PDR / GNSS autonomous navigation method based on flexible adjustment of observation quantities
Through the PDR/GNSS autonomous navigation method based on observational elastic adjustment, the GNSS outliers are corrected using Kalman filtering and ARMA model, which solves the problem of unstable GNSS positioning in urban canyon environment and improves the positioning accuracy of pedestrian navigation in smartphones and robustness.
Patent Information
- Application Number
- CN202410996964.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-24
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2044-07-24
AI Technical Summary
In complex environments such as urban canyons, GNSS positioning results are easily affected by multipath effect, resulting in unstable positioning results of combined navigation of smartphones. It is difficult for the existing technology to improve the reliability and robustness of pedestrian navigation under harsh conditions.
Using the PDR/GNSS autonomous navigation method based on observation elastic adjustment, the step size is estimated through GNSS velocity information, the Kalman filtered state equation and measurement equation are established, and the abnormal observation value is observed using the IGG-III type equivalent weight function and ARMA model. Perform correction and depreciation processing to enhance the continuity and reliability of positioning results.
It improves the positioning accuracy and robustness of PDR/GNSS combined navigation, can effectively resist the impact of outliers on filter estimation, and enhances the positioning reliability and continuity of pedestrian navigation under harsh conditions.
Smart Images

Figure CN118518095B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to a PDR / GNSS autonomous navigation method based on elastic adjustment of observed quantities, belongs to a multi-source PNT sensor information fusion technology, and is particularly suitable for the field of multi-sensor elastic PNT. Background Art
[0002] Multi-sensor fusion positioning, navigation and timing (PNT) is one of the mainstream directions of PNT application at present, and the fusion of multiple sensors requires miniaturization and low power consumption, so micro PNT will also become the core technology of comprehensive PNT terminals. At present, smartphones provide people with portable, accurate and personalized navigation services. Their built-in sensors include global navigation satellite system (GNSS), accelerometers, gyroscopes, magnetometers, etc. By using multiple sensors, pedestrians can be provided with better positioning accuracy and related positioning information. At present, pedestrian dead reckoning (PDR) algorithm and GNSS technology are two commonly used positioning methods. The PDR algorithm can calculate the position of pedestrians in real time according to the length and heading of pedestrians, and is not affected by the environment; however, its positioning error will gradually accumulate over time and eventually diverge. Although GNSS can provide users with more accurate positioning accuracy when the signal is good, it is easily interfered by the environment, which leads to frequent cycle slips of GNSS signals and even loss of lock. The integration of the two complementary methods of PDR and GNSS is an effective combination. Therefore, the PDR / GNSS combined navigation method is one of the most widely used solutions for pedestrian navigation.
[0003] However, in complex environments such as urban canyons, GNSS positioning results are easily affected by multipath effects. Once abnormal GNSS observations occur, the positioning results of smartphone integrated navigation will be seriously affected. The elastic random model is a multi-sensor PNT data fusion idea based on the robust estimation criterion. This method can adjust the observation weights in real time by introducing robustness factors, thereby resisting the influence of outliers on filter estimation to a certain extent. On the other hand, due to the lack of redundancy in the loose combination method and factors such as smartphone consumer-grade hardware, simple weight reduction cannot ensure that the positioning results of filter estimation are within a reasonable range. In order to further improve the reliability and robustness of smartphone pedestrian navigation positioning in harsh conditions such as urban canyons, a more effective data fusion method is urgently needed. Summary of the invention
[0004] Purpose of the invention: In order to solve the problem of reliable positioning of smartphone pedestrian navigation in harsh conditions such as urban canyons, the present invention provides a PDR / GNSS autonomous navigation method based on elastic adjustment of observation quantities.
[0005] In order to achieve the above object, the present invention can be implemented by the following technical solution: A PDR / GNSS autonomous navigation method based on elastic adjustment of observation quantity comprises the following steps:
[0006] Step 1, using GNSS speed information to estimate the step length;
[0007] Step 2, establishing the Kalman filter state equation and measurement equation according to the PDR and GNSS models respectively;
[0008] Step 3, based on the Kalman filter principle, respectively calculate the estimated state and prior error covariance matrix of the system;
[0009] Step 4: Calculate the new information vector based on the Kalman filter principle and construct a standardized new information test statistic;
[0010] Step 5, using the IGG-III type equivalent weight function, the standardized innovation test statistic is divided into three segments: no fault, deviation and abnormal;
[0011] Step 6: For deviations, use the robustness factor to downgrade the deviated observations;
[0012] Step 7: For abnormal situations, the auto-regressive moving average (ARMA) model is used to predict the new information of the current abnormal epoch, and the fault amplitude is calculated based on the relationship between the abnormal new information and the predicted new information;
[0013] Step 8: Use the fault amplitude to correct the abnormal observation value of the current epoch, and re-integrate it with the one-step prediction value to output its state estimation and its covariance matrix.
[0014] The step 1 includes the following process:
[0015] Use GNSS speed information to estimate the step length;
[0016] ,
[0017] In the formula, is the estimated step size of GNSS, and Indicates the speed at the current time and the previous time, and Indicates the timestamp of the current moment and the previous moment.
[0018] Step 2 includes the following process: establishing state equations and measurement equations according to PDR and GNSS models respectively,
[0019] ,
[0020] in, is a 4-dimensional state vector, which includes the north position error, the east position error, the step error of the current epoch, and the heading angle error of the current epoch; Indicates the time from epoch arrive The state transition matrix of is the 4-dimensional state vector of the previous epoch; is the observation vector, including the difference between the northing, easting, step length and heading of PDR and GNSS respectively; represents the observation matrix; It is the epoch The process noise vector of epoch The observation noise vector of The process noise vector and epoch The observation noise vector They are independent of each other and are all Gaussian white noise.
[0021] The step 3 includes the following process:
[0022] Based on the Kalman filter principle, the estimated state and error covariance matrix of the system are calculated respectively.
[0023] ,
[0024] in, is the a priori estimate vector for the current epoch; express To epoch The error covariance matrix of is the error covariance matrix of the previous epoch; for System process noise matrix for the epoch.
[0025] Step 4 includes the following process: calculating the standard innovation vector based on the Kalman filter principle and constructing the standardized residual test statistic.
[0026] Step 4-1, calculate the new information vector and its covariance matrix based on the Kalman filter principle:
[0027] ,
[0028] in, is the innovation vector, is the covariance matrix of the corresponding new information vector; in the set Kalman filter, the new information vector should obey the 2D Gaussian distribution, that is , is the observation noise matrix;
[0029] Step 4-2, construct the standardized innovation vector:
[0030] ,
[0031] in, is the normalized innovation vector, is the innovation vector, Represents the diagonal elements of the inverse covariance matrix of the innovation vector.
[0032] Step 5 includes the following process,
[0033] Using the IGG-III type M-estimator and standardized residual test statistics, the detection interval is divided into three sections: no fault, deviation and abnormality.
[0034] ,
[0035] in, Indicates different fault detection events; The current epoch The standardized residuals of observations; and They are two robustness parameters of the equivalent weight function. Usually , .
[0036] Step 6 includes the following process: for deviation, the deviation observation value is downgraded by using the robustness factor.
[0037] ,
[0038] in, is the Kalman gain after the expansion of the observation noise; is the observation vector The equivalent weight matrix; then the equivalent weight function of the IGG-III scheme is selected:
[0039] ,
[0040] in, Indicates For deviation cases, zero weight processing is not considered.
[0041] Step 7 includes the following process: for abnormal situations, the ARMA model is used to predict the new information of the current abnormal epoch, and the fault amplitude is calculated based on the relationship between the abnormal new information and the predicted new information.
[0042] Step 7-1, ARMA new information forecast:
[0043] First, we use the autoregressive moving average model ARMA Predicting time-based innovation sequences, including AR model, Order MA model:
[0044] ,
[0045] in, Indicates the current epoch The new information sequence value, and denote the parameters of the autoregressive model and the moving average model, respectively. The mean is 0 and the variance is Gaussian white noise,
[0046] Then, the minimum value is obtained according to the Akaike information criterion (AIC). ,generally , ,
[0047] ,
[0048] In the formula, Indicates the number of parameters. is the likelihood function;
[0049] Finally, for real-time positioning, the ARMA model needs to be trained and saved at regular intervals. Once an abnormality is detected in the new information of the current epoch, the model predicts the new information data in real time.
[0050] Step 7-2, predict the fault amplitude of the current abnormal observation epoch,
[0051] ,
[0052] In the formula, The current epoch The vector of abnormal magnitudes of ; and are the fault values in the north and east directions respectively Indicates the abnormal new information of the current epoch; is the predicted new information of the current abnormal epoch, where , Respectively represent the new forecast information for the north and east directions.
[0053] Step 8 includes the following process:
[0054] The abnormal observation value of the current epoch is corrected by using the fault amplitude, and then re-fused with the one-step prediction value to output its state estimation and its covariance matrix.
[0055] Step 8-1, use the fault amplitude to correct the current abnormal observation value,
[0056] ,
[0057] in, represents the corrected observed value, is the observed value of anomaly in the north or east direction, is the corresponding repair amplitude,
[0058] Step 8-2: The corrected observation value is re-fused with the one-step prediction value to obtain the optimal estimation result.
[0059] ,
[0060] ,
[0061] in, is the optimal estimate at the current moment, is the Kalman gain of the current epoch; for The state estimate covariance matrix for the epoch, It is a fourth-order unit matrix. The optimal estimate and covariance matrix obtained will be used as the prior estimate and covariance matrix of the next epoch, and the data will be fused again with the observations of the new epoch.
[0062] An electronic device comprises: a processor; and a memory for storing instructions executable by the processor; wherein the processor is configured to execute the instructions to implement the PDR / GNSS autonomous navigation method based on elastic adjustment of observed quantities.
[0063] Compared with the prior art, the advantages of the present invention are as follows: The present invention provides a PDR / GNSS autonomous navigation method based on flexible adjustment of observation quantities. 1) Based on position fusion, this method uses the step size and heading of GNSS as extended observation quantities and performs effective fusion, thereby improving the overall positioning accuracy of PDR / GNSS; 2) Compared with the classic two-segment detection interval, the division of the three-segment detection interval can assign different weights and classification processing according to different degrees of abnormal values, so as to flexibly handle the impact of gross errors on filter estimation; 3) The fault repair enhancement method can make full use of historical new information data to predict the current abnormal new information, and correct the observation value at the current moment, thereby further enhancing the continuity and reliability of pedestrian navigation positioning results. BRIEF DESCRIPTION OF THE DRAWINGS
[0064] The present invention is further described in conjunction with the accompanying drawings and specific implementation methods, highlighting the characteristics and advantages of the present invention in the field of multi-sensor data fusion positioning.
[0065] Figure 1 It is a system block diagram of a PDR / GNSS autonomous navigation method based on elastic adjustment of observation quantities;
[0066] Figure 2 The present invention provides a specific implementation method of the fault recovery enhanced robustness filtering core algorithm execution process. DETAILED DESCRIPTION
[0067] In order to deepen the recognition and understanding of the present invention, the scheme is described in detail below with reference to the accompanying drawings and embodiments.
[0068] Embodiment: A PDR / GNSS autonomous navigation method based on elastic adjustment of observed quantities in an embodiment of the present application is described below. In an embodiment of the present application, the built-in microelectromechanical systems (MEMS) sensors of a smartphone are first used to collect the user's dynamic raw data, such as accelerometers, gyroscopes, and magnetometers. The PDR algorithm system detects the gait by analyzing these sensor data, and estimates the step length of each step and the direction of travel of the user, and finally outputs information such as the user's position, step length, and heading at each moment. In addition, the system's GNSS chip is used to receive satellite signals to calculate the user's absolute positioning results. After coordinate system conversion and time synchronization, these solution data will be subjected to fault detection and fusion processing in the LC FR-RKF module. The specific fault detection and processing methods are as follows. Figure 2 Finally, the processed and fused data will be fed back to the PDR system to compensate for the position, step length and heading information, and finally output the solution result.
[0069] For the overall algorithm flow, see Figure 1 , Figure 2 , a PDR / GNSS autonomous navigation method based on elastic adjustment of observation quantity, comprising the following steps:
[0070] Step 1: Use the GNSS speed and corresponding timestamp information collected by the mobile phone APP to estimate the step length at the current moment:
[0071] Step 1 is as follows: use the speed information of GNSS to estimate the step length;
[0072] ,
[0073] In the formula, is the estimated step size of GNSS, and Indicates the speed at the current time and the previous time, and Indicates the timestamp of the current moment and the previous moment. In addition, GNSS heading information can be directly collected by the mobile phone APP. Therefore, PDR and GNSS systems can effectively integrate position, step length and heading information.
[0074] Step 2: Establish the state equation and measurement equation according to the PDR and GNSS models respectively, including the following process:
[0075] According to the PDR and GNSS models, the state equation and measurement equation are established respectively.
[0076] ,
[0077] in, is a 4-dimensional state vector, which includes the north position error, the east position error, the step error of the current epoch, and the heading angle error of the current epoch; Indicates the time from epoch arrive The state transition matrix of is the 4-dimensional state vector of the previous epoch; is the observation vector, including the difference between the northing, easting, step length and heading of PDR and GNSS respectively; represents the observation matrix; It is the epoch The process noise vector of epoch The observation noise vector of The process noise vector and epoch The observation noise vector They are independent of each other and are all Gaussian white noise.
[0078] Step 3: Calculate the estimated state and error covariance matrix of the system based on the Kalman filter principle.
[0079] Step 3 includes the following process:
[0080] Based on the Kalman filter principle, the estimated state and error covariance matrix of the system are calculated respectively.
[0081] ,
[0082] in, is the a priori estimate vector for the current epoch; express To epoch The error covariance matrix of is the error covariance matrix of the previous epoch; for The system process noise matrix of the epoch. Since the fusion scheme of the present application is based on the Kalman filter fusion processing of the error state, the error state vector at each moment needs to be set to zero.
[0083] Step 4: Calculate the standard innovation vector based on the Kalman filter principle and construct the standardized residual test statistic.
[0084] Based on the Kalman filter principle, the standard innovation vector is calculated and the standardized residual test statistic is constructed.
[0085] Step 4-1, calculate the new information vector and its covariance matrix based on the Kalman filter principle:
[0086] ,
[0087] in, is the innovation vector, is the covariance matrix of the corresponding new information vector; in the set Kalman filter, the new information vector should obey the 2D Gaussian distribution, that is , is the observation noise matrix;
[0088] Step 4-2, construct the standardized innovation vector:
[0089] ,
[0090] in, is the normalized innovation vector, is the innovation vector, Represents the diagonal elements of the inverse covariance matrix of the innovation vector.
[0091] In step 5, the detection interval is divided into three sections using the IGG-III type M-estimator and the standardized residual test statistic: no fault, deviation, and abnormal. Compared with the original IGG-III scheme, each detection interval corresponds to a different treatment measure.
[0092] ,
[0093] in, Indicates different fault detection events; The current epoch The standardized residuals of observations; and They are two robustness parameters of the equivalent weight function. Usually , .
[0094] Step 6: For deviations, use the robustness factor to downgrade the deviated observations.
[0095] ,
[0096] in, is the Kalman gain after the expansion of the observation noise; is the observation vector The equivalent weight matrix; then the equivalent weight function of the IGG-III scheme is selected:
[0097] ,
[0098] in, Indicates For deviation cases, zero weight processing is not considered.
[0099] Step 7: For abnormal situations, the ARMA model is used to predict the new information of the current abnormal epoch, and the fault amplitude is calculated based on the relationship between the abnormal new information and the predicted new information.
[0100] Step 7-1, ARMA new information forecast:
[0101] First, we use the autoregressive moving average model ARMA Predicting time-based innovation sequences, including AR model, Order MA model:
[0102] ,
[0103] in, Indicates the current epoch The new information sequence value, and denote the parameters of the autoregressive model and the moving average model, respectively. The mean is 0 and the variance is Gaussian white noise,
[0104] Then, the minimum value is obtained according to the Akaike information criterion (AIC). ,generally , ,
[0105] ,
[0106] In the formula, Indicates the number of parameters. is the likelihood function;
[0107] Finally, for real-time positioning, the ARMA model needs to be trained and saved at regular intervals. Once an abnormality is detected in the new information of the current epoch, the model can predict the new information data in real time and use the fault amplitude to correct the abnormal GNSS observation value in time.
[0108] Step 7-2, predict the fault amplitude of the current abnormal observation epoch,
[0109] ,
[0110] In the formula, The current epoch The vector of abnormal magnitudes of ; and are the fault values in the north and east directions respectively Indicates the abnormal new information of the current epoch; is the predicted new information of the current abnormal epoch, where , Respectively represent the new forecast information for the north and east directions.
[0111] Step 8: Use the fault amplitude to correct the abnormal observation value of the current epoch, and re-integrate it with the one-step prediction value to output its state estimation and its covariance matrix.
[0112] Step 8-1, use the fault amplitude to correct the current abnormal observation value,
[0113] ,
[0114] in, represents the corrected observed value, is the observed value of anomaly in the north or east direction, is the corresponding repair amplitude. It should be noted that this scheme uses the size of the new information data to judge the relationship between the abnormal observation and the amplitude in real time.
[0115] Step 8-2: The corrected observation value is re-fused with the one-step prediction value to obtain the optimal estimation result.
[0116] ,
[0117] ,
[0118] in, is the optimal estimate at the current moment, is the Kalman gain of the current epoch; for The state estimate covariance matrix for the epoch, It is a fourth-order unit matrix. The optimal estimate and covariance matrix obtained will be used as the prior estimate and covariance matrix of the next epoch, and the data will be fused again with the observations of the new epoch.
[0119] The present invention provides a PDR / GNSS autonomous navigation method based on elastic adjustment of observations. First, the position, estimated step length and heading of GNSS are effectively fused as extended observations of Kalman filtering and the inferred results of PDR algorithm to improve the positioning accuracy of PDR / GNSS combined navigation system. Then, a robust factor is introduced under the framework of Kalman filtering to make corresponding elastic adjustments to abnormal observations. Secondly, in view of the fact that there is no observation redundancy in the loose combination method, the detection interval of the robust Kalman filter is divided into three cases: under normal circumstances, no processing is performed; when deviation occurs, the abnormal observations are downgraded; under abnormal circumstances, the ARMA model is used to reconstruct the new information of the current epoch and repair its amplitude. Then, the observation value of the current abnormal epoch is corrected according to the repaired amplitude. Finally, the one-step prediction value and the corrected observation value are re-processed by Kalman filter data fusion, thereby further improving the robustness and accuracy of the pedestrian navigation system. It should be pointed out that based on the inventive principle of the present invention, it can still be further optimized and improved, and these optimizations and improvements still belong to the protection scope of the present invention, and the scope of protection of the present invention should be based on the claims.
Claims
1. A PDR / GNSS autonomous navigation method based on flexible adjustment of observation quantities, characterized in that: The steps include: Step 1, using GNSS speed information to estimate the step length; Step 2, establishing the Kalman filter state equation and measurement equation according to the PDR and GNSS models respectively; Step 3, based on the Kalman filter principle, respectively calculate the estimated state and prior error covariance matrix of the system; Step 4: Calculate the new information vector based on the Kalman filter principle and construct a standardized new information test statistic; Step 5, using the IGG-III type equivalent weight function, the standardized innovation test statistic is divided into three segments: no fault, deviation and abnormal; Step 6: For deviations, use the robustness factor to downgrade the deviated observations; Step 7: For abnormal situations, the ARMA model is used to predict the new information of the current abnormal epoch, and the fault amplitude is calculated based on the relationship between the abnormal new information and the predicted new information; Step 8: Use the fault amplitude to correct the abnormal observation value of the current epoch, and re-integrate it with the one-step prediction value to output its state estimation and its covariance matrix; Among them, step 1 is as follows: Use GNSS speed information to estimate the step length; In the formula, SL GNSS is the estimated step size of GNSS, v k and v k-1 Indicates the speed at the current moment and the previous moment, T k and T k-1 Indicates the timestamp of the current moment and the previous moment. Wherein, step 2 includes the following process: According to the PDR and GNSS models, the state equation and measurement equation are established respectively. Among them, O k is a 4-dimensional state vector, including the north position error, the east position error, the step error of the current epoch, and the heading angle error of the current epoch; Γ k,k-1 represents the state transition matrix from epoch k-1 to k; O k-1 is the 4-dimensional state vector of the previous epoch; z k is the observation vector, including the difference between the northing, easting, step length and heading of PDR and GNSS; H k represents the observation matrix; w k-1 is the process noise vector of epoch k-1; ε k is the observation noise vector at epoch k; the process noise vector w at epoch k-1 k-1 and the observation noise vector ε at epoch k k are independent of each other and are all Gaussian white noise. Wherein, step 3 includes the following process: Based on the Kalman filter principle, the estimated state and error covariance matrix of the system are calculated respectively. in, is the prior estimate vector of the current epoch; P k,k-1 represents the error covariance matrix from k-1 to epoch k; P k-1 is the error covariance matrix of the previous epoch; Q k-1 is the system process noise matrix of k-1 epochs, Wherein, step 4 includes the following process: Based on the Kalman filter principle, the standard innovation vector is calculated and the standardized residual test statistic is constructed. Step 4-1, calculate the new information vector and its covariance matrix based on the Kalman filter principle: Among them, γ k is the innovation vector, S k is the covariance matrix of the corresponding new information vector; in the set Kalman filter, the new information vector should obey the 2D Gaussian distribution, that is, γ k ~N(0,S k ), R k is the observation noise matrix; Step 4-2, construct the standardized innovation vector: in, is the normalized innovation vector, γ k is the innovation vector, diag(·) represents taking the diagonal elements of the inverse covariance matrix of the innovation vector, Wherein, step 5 includes the following process: Using the IGG-III type M-estimator and standardized residual test statistics, the detection interval is divided into three sections: no fault, deviation and abnormality. Among them, A i Represents different fault detection events; δ i is the standardized residual of the i-th observation data in the current epoch; k0 and k1 are two robustness parameters of the equivalent weight function, k0 = 1.0 ~ 1.5, k1 = 3.0 ~ 4.5, Wherein, step 6 includes the following process: For deviations, the robustness factor is used to downgrade the deviated observations. in, is the Kalman gain after the expansion of the observation noise; is the observation vector z k The equivalent weight matrix; then the equivalent weight function of the IGG-III scheme is selected: in, Represents the robustness factor of the i-th dimension; for deviation, zero weight processing is not considered. Wherein, step 7 includes the following process: For abnormal situations, the ARMA model is used to predict the new information of the current abnormal epoch, and the fault amplitude is calculated based on the relationship between the abnormal new information and the predicted new information. Step 7-1, ARMA new information forecast: First, the autoregressive moving average model ARMA (p, q) is used to predict the time-based new information series, including the p-order AR model and the q-order MA model: Among them, y k represents the value of the new information sequence at the current epoch k, ψ i and θ j denote the parameters of the autoregressive model and the moving average model, respectively. The mean is 0 and the variance is σ 2 Gaussian white noise, Then, according to the AIC criterion, the minimum (p, q) is found, p∈[1, 2, 3], q∈[1, 2, 3], AIC=2N-2ln(L), In the formula, N represents the number of parameters, and L is the likelihood function; Finally, for real-time positioning, the ARMA model needs to be trained and saved at regular intervals. Once an abnormality is detected in the new information of the current epoch, the model predicts the new information data in real time. Step 7-2, predict the fault amplitude of the current abnormal observation epoch, Where M is the vector of anomaly amplitudes at the current epoch k; and are the fault values in the north and east directions respectively Indicates the abnormal new information of the current epoch; is the predicted new information of the current abnormal epoch, where Respectively represent the new forecast information in the north and east directions, Step 8 includes the following process: The abnormal observation value of the current epoch is corrected by using the fault amplitude, and then re-fused with the one-step prediction value to output its state estimation and its covariance matrix. Step 8-1, use the fault amplitude to correct the current abnormal observation value, in, represents the corrected observed value, is the observed value of anomaly in the north or east direction, is the corresponding repair amplitude, Step 8-2: The corrected observation value is re-fused with the one-step prediction value to obtain the optimal estimation result. Pk=(I-K k H)P k,k-1 , in, is the optimal estimate at the current moment, K k is the Kalman gain of the current epoch; P k is the state estimation covariance matrix of the k epoch, I is the fourth-order unit matrix, and the optimal estimate and covariance matrix obtained will be used as the prior estimate and covariance matrix of the next epoch, and will be re-fused with the observations of the new epoch.
2. An electronic device, characterized in that: include: processor; A memory for storing instructions executable by the processor; wherein the processor is configured to execute the instructions to implement the PDR / GNSS autonomous navigation method based on elastic adjustment of observation quantities as claimed in claim 1.
