Long cruise unmanned platform combined positioning method and system

Abstract

The invention discloses a long cruise unmanned platform combined positioning method and system, and the method comprises the steps: building a nonlinear state equation of a long cruise unmanned platform based on the real-time three-dimensional acceleration of the long cruise unmanned platform; the nonlinear state equation contains non-Gaussian state noise; acquiring a signal arrival angle and signal arrival time between the long cruise unmanned platform and the signal emission base station calibrated by the initial coordinate; establishing a measurement equation of the long cruise unmanned platform by combining the height value; the measurement equation contains non-Gaussian measurement noise; performing approximate distribution modeling on the non-Gaussian state noise and the non-Gaussian measurement noise; and based on the approximate distribution model of the non-Gaussian state noise and the non-Gaussian measurement noise, designing a particle filter probability density function and a particle weight, and calculating by adopting a particle filter algorithm to obtain a positioning result of the long cruise unmanned platform.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent detection of unmanned platforms, and in particular relates to a combined positioning method and system for a long-cruise unmanned platform. Background Art

[0002] The rapid development of robotics, sensor network technology, motion control technology, and artificial intelligence has promoted the rapid development and penetration of unmanned systems into deep-sea and deep-space exploration, and also promoted the transformation of exploration methods from network-centric to data- and decision-centric. A large number of unmanned platforms distributed in deep-sea and deep-space environments, including underwater robots, surface unmanned boats, ground unmanned vehicles, and aerial drones, have the characteristics of dispersed deployment, single functions, flexible networking, and strong anti-destruction capabilities. Since unmanned platforms are the basis of tasks including data collection, data transmission, and formation movement, and location is the natural attribute of people or objects in the physical world, it can explain "what kind of event occurred at what location or in what area", so it is crucial to the positioning and detection of unmanned platforms.

[0003] At present, the positioning technology of unmanned platforms is mainly divided into satellite positioning, inertial navigation, image matching and combined positioning according to different sensors and working modes. The principle of satellite positioning is that the unmanned platform receives signals from at least four satellites, and can obtain the three-dimensional coordinates of the unmanned platform through distributed solution. However, when the satellite signal is interfered or shielded, there is a large positioning error or even no output; the principle of inertial navigation is to measure the angular rate and acceleration of the unmanned platform fuselage, and solve the three-axis attitude and position, but there is a large cumulative error in long cruise and external correction is required; the principle of image matching is to obtain the position of the unmanned platform by comparing it with the images in the map library taken in advance by the unmanned platform, but the weather conditions and flight attitude have a great influence on the positioning accuracy. Summary of the invention

[0004] Purpose of the invention: In order to solve the problem of unpredictable positioning errors caused by complex noise in long cruise unmanned platforms due to the unreliability of communication networks in complex sea and air environments, uncertainty in sensor measurements, and volatility in positioning solution errors, the present invention proposes a combined positioning method and system for long cruise unmanned platforms, which uses multi-sensor fusion to improve the positioning performance of long cruise unmanned platforms and solves the problem of error divergence of unmanned platforms under non-Gaussian noise measurements.

[0005] Technical solution: A combined positioning method for a long cruise unmanned platform, comprising the following steps:

[0006] Based on the real-time three-dimensional acceleration of the long cruise unmanned platform, a nonlinear state equation of the long cruise unmanned platform is established; the nonlinear state equation contains non-Gaussian state noise;

[0007] Obtain the signal arrival angle and signal arrival time between the long cruise unmanned platform and the signal transmitting base station calibrated by the initial coordinates;

[0008] The signal arrival angle and the signal arrival time are combined with the height value of the long cruise unmanned platform to establish a measurement equation for the long cruise unmanned platform; the measurement equation contains non-Gaussian measurement noise;

[0009] Approximate distribution modeling is performed for non-Gaussian state noise and non-Gaussian measurement noise respectively;

[0010] Based on the approximate distribution model of non-Gaussian state noise and non-Gaussian measurement noise, the particle filter probability density function and particle weight are designed, and the state of the long cruise unmanned platform is calculated using the particle filter algorithm.

[0011] Furthermore, the nonlinear state equation of the long cruise unmanned platform is established based on the real-time three-dimensional acceleration of the long cruise unmanned platform, which specifically includes:

[0012] The state position component of the long cruise unmanned platform at time k is [x k ,y k ,z k ] T , the velocity component is The acceleration component is Based on the real-time three-dimensional acceleration of the long cruise unmanned platform, the nonlinear state equation of the long cruise unmanned platform at time k is established and expressed as:

[0013] x k+1 =Φ k (x k )+w k

[0014] In the formula, represents the state vector, w k represents non-Gaussian state noise, Φ k (x k )=Ax k +Gf k (x k ) represents the state transfer matrix, T represents the sampling interval, nonlinear term

[0015] Furthermore, the signal arrival angle and signal arrival time are combined with the altitude value to establish a measurement equation for a long cruise unmanned platform. The specific operations include:

[0016] Assume that the signal arrival angle at time k is θ k The signal arrival time is t k , signal arrival time t kMultiply by the signal propagation speed to get the distance d between the long cruise unmanned platform and the space signal transmitting base station to be positioned k , combined with the altitude value z of the long cruise unmanned platform k , establish the measurement equation of the long cruise unmanned platform, expressed as:

[0017] z k =h(x k )+v k

[0018] In the formula, v k is non-Gaussian measurement noise.

[0019] Furthermore, the approximate distribution modeling of the non-Gaussian state noise and the non-Gaussian measurement noise respectively includes:

[0020] Assume the non-Gaussian state noise is w k , the non-Gaussian measurement noise is v k , respectively, approximate distribution modeling is performed, expressed as:

[0021] p(w k )=St(w k ; 0,Q k ,υ1)

[0022] p(v k )=St(v k ; 0,R k ,υ2)

[0023] In the formula, Q k and R k They represent the scale matrices of non-Gaussian state noise and non-Gaussian measurement noise respectively; υ1 and υ2 represent the degree of freedom parameters of non-Gaussian state noise and non-Gaussian measurement noise respectively.

[0024] Furthermore, the approximate distribution model based on non-Gaussian state noise and non-Gaussian measurement noise is used to design a particle filter probability density function and a particle weight, which specifically includes:

[0025] According to the state prior distribution p(x0)~Nx0,P0) of the long cruise unmanned platform, the initialization state is sampled Among them, P0 is the initial covariance matrix;

[0026] Based on the state transfer matrix Φ k (x k ) and the particle at the previous moment, through Generate predicted state particles, where According to the state transfer matrix Φ k (x k) generated by the state prior sampling density function, is the sampling particle;

[0027] According to the state prior density function Sampled Particles Screening is performed, and the particle importance weight is calculated by combining the measurement equation of the long cruise unmanned platform and the signal arrival angle, signal arrival time and height value of the long cruise unmanned platform at time k:

[0028]

[0029] in, is the probability density function generated according to the measurement equation of the long cruise unmanned platform;

[0030] Normalized importance weights:

[0031]

[0032] Particles with small weights are eliminated, and particles with large weights are copied to generate new particles. The state of the long-range cruise unmanned platform at time k is output based on the average of the resampled particles, which is expressed as:

[0033]

[0034] Where N is the number of particles.

[0035] The present invention also discloses a long cruise unmanned platform combined positioning system, comprising:

[0036] A signal transmitting base station is installed in the space to be positioned;

[0037] Long cruise unmanned platform,

[0038] An acceleration sensor, installed on the long cruise unmanned platform, for obtaining the real-time three-dimensional acceleration of the long cruise unmanned platform;

[0039] A signal receiver, installed on the long cruise unmanned platform, for receiving information from a signal transmitting base station;

[0040] An altitude sensor, installed on the long cruise unmanned platform, for obtaining an altitude value of the long cruise unmanned platform;

[0041] A positioning data processing module is installed on the long cruise unmanned platform and is used to execute the steps of a long cruise unmanned platform combined positioning method.

[0042] Beneficial effects: The present invention establishes a state equation of a long cruise unmanned platform based on an acceleration measurement sensor, establishes a measurement equation of a long cruise unmanned platform based on signal arrival angle, arrival time, and altitude measurement, performs approximate distribution modeling on state noise and measurement noise of non-Gaussian distribution, calculates the position value of the long cruise unmanned platform based on an improved particle probability density function and particle weight, performs posterior Cramer-Rao iterative derivation in combination with the approximate distribution of non-Gaussian noise, calculates the lower bound of the positioning error variance of the long cruise unmanned platform, and performs positioning performance analysis of the long cruise unmanned platform. Compared with the prior art, the present invention has the following advantages:

[0043] (1) The present invention takes into account the non-Gaussian heavy-tailed noise in the state equation and measurement equation of the long cruise unmanned platform, and reduces the influence of the unmodeled multi-sensor non-Gaussian noise on the combined positioning accuracy of the long cruise unmanned platform by approximate distribution modeling of the non-Gaussian noise;

[0044] (2) Due to the uncertainty in the positioning detection of the long cruise unmanned platform, the particle set passing through the Gaussian probability density function cannot represent the real state, which makes the unmanned platform filter positioning error diverge and affects the positioning performance. Therefore, the present invention adopts an improved particle filter probability density function and particle weight for iterative calculation;

[0045] (3) The positioning accuracy of the long cruise unmanned platform is crucial to evaluating the performance of the combined positioning method. The present invention obtains the error variance of the entire positioning space through the a posteriori Cramer-Rao equation, thereby providing a theoretical basis for subsequent location-based related service applications;

[0046] (4) The present invention takes into account the non-Gaussian noise in the positioning measurement process of the long cruise unmanned platform, and carries out iterative estimation and accuracy evaluation of the nonlinear positioning equation after approximate distribution modeling, which can provide a basis for the coordinated control of the unmanned platform and related location-based applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 A schematic diagram of a combined positioning method and system for a long cruise unmanned platform proposed by the present invention;

[0048] Figure 2 This is a combined positioning performance diagram of a long cruise unmanned platform proposed by the present invention. DETAILED DESCRIPTION

[0049] The technical solution of the present invention is now further described in conjunction with the accompanying drawings and embodiments.

[0050] like Figure 1 As shown, this embodiment proposes a combined positioning method for a long cruise unmanned platform, which mainly includes the following steps:

[0051] Step 1: Use the acceleration sensor installed on the fuselage of the long cruise unmanned platform to measure the real-time three-dimensional acceleration, which is used to establish the nonlinear state equation of the long cruise unmanned platform under the non-Gaussian noise approximate distribution; specifically including:

[0052] The state component of the long cruise unmanned platform at time k is [x k ,y k ,z k ] T , the velocity component is The acceleration component is

[0053] The acceleration sensor measures the acceleration value in real time, and the nonlinear state equation of the long cruise unmanned platform under the approximate distribution of non-Gaussian noise at time k is established, which is expressed as:

[0054] x k+1 =Φ k (x k )+w k

[0055] In the formula, represents the state vector; w k represents state noise; Φ k (x k )=Ax k +Gf k (x k ) represents the state transfer matrix, T represents the sampling interval; nonlinear term

[0056] Step 2: Communicate with the signal transmitting base station of the space to be positioned through the signal receiver installed on the long cruise unmanned platform, and measure the signal arrival angle and signal arrival time between the long cruise unmanned platform and the signal transmitting base station calibrated by the initial coordinates; specifically including:

[0057] The initial coordinates of the signal transmitting base station to be deployed in the positioning space are [x0, y0, z0] T , communicate with the signal receiver installed on the fuselage of the long cruise unmanned platform, and obtain the signal arrival angle θ between the long cruise unmanned platform and the signal transmitting base station of the space to be positioned at time k k and signal arrival time t k .

[0058] Step 3: The height of the long cruise unmanned platform is measured by the height sensor installed on the long cruise unmanned platform. The signal arrival angle and arrival time are combined with the height of the long cruise unmanned platform to establish the measurement equation of the long cruise unmanned platform under the non-Gaussian noise approximate distribution. The specific operations include:

[0059] Signal arrival time t k Multiply by the signal propagation speed to get the distance d between the long cruise unmanned platform and the signal transmitting base station in the space to be positioned k , combined with the height value z measured by the height sensor k The measurement equation of the long cruise unmanned platform under the non-Gaussian noise approximate distribution is established and expressed as:

[0060] z k =h(x k )+v k

[0061] Where: v k To measure noise.

[0062] Step 4: Considering that the non-Gaussian characteristics of noise in the state equation and measurement equation affect the positioning performance, approximate distribution modeling is performed on the state noise and measurement noise with non-Gaussian heavy-tail characteristics to reduce the impact of modeled non-Gaussian noise; the specific operations include:

[0063] Considering the state noise w k and measurement noise v k It has non-Gaussian characteristics and is prone to cause large positioning errors, so the state noise w k and measurement noise v k Approximate distribution modeling is performed respectively, expressed as:

[0064] p(w k )=St(w k ; 0,Q k ,υ1)

[0065] p(v k )=St(v k ; 0,R k ,υ2)

[0066] Where: Q k and R k They represent the scale matrices of state noise and measurement noise respectively; υ1 and υ2 represent the degree of freedom parameters of state noise and measurement noise respectively.

[0067] Step 5: Based on the non-Gaussian approximate distribution modeling of state noise and measurement noise with non-Gaussian heavy-tail characteristics, an improved particle filter probability density function is designed, and the particle weights are designed in combination with the observed state and the latest observation value under the non-Gaussian noise approximate distribution. After iterative calculation, the filtered positioning estimate of the long cruise unmanned platform is obtained. Specifically, the following steps are included:

[0068] According to the state prior distribution p(x0)~Nx0,P0) of the long cruise unmanned platform, the initialization state is sampled Where P0 is the initial covariance matrix;

[0069] Based on the state transfer matrix and the particles at the previous moment, Generate predicted state particles, where According to the state transfer matrix Φ k (x k ) generated by the state prior sampling density function, is the sampling particle;

[0070] According to the state prior density function Sampled Particles Screening, combined with measurement equation z k =h(x k )+v k As well as the signal arrival angle, signal arrival time and the height of the long cruise unmanned platform at time k, calculate the particle importance weight in: is the probability density function generated according to the measurement equation;

[0071] Normalized importance weights

[0072] Particles with small weights are eliminated, particles with large weights are copied to generate new particles, and the state of the long-range cruise unmanned platform at time k is output based on the average of the resampled particles: Where N is the number of particles.

[0073] Step 6: Establish the a posteriori Cramer-Rao estimation equation for the performance evaluation of the positioning error variance, that is, based on the probability density function of the long cruise unmanned platform filter estimate and the non-Gaussian noise approximate modeling, derive the a posteriori Cramer-Rao estimation equation of the long cruise unmanned platform under non-Gaussian noise, calculate the positioning error variance of the long cruise unmanned platform, and compare it with the filtering positioning estimation error variance, which specifically includes the following steps:

[0074] Estimating the state of a k-time long-cruise unmanned platform based on an improved particle filter probability density function Substituting it into the state equation and measurement equation, the posterior Cramer-Rao lower bound of the target at time k+1 is derived. The lower bound of the estimated variance of the long cruise unmanned platform is expressed as:

[0075]

[0076] Where: J k is the Fisher information matrix, i,j=1,…,n, where p(X k ,Z k ) is X k and Z k The joint probability density function, X k={x1,x2,…,x k}, Z k ={z1,z2,…,z k}. Substituting it into the Fisher information matrix, we get J k The recursive expression is:

[0077]

[0078]

[0079] Initial Fisher information matrix Where p(x0) is the state prior distribution p(x0)~N(x0,P0).

[0080] According to the established state equation and measurement equation of the long cruise unmanned platform, the state probability density function and measurement probability density function are obtained:

[0081]

[0082] Where: Γ(·) is the gamma function; m is the state dimension.

[0083] Based on filter estimation Calculate the mean Then there is

[0084]

[0085] Where: n is the number of measurement states, the mean

[0086] Taking the logarithmic derivative of the conditional probability density function of the predicted state, we get:

[0087]

[0088] Substituting it into the Fisher information matrix recursion formula, we can get the approximate posterior Cramer-Rao lower bound of the state of the long cruise unmanned platform. To evaluate the positioning performance of the next long-cruise unmanned platform.

[0089] Figure 2 The particle filter error, the improved particle filter error after non-Gaussian modeling, the approximate posterior Cramer-Rao estimate, and the ideal posterior Cramer-Rao estimate are shown respectively. Due to the thick-tailed noise caused by the complex environment, the average error of the position estimation of the improved particle filter algorithm after non-Gaussian noise approximate modeling compared with the particle filter algorithm is reduced from 3.46m to 2.72m; the approximate posterior Cramer-Rao lower bound derived from the probability density function after the improved filter estimate and the non-Gaussian approximate modeling is 0.12m different from the ideal posterior Cramer-Rao, which can be used as a filter algorithm evaluation standard.

[0090] Embodiment 2:

[0091] The present embodiment discloses a combined positioning system for a long cruise unmanned platform, which mainly includes: a long cruise unmanned platform, an acceleration sensor, a signal sending base station, a signal receiver, an altitude sensor and a positioning data processing module, wherein the acceleration sensor, the signal receiver, the altitude sensor and the data processor are installed on the fuselage of the long cruise unmanned platform, and the signal sending base station is installed in the space to be positioned. The collected acceleration value, signal arrival angle, signal arrival time and altitude value are transmitted to the positioning data processor, and the positioning data processor executes the steps of a combined positioning method for a long cruise unmanned platform proposed in Example 1.

Claims

1. A combined positioning method for a long cruise unmanned platform, characterized in that: The following steps are involved: Based on the real-time three-dimensional acceleration of the long cruise unmanned platform, a nonlinear state equation of the long cruise unmanned platform is established; the nonlinear state equation contains non-Gaussian state noise; Obtain the signal arrival angle and signal arrival time between the long cruise unmanned platform and the signal transmitting base station calibrated by the initial coordinates; The signal arrival angle and the signal arrival time are combined with the height value of the long cruise unmanned platform to establish a measurement equation for the long cruise unmanned platform; the measurement equation contains non-Gaussian measurement noise; Approximate distribution modeling is performed for non-Gaussian state noise and non-Gaussian measurement noise respectively; Based on the approximate distribution model of non-Gaussian state noise and non-Gaussian measurement noise, the particle filter probability density function and particle weight are designed, and the state of the long cruise unmanned platform is calculated using the particle filter algorithm.

2. A combined positioning method for a long cruise unmanned platform according to claim 1, characterized in that: The nonlinear state equation of the long cruise unmanned platform is established based on the real-time three-dimensional acceleration of the long cruise unmanned platform, which specifically includes: The state position component of the long cruise unmanned platform at time k is [x k ,y k ,z k ] T , the velocity component is The acceleration component is Based on the real-time three-dimensional acceleration of the long cruise unmanned platform, the nonlinear state equation of the long cruise unmanned platform at time k is established and expressed as: x k+1 =Φ k (x k )+w k In the formula, represents the state vector, w k represents non-Gaussian state noise, Φ k (x k )=Ax k +Gf k (x k ) represents the state transfer matrix, T represents the sampling interval, nonlinear term 3. A combined positioning method for a long cruise unmanned platform according to claim 2, characterized in that: The signal arrival angle and signal arrival time are combined with the altitude value to establish a measurement equation for the long cruise unmanned platform. The specific operations include: Assume that the signal arrival angle at time k is θ k The signal arrival time is t k , signal arrival time t k Multiply by the signal propagation speed to get the distance d between the long cruise unmanned platform and the space signal transmitting base station to be positioned k , combined with the altitude value z of the long cruise unmanned platform k , establish the measurement equation of the long cruise unmanned platform, expressed as: z k =h(x k )+v k In the formula, v k is non-Gaussian measurement noise.

4. The long cruise unmanned platform combined positioning method according to claim 1 is characterized by: The approximate distribution modeling of the non-Gaussian state noise and the non-Gaussian measurement noise respectively specifically includes: Assume the non-Gaussian state noise is w k , the non-Gaussian measurement noise is v k , respectively, approximate distribution modeling is performed, expressed as: p(w k )=St(w k ;0,Q k ,υ1) p(v k )=St(v k ;0,R k ,υ2) In the formula, Q k and R k They represent the scale matrices of non-Gaussian state noise and non-Gaussian measurement noise respectively; υ1 and υ2 represent the degree of freedom parameters of non-Gaussian state noise and non-Gaussian measurement noise respectively.

5. The long cruise unmanned platform combined positioning method according to claim 3 is characterized by: The approximate distribution model based on non-Gaussian state noise and non-Gaussian measurement noise is used to design the particle filter probability density function and particle weight, which specifically includes: According to the state prior distribution p(x0)~N(x0,P0) of the long cruise unmanned platform, the initialization state is sampled Among them, P0 is the initial covariance matrix; Based on the state transfer matrix Φ k (x k ) and the particle at the previous moment, through Generate predicted state particles, where According to the state transfer matrix Φ k (x k ) generated by the state prior sampling density function, is the sampling particle; According to the state prior density function Sampled Particles Screening is performed, and the particle importance weight is calculated by combining the measurement equation of the long cruise unmanned platform and the signal arrival angle, signal arrival time and height value of the long cruise unmanned platform at time k: in, is the probability density function generated according to the measurement equation of the long cruise unmanned platform; Normalized importance weights: Particles with small weights are eliminated, and particles with large weights are copied to generate new particles. The state of the long-range cruise unmanned platform at time k is output based on the average of the resampled particles, which is expressed as: Where N is the number of particles.

6. A long cruise unmanned platform combined positioning system, characterized in that: include: A signal transmitting base station is installed in the space to be positioned; Long cruise unmanned platform, An acceleration sensor, installed on the long cruise unmanned platform, for obtaining the real-time three-dimensional acceleration of the long cruise unmanned platform; A signal receiver, installed on the long cruise unmanned platform, for receiving information from a signal transmitting base station; An altitude sensor, installed on the long cruise unmanned platform, for obtaining an altitude value of the long cruise unmanned platform; A positioning data processing module is installed on the long cruise unmanned platform and is used to execute the steps of a long cruise unmanned platform combined positioning method as described in any one of claims 1 to 5.

Images