, Cx,k=diag(Gσ(d1,k-w1,kxk),…,Gσ(dn,k-wn,kxk)),
Cy,k=diag(Gσ(dn+1,k-wn+1,kxk),…,Gσ(dn+m,k-wn+m,kxk))。
The state parameter ellipsoid boundary for updating the system state variable by using the linear ellipsoid collective filtering algorithm is as follows:wherein, PkRepresenting the system state variable estimation error matrix at time k, EkAn ellipsoid representing the updated system state variables;
the predicted value of the system state vector at the kth moment calculated according to the state parameter ellipsoid boundary is as follows:
wherein the content of the first and second substances,representing an ellipsoid collector filter gain matrix;
the prediction variance matrix is:
wherein the content of the first and second substances, ρk∈(0,1),is a first order difference operator matrix of the observation equation.
The beneficial effect that this technical scheme can produce: the invention introduces a novel maximum co-entropy MCC criterion in the observation updating step on the basis of the traditional ellipsoid collective filtering algorithm based on the minimum mean square error criterion, obtains a novel system noise expression by carrying out system model expansion operation by combining the nonlinear system prediction noise error and the observation noise, and constructs an error cost function expressed by a second-order information potential energy formula according to the maximum co-entropy criterion based on the system state variable prediction vector and the observation vector, thereby designing the observation updating calculation process of the ellipsoid collective filtering algorithm and constructing a novel ellipsoid collective filtering algorithm calculation frame based on the maximum co-entropy; by utilizing the method disclosed by the invention to carry out the pose calculation simulation verification of the land-based robot, the calculation precision of the method disclosed by the invention is improved, and the calculation stability is obviously improved and enhanced compared with that of the traditional ellipsoid collective filtering algorithm.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
FIG. 1 is a computational flow diagram of the present invention.
FIG. 2 is a schematic diagram of a mobile robot motion model in the method of the present invention.
FIG. 3 is a diagram of the error data calculated by the mobile robot for MCC-SMF according to the present invention.
FIG. 4 is a diagram of the mobile robot carrier trajectory calculation data of the MCC-SMF of the present invention.
Fig. 5 is a graph of the calculated error data of the mobile robot carrier obtained by the EKF algorithm.
Fig. 6 is a graph of the mobile robot carrier trajectory calculation data obtained by the EKF algorithm.
Fig. 7 is a diagram of the mobile robot carrier calculation error data obtained by the SUKF algorithm.
Fig. 8 is a diagram of the mobile robot carrier trajectory calculation data obtained by the SUKF algorithm.
Detailed Description
The technical solutions in the embodiments of the present invention will be described clearly and completely with reference to the accompanying drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without any inventive step, are within the scope of the present invention.
Preliminary knowledge
The co-entropy criterion is a local similarity measurement concept which defines a similarity measurement between two random vectors in a kernel space, and has the advantage of insensitivity to large errors, so that the computation efficiency of the system state variable is not influenced by impulse noise or abnormal data processing. For two random variables X ∈ Rn,Y∈RnWith a joint probability density function of pX,Y(x, y), then the coventropy of two random variables is defined as,
kappa (X, Y) in the definition of coventropy is a radial kernel function, E [ kappa (X, Y) ] represents the desired operator, the radial kernel function is generally represented by a Gaussian symmetric kernel function,
wherein x-y-e represents an error vector of two variable vectors, and σ represents a bandwidth value σ > 0 of the kernel function; for an asymmetric gaussian kernel distribution, however, its kernel function can be expressed as,
σ + and σ -as defined herein represent the "+" and "-" portions of the error vector, respectively, then asymmetric coventropy may be defined as,
for limited data setsA simple sample mean estimator can be used to approximate the coventropy expression of a continuous gaussian kernel as,
further obtains the Taylor series expansion expression as,
then, if the state parameter x and the output variable y of the state space model are considered, the optimal estimated target value of the system state variable obtained by performing iterative recursive optimal filtering calculation based on the maximum covariance criterion can be expressed as,
then, a cost function can be constructed by utilizing a maximum co-entropy criterion to carry out the design work of the nonlinear system state space model state parameter filtering algorithm.
Aiming at the computational instability of a central differential filtering algorithm, the filtering computation criterion is revised from Minimum Mean Square Error (MMSE) to maximum negotiation entropy criterion MCC, so that a novel maximum negotiation entropy ellipsoid collective filtering algorithm is designed and obtained, and is applied to an autonomous mobile robot pose model to develop the state parameter computation of a robot system state space model.
The invention discloses a method for constructing a computing frame of an ellipsoid collective filtering algorithm based on maximum entropy, which is based on a traditional ellipsoid collective filtering algorithm based on a minimum mean square error criterion, faces to a nonlinear system state space model, reserves the computing process of a time updating prediction step of the traditional ellipsoid collective filtering algorithm, introduces a novel maximum entropy negotiation MCC criterion in an observation updating step, obtains a novel system noise expression by carrying out system model expansion operation by combining nonlinear system prediction noise error and observation noise, constructs an error cost function expressed by a second-order information potential energy formula according to the maximum entropy criterion based on a system state variable prediction vector and an observation vector, and designs the observation updating computing process of the ellipsoid collective filtering algorithm based on the maximum entropy. The algorithm is used for carrying out the pose calculation simulation verification of the ground-based robot, the calculation precision of the method is improved, and the calculation stability is obviously improved compared with that of the traditional ellipsoid collective filtering algorithm.
The maximum entropy criterion is a method for calculating the maximum value of error information based on a Renyis entropy formula, an information potential energy concept in a machine learning theory is introduced, a second-order information potential energy expression is abstracted to describe the prediction error of a nonlinear system state variable in an ellipsoid collective filtering algorithm, a nonlinear system extended error state model is obtained together with an observation equation of a nonlinear system state space model, the noise variance of the extended error state model is calculated, Cholesky decomposition is carried out on the noise variance model to obtain a square root variance positive definite matrix of extended noise, the extended error state model is transformed and sorted by using the square root noise variance matrix to obtain a prediction error and observation noise combined error expression, then a prediction error cost function of the ellipsoid collective filtering algorithm is synthesized by using a second-order information expression of the maximum entropy criterion, then the optimal calculation of the system state variable is obtained by using the minimum maximum entropy negotiation cost function, calculating partial differential of a minimized error entropy cost function and setting the partial differential to be 0 to obtain the minimum error entropy cost function, simultaneously realizing an explicit expression of system state variable optimization calculation by utilizing a matrix inverse theorem, setting a small parameter in computer algorithm compilation to judge a system state variable estimation value of each iterative calculation, and if the judgment expression is true, continuing to calculate a system state variable estimation variance matrix; otherwise, the calculation process of the previous step is continued, and finally the calculation of the system state variable estimation variance matrix is obtained.
As shown in fig. 1, an embodiment of the present invention provides a maximum covariance extended ellipsoid membership filtering method, which includes the following steps:
the method comprises the following steps that firstly, a nonlinear discrete system state space model of the ground-based autonomous mobile robot is constructed, system state variables of a nonlinear discrete system are initialized, and an ellipsoid set of the system state variables is given; the nonlinear discrete system state space model comprises a dynamic motion model equation and a discretization observation model equation;
the nonlinear discrete system state space model is as follows:
wherein x isk∈RnState variable, x, representing time kk-1Representing the state variable at time k-1, f (-) and h (-) each representing a non-linear second order derivative function, qk-1Representing the process noise term at time k-1, rk∈RmRepresenting the observed noise term, y, over timekRepresenting an observation vector; q. q.sk-1∈RnAnd rk∈RmRepresenting process noise and observation noise over time, and satisfying Unknown But Bounded (UBB) assumptions,note qk∈(0,Qk) And rk∈(0,Rk). Initial state x of the system0Belonging to a known bounded set x0∈X0The set may be determined from a priori knowledge of the system state, while for the measurement sequence vectorThen the state feasible set of the ellipsoid membership filtering algorithm at time k is defined as XkIt consists of all possible state points that are consistent with all available information, including the system model, noise assumptions, and initial state set.
The ellipsoid set of the system state variables is:
E(a,P)={x∈Rn|(x-a)TP-1(x-a)≤1} (9),
wherein a represents the center of the ellipsoid set, P is a positive qualitative ellipsoid envelope matrix, and the system initial state estimation ellipsoid set isThe system state ellipsoid set obtained by k-1 time estimation isThe iterative process of the non-linear ellipsoid membership filtering algorithm at the time k consists of steps two to eight.
Step two, obtaining the estimated value of the system state variable at the k-1 moment according to the initialization result of the nonlinear discrete system state space modelAnd a system state variable ellipsoidal shape matrix Pk-1And for the system state variable ellipsoid shape matrix Pk-1A Cholesky decomposition operation is performed,the uncertainty interval of the state component at time k-1 is set to,
Thirdly, in the process of designing the nonlinear system state space model, carrying out linear approximation calculation on the nonlinear system function of the moving target, wherein Stirling interpolation polynomial is adopted to carry out linear approximation calculation on the nonlinear function, and the Stirling interpolation polynomial is adopted to carry out linear approximation calculation on the estimated value of the system state variable at the k-1 momentPerforming linear approximation calculation, and taking the linearized second-order polynomial as a Lagrange remainder formula of the nonlinear discrete system;
the system state variable estimation value at the k-1 time is subjected to a Stirling interpolation polynomialThe expression for the calculation of the linearized approximation is:
wherein x iskState variable representing time k, DΔxThe term is called a difference operator;
wherein, Δ xpRepresenting the estimated deviation, μ, of the decoupled state variable of the systempAs deviation operator, δpIn order to average the operators, the operator is,representing the estimation deviation of the system state variable at the k-1 moment, wherein s is an interpolation step length;
said deviation operator mupComprises the following steps:
the average operator deltapComprises the following steps:
as can be seen from the approximation expression of the Stirling interpolation polynomial, the computation accuracy of the Stirling interpolation expansion is higher than that of the Taylor series expansion, and the accuracy can be controlled by the interpolation step length s. Taking the first two terms of the Stirling interpolation polynomial as linearized approximations of the nonlinear system process function, wherein the Lagrange remainder formula is as follows:
wherein the content of the first and second substances,indicating Lagrange remainder terms centered on the state estimate at time k-1.
Step four, utilizing an ellipsoid to outsource Lagrange remainder sub-type of the nonlinear discrete system to obtain a linearization error, and calculating an outsourcing ellipsoid of the linearization error;
the method for utilizing the ellipsoid to outsource the Lagrange remainder formula of the nonlinear discrete system to obtain the linearization error comprises the following steps:
wherein the content of the first and second substances,represents the diagonal elements of the linearized error shape matrix determined by Lagrange's remainder terms at time k-1,representing the non-diagonal elements of the linearization error shape matrix determined by Lagrange remainder items at the k-1 time, i and j representing the ith row and the jth column of the linearization error shape matrix, and the value of the i and the jth columns is more than or equal to 1i, j is less than or equal to n, and n is the dimension of the system state vector;
the outer ellipsoid of the linearization error is:whereinRepresenting a linearized error shape matrix determined by Lagrange's remainder terms. .
Adding the linearization error and the process noise of the nonlinear discrete system to obtain a virtual process noise error, calculating a virtual process noise error ellipsoid, and performing Cholesky decomposition operation on the virtual process noise error ellipsoid;
the virtual process noise error is:
wherein the content of the first and second substances,representing the virtual process noise at time k-1,representing a virtual process noise variance matrix, Qk-1A process noise matrix is represented that represents the process noise matrix,a set of add operations is represented and,representing a process noise error optimization factor;
the virtual process noise error ellipsoid is:wherein the content of the first and second substances,representing a virtual process noise at time k;
performing Cholesky decomposition operation on the virtual process noise error ellipsoid to obtain:wherein the content of the first and second substances,representing the virtual process noise variance matrix square root.
Step six, calculating the state parameter ellipsoid boundary of the system state variable and the predicted value of the system state variable by utilizing a linear ellipsoid collective filtering algorithmAn ellipsoid-shaped matrix of the system state variable is predicted according to the virtual process noise error ellipsoid; calculating the boundary of the prediction state parameter ellipsoid by using the prediction step of the linear ellipsoid collective filtering algorithm, which is a linearized prediction ellipsoidAnd a virtual process noise direct sum calculation process;
the ellipsoid-shaped matrix of the system state variables is:
wherein the content of the first and second substances,a matrix of predicted shapes representing the state variables of the system,is represented by P'k,k-1The square root of (a);
seventhly, estimating the system state variable value based on the discretization observation model equation and the k-1 momentPredicting and updating the state variables of the nonlinear dispersion system to obtain a predicted value of the observation vector and a predicted ellipsoidal shape matrix, and calculating a predicted covariance matrix according to the square root of the predicted ellipsoidal shape matrix and the square root of the ellipsoidal shape matrix in the step six;
on the basis of predicting and updating the state variables of the nonlinear system, one-step prediction is carried out on the observation vector, and the prediction value of the obtained observation vector is as follows:
wherein the content of the first and second substances,representing a system observation vector predicted value;
prediction covariance matrix Pxy,k,k-1Comprises the following steps:
wherein the content of the first and second substances,expressing a system observation equation prediction error matrix;
step eight, predicting values according to the system state variables in the step sixCalculating the prediction error of the state variable, and predicting the system state variable in the sixth stepCombining the system error expansion equation with an observation equation to obtain a system error expansion equation; obtaining a conversion expansion state model according to a variance matrix constructed by utilizing an expansion noise item of a system error expansion equation;
the state variable prediction error is:
wherein the content of the first and second substances,representing a state variable prediction error;
the system error expansion equation is:
wherein the content of the first and second substances,first order difference operator matrix representing a non-linear observation equation, BkA matrix of first order difference operators, R, representing the k-th momentkRepresenting the observation noise term, I representing the identity matrix,is the spreading noise term;
the variance matrix constructed by using the extended noise term of the system error extension equation is as follows:
wherein, thetakRepresenting the spread noise variance matrix, SkRepresenting the spread noise variance matrix ΘkCholesky decomposition matrix of (S)r,kRepresents the square root of the observed noise error matrix,representing a spread noise square root matrix; (ii) a
The transition extended state model is:
Dk=Wkxk+ek (30),
wherein the content of the first and second substances,indicating the expanded state at the time of the k-th instant,represents the extended state transition matrix at time k,and is
Step nine, constructing a cost function by using an MCC (Motor control center) criterion according to the conversion expansion state model in the step eight, and obtaining the optimal estimation value of the system state variable at the kth moment by calculating the optimal solution of the cost function
The cost function is:
wherein d isi,kIs a matrix DkThe ith row vector ofi,kIs a matrix WkL ═ n + m denotes the matrix DkDimension of, JL(xk) Representing the MCC cost function, GσRepresenting an error cost function; (ii) a
Optimal estimation of system state variablesIs the optimal solution of equation (31), and therefore, the optimal estimated value of the system state variable at the k-th timeComprises the following steps:
further finishing to obtain:
wherein, Cx,k=diag(Gσ(d1,k-w1,kxk),…,Gσ(dn,k-wn,kxk)),
Cy,k=diag(Gσ(dn+1,k-wn+1,kxk),…,Gσ(dn+m,k-wn+m,kxk))。
Step ten, updating the state parameter ellipsoid boundary of the system state variable by utilizing a linear ellipsoid collective filtering algorithmWherein, PkRepresenting the system state variable estimation error matrix at time k, EkAn ellipsoid representing an updated system state variable; the essence is to predict the state ellipsoidAnd observation setThe calculation of the straight intersection and the intersection is carried out,
calculating the predicted value and the prediction variance matrix of the system state vector at the k moment according to the ellipsoid boundary of the state parameters,
the prediction variance matrix is:
wherein the content of the first and second substances, is a first order difference operator matrix of the observation equation.
The method has the advantages that the method adopts the Stirling interpolation polynomial to implement the linearization operation, effectively avoids the complex calculation of the first-order Jacobian matrix and the second-order Hessian matrix of the Taylor series expansion, and reduces the calculation complexity of the algorithm; the calculation precision can be controlled by utilizing the interpolation step length s; compared with the traditional nonlinear ensemble filtering algorithm based on Taylor series expansion, the method provided by the invention has higher calculation precision.
In addition, the algorithm introduces four parameters, interpolation step length s and three scale factor parametersβk-1And ρkThe numerical determination method is as follows:
for the interpolation step length s parameter, in general, if the system state vector satisfies the Gauss distribution,to satisfy this condition, the estimated variance matrix of the system state vector computed at each iteration implements a cholesky decomposition, P ═ SSTTherefore, decoupling transformation operation is carried out on the system state vector to enable the system state vector to meet Gauss distribution conditions.
Scale factor parameterAnd betak-1Involving the outsourcing of two ellipsoidal direct sumsThe method has simple solving form, and has stronger performance robustness compared with the optimization criterion of minimizing the volume of the outer-wrapped ellipsoid. Namely haveThereby can adopt the formulaObtaining optimal scale factor parametersAnd betak-1。
Scale factor parameterRequires E (0, Q)k-1) Andand (3) calculating the direct sum, and then calculating the formula of the calculation criterion as follows:
For the scale factor parameter betak-1Two ellipsoids are requiredAndthe calculation formula of the variance matrix under the condition of considering the updating of the observation vector is as follows:
thereby obtaining the scale factor parameter betak-1Is calculated by the formula
In the iterative calculation process, an observation set SyThe form is generally complex, resulting in a system state vector variance matrix PkWhether using the minimized ellipsoid volume method or the minimized ellipsoid trajectory criterion, the calculation complexity of (2) makes the scale factor parameter ρkThe optimization calculation is difficult, even an analytic solution cannot be obtained, and the calculation complexity is high if a numerical calculation method is adopted. In the present invention, the minimum performance index delta is adoptedkIs calculated in upper bound form
Thus, the scale factor parameter ρ can be obtainedkA suboptimum calculation formula
Wherein p ismIs a matrixMaximum singular value of cmIs thatThe maximum singular value of the matrix.
Examples of the applications
In order to verify the calculation efficiency of the minimum error entropy center differential filtering algorithm provided by the invention, the method is utilized to carry out simulation verification calculation on a land-based robot positioning system model to prove the effectiveness and the calculation advantages of the method, and simulation verification test data are provided. Here consider a groundA surface mobile robot system adopts a front wheel driving mode, as shown in figure 2, the steering angle of a front wheel is defined as alpha, the anticlockwise direction is defined as the positive direction, and a robot coordinate system xy is adoptedRobotRelative to the ground coordinate system xygroundWith steering angle psi, counter-clockwise positive and rear wheel speed defined as vrearwheel. The distance R between an observation point and the center of the rear wheel and the wheel base L of the front wheel and the rear wheel satisfy
So that the observation point distance can be obtained,simultaneously, the expression of the rear wheel speed can be obtained as follows:the steering angle equation of the robot can be obtained by arranging,
is finished to obtain
If the movement speed of the mobile robot system is considered, the dynamic equation of the mobile robot system can be obtained in the robot coordinate system as
The dynamic equation of the robot system is converted into a ground coordinate system, and the final mobile equation of the mobile robot system in the ground coordinate system can be obtained
The parameter alpha (t) in the system equation can be used as a modulation parameter and used as an input variable of the system to generate a control parameter of the front wheel, which meets the control law
The parameter psi heredesRepresenting the desired heading angle and the parameter ψ represents the current moving heading angle, the desired heading angle can be expressed as
Generally, the steering angle ranges from (+/-pi/4), and the G parameter influences the speed of the steering of the robot. For the purpose of carrying out system state variable estimation and motion path tracking on a motion equation of a mobile robot system, a plurality of sensors are used for sensing the coordinate position of the system, such as a GPS or a forward direction coding direction finder and other equipment for completing target tracking observation, and the two-direction position coordinates of the mobile robot in a ground coordinate system are selected as observation variables, so that the observation equation is linear and can be directly obtained. Consider a system initial variance matrix of
In addition, system parameters are set as that the wheel base L of the front wheel and the rear wheel of the mobile robot is 2m, the control law gain G is 2, the moving speed is kept as v is 1m/s, and the sampling time interval delta t is 0.1 s; assuming that only the heading angle psi has interference noise in the process variable, the process noise variance can be set to Q0.052. The observed variable is the mobile robot position coordinates, so the observed noise variance matrix can be set to
Therefore, the moving track simulation result of the mobile robot system can be obtained, and the results of the method, the EKF algorithm and the SUKF algorithm are shown in FIGS. 3-8. Comparing the MCC-SMF algorithm with the SUKF algorithm, the invention has the advantages of better calculation stability of the MCC-SMF algorithm, high calculation convergence speed and obviously improved and improved calculation precision.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, but rather as the subject matter of the invention is to be construed in all aspects and equivalents thereof.
Claims (8)
1. A maximum covariance extended ellipsoid collective filtering method is characterized by comprising the following steps:
the method comprises the following steps that firstly, a nonlinear discrete system state space model of the ground-based autonomous mobile robot is constructed, system state variables of a nonlinear discrete system are initialized, and an ellipsoid set of the system state variables is given; the nonlinear discrete system state space model comprises a dynamic motion model equation and a discretization observation model equation;
step two, obtaining the estimated value of the system state variable at the k-1 moment according to the initialization result of the nonlinear discrete system state space modelAnd a system state variable ellipsoidal shape matrix Pk-1And for the system state variable ellipsoid shape matrix Pk-1Performing Cholesky decomposition operation;
step three, adopting Stirling interpolation polynomial to estimate the system state variable at the k-1 th momentPerforming linear approximation calculation, and performing linear approximationThe second-order polynomial is used as Lagrange remainder of the nonlinear discrete system;
step four, utilizing an ellipsoid to outsource Lagrange remainder sub-type of the nonlinear discrete system to obtain a linearization error, and calculating an outsourcing ellipsoid of the linearization error;
adding the linearization error and the process noise of the nonlinear discrete system to obtain a virtual process noise error, calculating a virtual process noise error ellipsoid, and performing Cholesky decomposition operation on the virtual process noise error ellipsoid;
step six, calculating the state parameter ellipsoid boundary of the system state variable and the predicted value of the system state variable by utilizing a linear ellipsoid collective filtering algorithmAn ellipsoid-shaped matrix of the system state variable is predicted according to the virtual process noise error ellipsoid;
seventhly, estimating the system state variable value based on the discretization observation model equation and the k-1 momentPredicting and updating the state variables of the nonlinear discrete system to obtain a predicted value of the observation vector and a predicted ellipsoidal matrix, and calculating a predicted covariance matrix according to the square root of the predicted ellipsoidal matrix and the square root of the ellipsoidal matrix in the step six;
step eight, predicting values according to the system state variables in the step sixCalculating the prediction error of the state variable, and predicting the system state variable in the sixth stepCombining the system error expansion equation with an observation equation to obtain a system error expansion equation; obtaining a conversion expansion state model according to a variance matrix constructed by utilizing an expansion noise item of a system error expansion equation;
step nine, according to the conversion expansion state model in the step eight, constructing a cost function by using the maximum covariance MCC criterion, and obtaining the optimal estimation value of the system state variable at the kth moment by calculating the optimal solution of the cost function
The cost function is:
wherein d isi,kIs a matrix DkThe ith row vector ofi,kIs a matrix WkL ═ n + m denotes the matrix DkThe dimension(s) of (a) is,representing the maximum co-entropy MCC cost function, GσRepresenting an error cost function;
the optimal predicted value x of the system state variable at the kth momentk,k-1Comprises the following steps:
the cost function is subjected to partial derivation and sorting to obtain:
step ten, updating the state parameter ellipsoid boundary of the system state variable by using a linear ellipsoid collective filtering algorithm, and calculating a predicted value and a predicted variance matrix of the system state vector at the kth moment according to the state parameter ellipsoid boundary;
the state parameter ellipsoid boundary for updating the system state variable by using the linear ellipsoid collective filtering algorithm is as follows:
wherein, PkRepresenting the system state variable estimation error matrix at time k, EkAn ellipsoid representing the updated system state variables;
the estimation value of the system state vector at the kth moment calculated according to the state parameter ellipsoid boundary is as follows:
wherein the content of the first and second substances,representing an ellipsoidal collector-driver filter gain matrix, Pk,k-1Ellipsoidal-shaped matrix, y, representing system state variableskRepresenting an observation vector, h (-) representing a non-linear second order derivative function, xk,k-1The optimal predicted value of the system state variable at the kth moment is represented;
the system state variable estimation error matrix at the k-th moment is as follows:
2. The maximum covariance extended ellipsoid membership filtering method of claim 1, wherein the nonlinear discrete system state space model is:
wherein x isk∈RnState variable, x, representing time kk-1Representing the state variable at time k-1, f (-) and h (-) each representing a non-linear second order derivative function, qk-1Representing the process noise term at time k-1, rk∈RmRepresenting the observed noise term, y, over timekRepresenting an observation vector;
the ellipsoid set of the system state variables is:
E(a,P)={x∈Rn|(x-a)TP-1(x-a)≤1},
3. The maximum covariance spreading ellipsoid collective filtering method of claim 2, wherein the system state variable estimation value at the k-1 th time is obtained by using Stirling interpolation polynomialThe expression for the calculation of the linearized approximation is:
wherein x iskState variable representing time k, D△xReferred to as difference operator;
wherein, Δ xp'Representing the estimated deviation, μ, of the decoupled state variable of the systemp'As deviation operator, δp'In order to average the operators, the operator is,representing the estimation deviation of the system state variable at the k-1 moment, wherein s is an interpolation step length;
said deviation operator mup'Comprises the following steps:
the average operator deltap'Comprises the following steps:
the Lagrange remainder formula is:
4. The maximum covariance spreading ellipsoid membership filtering method according to claim 3, wherein the Lagrange remainder sub-formula of the nonlinear discrete system is outsourced by using an ellipsoid to obtain a linearization error as follows:
wherein the content of the first and second substances,represents the diagonal elements of the linearized error shape matrix determined by Lagrange's remainder terms at time k-1,representing the non-diagonal elements of the linearization error shape matrix determined by Lagrange remainder items at the k-1 moment, wherein i and j represent the ith row and the jth column of the linearization error shape matrix, the value is i more than or equal to 1, j more than or equal to n, and n is the dimension of the system state vector;
5. The maximum covariance spreading ellipsoid membership filtering method of claim 4, wherein the virtual process noise error is:
wherein the content of the first and second substances,representing the virtual process noise at time k-1,representing a virtual process noise variance matrix, Qk-1A process noise matrix is represented that represents the process noise matrix,represents a set addition operation, βQk-1E (0,1) represents a process noise error optimization factor;
the virtual process noise error ellipsoid is:wherein the content of the first and second substances,representing virtual process noise at time k;
6. The maximum covariance spreading ellipsoid membership filtering method of claim 5, wherein the system state variable prediction valueComprises the following steps:
the ellipsoid-shaped matrix of the system state variables is:
wherein the content of the first and second substances,a matrix of predicted shapes representing the state variables of the system,represents P'k,k-1The square root of (a);
7. the maximum co-entropy extended ellipsoid membership filtering method of claim 6, wherein a predictor of the observation vector
Wherein the content of the first and second substances,representing a system observation vector predicted value;
prediction covariance matrix Pxy,k,k-1Comprises the following steps:
wherein the content of the first and second substances,expressing a system observation equation prediction error matrix;
8. the maximum covariance spreading ellipsoid membership filtering method of claim 7, wherein the state variable prediction error is:
wherein the content of the first and second substances,representing a state variable prediction error;
the system error expansion equation is:
wherein the content of the first and second substances,first order difference operator matrix representing a non-linear observation equation, BkA matrix of first order difference operators, R, representing the k-th momentkRepresenting the observation noise term, I representing the identity matrix,is the spreading noise term;
the variance matrix constructed by using the extended noise term of the system error extension equation is as follows:
wherein, thetakRepresenting the spread noise variance matrix, SkRepresenting the spread noise variance matrix ΘkCholesky decomposition matrix of (S)r,kRepresents the square root of the observed noise error matrix,representing a spread noise square root matrix;
the transition extended state model is:
Dk=Wkxk+ek,
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010893973.8A CN111983926B (en) | 2020-08-31 | 2020-08-31 | Maximum co-entropy extended ellipsoid collective filtering method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010893973.8A CN111983926B (en) | 2020-08-31 | 2020-08-31 | Maximum co-entropy extended ellipsoid collective filtering method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111983926A CN111983926A (en) | 2020-11-24 |
CN111983926B true CN111983926B (en) | 2022-04-12 |
Family
ID=73441363
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010893973.8A Active CN111983926B (en) | 2020-08-31 | 2020-08-31 | Maximum co-entropy extended ellipsoid collective filtering method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111983926B (en) |
Families Citing this family (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112683265B (en) * | 2021-01-20 | 2023-03-24 | 中国人民解放军火箭军工程大学 | MIMU/GPS integrated navigation method based on rapid ISS collective filtering |
CN112883508B (en) * | 2021-01-22 | 2024-03-08 | 江南大学 | Parallel spatial filtering-based spring damping system state estimation method |
CN113110548B (en) * | 2021-04-21 | 2023-05-12 | 北京控制工程研究所 | Design method for spacecraft ellipsoid set evolution |
CN113950018B (en) * | 2021-10-13 | 2023-10-27 | 华东理工大学 | Asynchronous multi-sensor network system and global ellipsoid state estimation method |
CN114353805B (en) * | 2022-01-13 | 2022-10-21 | 中国人民解放军国防科技大学 | Robot cluster navigation method, device and equipment based on manual assistance |
CN116518983B (en) * | 2023-07-05 | 2023-10-17 | 西安羚控电子科技有限公司 | Self-adaptive fusion method and device for mobile robot positioning |
Family Cites Families (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2016176684A1 (en) * | 2015-04-30 | 2016-11-03 | The Regents Of The University Of California | Entropy field decomposition for image analysis |
CN105222780B (en) * | 2015-09-07 | 2016-08-24 | 郑州轻工业学院 | A kind of ellipsoid set-membership filtering method approached based on Stirling interpolation polynomial |
CN105740947B (en) * | 2016-02-01 | 2018-01-12 | 金陵科技学院 | Robot performance's optimization method based on Pareto entropy particle cluster algorithms |
CN106767780B (en) * | 2016-11-28 | 2017-10-17 | 郑州轻工业学院 | The extension ellipsoid set-membership filtering method approached based on Chebyshev polynomial interopolations |
CN108508463B (en) * | 2018-03-28 | 2020-02-18 | 郑州轻工业学院 | Fourier-Hermite orthogonal polynomial based extended ellipsoid collective filtering method |
CN109901153B (en) * | 2019-03-29 | 2022-11-01 | 西安电子科技大学 | Target track optimization method based on information entropy weight and nearest neighbor data association |
-
2020
- 2020-08-31 CN CN202010893973.8A patent/CN111983926B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN111983926A (en) | 2020-11-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111983926B (en) | Maximum co-entropy extended ellipsoid collective filtering method | |
CN109885883B (en) | Unmanned vehicle transverse motion control method based on GK clustering algorithm model prediction | |
CN111983927B (en) | Ellipsoid collective filtering method for maximum covariance MCC (MCC) criterion | |
CN111258218B (en) | Intelligent vehicle path tracking method based on maximum correlation entropy criterion | |
CN111783307B (en) | Hypersonic aircraft state estimation method | |
Xu et al. | Enhanced bioinspired backstepping control for a mobile robot with unscented Kalman filter | |
CN110095983B (en) | Mobile robot prediction tracking control method based on path parameterization | |
CN113296514B (en) | Local path optimization method and system based on sparse banded structure | |
CN111969979B (en) | Minimum error entropy CDKF filter method | |
Gu et al. | Trajectory planning and tracking control of a ground mobile robot: A reconstruction approach towards space vehicle | |
Qin et al. | An improved real-time slip model identification method for autonomous tracked vehicles using forward trajectory prediction compensation | |
Cossette et al. | The complex-step derivative approximation on matrix lie groups | |
Wu et al. | LiDAR based trajectory-tracking of an autonomous differential drive mobile robot using fuzzy sliding mode controller | |
Demim et al. | Visual SVSF-SLAM algorithm based on adaptive boundary layer width | |
Jiang et al. | Learning based predictive error estimation and compensator design for autonomous vehicle path tracking | |
CN107292445A (en) | A kind of space non-cooperative target pose Forecasting Methodology | |
Yu et al. | An adaptive unscented particle filter algorithm through relative entropy for mobile robot self-localization | |
CN108681621B (en) | RTS Kalman smoothing method based on Chebyshev orthogonal polynomial expansion | |
CN113670315B (en) | Variation iterative Kalman filtering-based method for estimating attitude of Liqun heavy tail interference noise dynamic aircraft | |
Quang et al. | Design a nonlinear MPC controller for autonomous mobile robot navigation system based on ROS | |
CN115933647B (en) | OMR track tracking control method based on composite control algorithm and storage medium | |
CN111998854A (en) | Cholesky decomposition calculation-based accurate expansion Stirling interpolation filtering method | |
Wang et al. | Ground moving target tracking with quadrotors using nonlinear model predictive control | |
Gao et al. | Improved 2D laser slam graph optimization based on Cholesky decomposition | |
Moore et al. | Utilising fuzzy models in the design of estimators and predictors: An agile target tracking example |
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 | ||
TR01 | Transfer of patent right | ||
TR01 | Transfer of patent right |
Effective date of registration: 20230413 Address after: No. 3a20, incubation Building 1, No. 11, Changchun Road, high tech Industrial Development Zone, Zhengzhou, Henan 450000 Patentee after: Zhengzhou ousma Intelligent Control Technology Co.,Ltd. Address before: 450002 No. 5 Dongfeng Road, Jinshui District, Henan, Zhengzhou Patentee before: Zhengzhou University of light industry |