An adaptive integrated navigation system for vehicles

Abstract

The invention discloses an adaptive combined navigation system applied to a vehicle, and relates to the field of navigation technology. The system comprises a data acquisition module, a data storage module, a state analysis module, a model optimization module, a factor optimization module, a filtering analysis module, and a combination analysis module; the data acquisition module acquires navigation data in the vehicle running process through a vehicle sensor, and stores the data in the data storage module; the state analysis module then establishes a state prediction analysis model according to the navigation data, predicts and analyzes the running state of the vehicle at the next moment, obtains a predicted driving state, and then judges the driving state according to a state pool, and selects different adaptive combined navigation algorithm optimization methods according to the judgment result, thereby solving the problem of inaccurate estimation caused by inaccurate selection of adaptive factors, and at the same time, timely optimizing and analyzing the adaptive combined navigation model when facing a complex driving environment, thereby greatly improving the accuracy of the adaptive combined navigation.

Description

Technical Field

[0001] The invention relates to the field of navigation technology, and in particular to an adaptive combined navigation system applied to a vehicle. Background Art

[0002] In order to adapt to the complex and changeable operating environment, the vehicle navigation adopts fused combined navigation to improve the overall navigation performance. Different combined navigation systems adopt different adaptive filtering algorithms for fusion. For example, in the prior art, the GNSS / INS combined navigation system adopts the Kalman filter algorithm for fusion, but the Kalman filter algorithm needs to have a definite and accurate dynamic model and noise statistical characteristics. The disturbance problem of abnormal observation and dynamic model will seriously affect the performance of the filter. In the fusion process, a reasonable adaptive factor can balance the weight ratio of the dynamic model and the observation information, and solve the influence of observation noise and system state disturbance on parameter estimation. Therefore, when the vehicle always appears in an abnormal state, the construction and selection of the adaptive factor are very important. The optimized adaptive factor is combined with the estimation process of the adaptive combined navigation filtering algorithm, so as to improve the stability and estimation accuracy of the system. In order to reduce the influence of observation error and system disturbance interference, the optimal adaptive factor is selected for parameter estimation. The present invention proposes an adaptive combined navigation system applied to vehicles. Summary of the invention

[0003] In view of the above situation, in order to overcome the defects of the prior art, the purpose of the present invention is to provide an adaptive combined navigation system applied to vehicles, wherein the state analysis module of the system predicts the driving state of the vehicle at the next moment by constructing a state prediction analysis model, and when the predicted driving state has appeared in the historical state pool, the optimal adaptive combined navigation analysis algorithm is selected according to the difference between different observation sequences under the same driving state, and the model is optimized to achieve accurate analysis of the adaptive process at the next moment; when a new driving state that is not in the state pool appears, it means that new data that has not been monitored has appeared and has an impact on the driving state, and the number of observation items of the prediction sequence is less than the observation items applied to the driving state image, that is, other observation items are retrieved through similarity analysis and combined with the prediction sequence to expand the analyzed data items, and then the adaptive factors are optimized and selected through the combination of the discrete node graph and the adaptive factor graph, and the adaptive combined navigation analysis algorithm is constructed according to the optimized adaptive factors, which solves the problem of adaptive factor selection when an abnormal state occurs and improves the adaptability to the abnormal state.

[0004] In order to improve the navigation accuracy of the vehicle's adaptive integrated navigation system, the present invention proposes an adaptive integrated navigation system applied to the vehicle, comprising a data acquisition module, a data storage module, a state analysis module, a model optimization module, a factor optimization module, a filtering analysis module, and a combination analysis module;

[0005] The data acquisition module acquires navigation data during vehicle operation through vehicle sensors, pre-processes the navigation data, and stores it in the data storage module;

[0006] The state analysis module establishes a state prediction analysis model according to the navigation data, performs prediction analysis on the running state of the vehicle at the next moment, obtains the predicted state analysis result of the vehicle, and constructs the state pool of the vehicle according to the state analysis result corresponding to the actual state of the vehicle at the historical moment;

[0007] Calculate the correlation between the predicted state analysis result and the state analysis result in the state pool, determine whether a driving state that is not in the state pool appears, when the predicted driving state appears in the state pool, calculate the state deviation between the driving state in the predicted state analysis result and the driving state at the current moment, when the state deviation exceeds the determination threshold, the predicted driving state and the optimal adaptive combined navigation algorithm corresponding to the current driving state are different, the model optimization module calls the optimal adaptive combined navigation algorithm corresponding to the predicted driving state, and performs adaptive combined navigation optimization according to the optimal adaptive combined navigation algorithm;

[0008] When the predicted driving state is a driving state that has not appeared in the state pool, the similarity between the predicted driving state and the historical driving state in the state pool is calculated, and the similarity relative to different historical driving states is mapped onto a discrete node graph;

[0009] The factor optimization module analyzes the adaptive factor graph corresponding to the adaptive combined navigation algorithm in the state pool in combination with the predicted driving state, combines the discrete node graph with the adaptive factor graph, optimizes and analyzes the node relationship of the adaptive factor, and selects the optimal adaptive factor;

[0010] Different adaptive factors correspond to different measurement covariance matrices. After the optimal adaptive factor is selected, the filtering analysis module constructs the corresponding filtering analysis model based on the selected optimal adaptive factor, and then uses the filtering analysis model to perform filtering analysis on the adaptive combined navigation process at the next moment. In the filtering analysis process, the adaptive factor is used to optimize the estimation process.

[0011] The combined analysis module performs adaptive combined navigation control on the combined navigation according to the error check result of the filtering analysis module.

[0012] Preferably, under the same driving operation, the driving state of the vehicle is the result of comprehensive analysis of the navigation data. The navigation data of the vehicle is used as input data of the dynamic analysis model to construct a state prediction analysis model of the driving state, as follows:

[0013] a. Convert navigation data into observation data and construct an optimization objective function. The optimized objective minimizes the overall error.

[0014]

[0015] Among them, J(T) is the error function, T is the vehicle's posture matrix, z i For point p i The observed data, the measurement equation is is the observation error, N is the number of observation data;

[0016] b. Let X k is the state matrix at the kth moment, u k is external input, w k is the noise, n k+1 is the landmark measurement pixel noise, A is the increment coefficient, and the prediction of the state at the next moment is:

[0017] X k+1 =f(X k ,u k , w k+1 )=X k exp(Ω(X k ,u k )+n k+1 )+AX k ;

[0018] c. Equation of state ξ k is the position change matrix, the length of the time period between Ω moments, X k and u k It is known that k In input mode, X k The state changes to state X k+1 ;

[0019] d. Determine the driving state X at the next moment based on data analysis of the observed data k+1 The current driving state is jointly determined by different parameters of the current navigation data. As the driving environment changes, the correlation between different parameters changes with time. The current navigation data is used as input to obtain the predicted state analysis result at the next moment. The predicted state analysis result includes the predicted driving state and the corresponding prediction sequence.

[0020] Furthermore, the driving state is presented as a result, which is obtained by analyzing different navigation data. When the driving state is the same, the difference between the corresponding navigation data is recorded as the state deviation. The observation sequence x={x i ,i=1,2,3,...,M},xi is the observed value, M is the number of observation items, and the difference set {Δ i =x i -y i , i = 1, 2, 3, ..., M}, the two observation sequences to be compared are recorded as two phase points in the M-dimensional phase space, x and y = {x i , i = 1, 2, 3, ..., M} is recorded as C(r), and the correlation function C(r) is calculated

[0021]

[0022] Where r is the given critical distance, θ is the unit function, C(r) is the number of phase points in the phase space whose distance is less than r, and 2 The proportion of the pairing in the data is then analyzed through the correlation function to obtain the correlation dimension d, Md is the deviation dimension, The state deviation is then judged, and the model optimization module optimizes the adaptive combined navigation algorithm according to the judgment result.

[0023] Furthermore, the model optimization module uses different adaptive combined navigation algorithms to perform adaptive combined analysis on the prediction sequence to obtain corresponding adaptive combined navigation analysis results. The model optimization module analyzes the adaptability between different models and observation data to obtain an adaptability vector P, and then constructs a transformation relationship between data changes and models through vector P, thereby selecting the optimal model parameters, and then selecting the optimal adaptive combined navigation analysis algorithm based on the optimal model parameters.

[0024] Preferably, when a driving state that has not appeared in the state pool appears, the optimal adaptive factor of the predicted driving state is to be determined, and the similarity between the predicted driving state and the historical driving state in the state pool is analyzed. Then, using the similarity as a proportional parameter, the observation items related to the predicted sequence in the state pool are extracted, and the extracted observation items are combined with the predicted sequence to obtain a new adaptive observation sequence. The adaptive combined navigation algorithm in the state pool is then used to adaptively analyze the adaptive observation sequence, and the adaptive factors corresponding to different algorithms are mapped to the node space to obtain a discrete node graph.

[0025] Preferably, different adaptive combined navigation algorithms in the state pool analyze the observation sequence of historical navigation data to obtain the corresponding adaptive factor graph, analyze the new adaptive observation sequence to obtain the discrete node graph, and then combine the adaptive factor graph with the discrete node graph, perform range check on the adaptive factor graph through the combination, and perform global search optimization on all nodes to select the optimal adaptive factor.

[0026] Preferably, the discrete node graph is combined with the adaptive factor graph, and the specific process is as follows:

[0027] Step 1: construct a Moral graph, insert different nodes in the discrete node graph into the adaptive factor graph, connect them with the adaptive factor graph according to the distribution position, connect the parent nodes of the same node in the adaptive factor graph to each other, and remove the truncated head of each connection;

[0028] Step 2: Triangulate the graph. For a ring containing 4 or more nodes, add an undirected edge to connect two non-adjacent nodes in the ring to complete the triangulation of the Moral graph.

[0029] Step 3: Distinguish the clique nodes and determine the unity points in the triangulated graph. Each unity point is a subgraph of the undirected graph.

[0030] Step 4: Establish a union tree. The established union tree must include all cluster nodes, and the intersection is a subgraph of the undirected graph that connects the two cluster nodes.

[0031] Preferably, according to the distribution probability of the joint nodes, the nodes in the discrete node graph and the adaptive factor graph are merged to obtain new adaptive factor distribution nodes, the adaptability of different adaptive factors is evaluated, and then the particle swarm optimization method is used to optimize the node relationship of the adaptive factor, search for the adaptive factor in the best position globally, optimize the adaptive factor, and select the optimal adaptive factor.

[0032] Preferably, the navigation data collected by the data acquisition module includes inertial navigation data, vehicle status data and driving environment data. The vehicle's adaptive combined navigation algorithm performs adaptive analysis based on the navigation data collected at different times, and constructs a vehicle status pool based on the data corresponding to the vehicle's historical driving status during the adaptive analysis process. The status pool includes all driving statuses that have occurred.

[0033] Due to the adoption of the above technical solution, the present invention has the following advantages compared with the prior art:

[0034] 1. The state analysis module of the present invention constructs a state prediction analysis model based on navigation data, predicts the driving state of the vehicle at the next moment, and judges the predicted driving state. When the predicted driving state has appeared in the historical state pool, the optimal adaptive combined navigation analysis algorithm is selected according to the difference between different observation sequences under the same driving state, and the model is optimized to achieve accurate analysis of the adaptive process at the next moment. When facing a complex formal environment, the predicted driving state is difficult to determine. When a new driving state that is not in the state pool appears, it means that new data that has not been monitored has appeared and has affected the driving state. The number of observation items in the prediction sequence is less than the observation items applied to the driving state image. The present invention uses similarity analysis to call other observation items and combine them with the prediction sequence, thereby expanding the data items for analysis.

[0035] 2. The model optimization module of the present invention optimizes the model of the driving states that have appeared in the state pool. The same driving state means that the driving state of the entire vehicle is the same, which is a result state. Maintaining the same driving state requires the result of the joint action of different observation data and sub-navigation systems. There are differences in the data collected by sensors of different vehicles at different times, but the driving states may be the same, so the corresponding adaptive combined navigation analysis algorithms are also different. The influence of different selected adaptive factors on parameter estimation is different. The model optimization module of the present invention optimizes and analyzes the adaptive combined navigation model when the driving states are the same, which greatly improves the accuracy of the adaptive combined navigation.

[0036] 3. The factor optimization module of the present invention is an analysis process for when a driving state that has not been in the state pool appears. When the driving state changes, the change of the observation sequence is observed. The state prediction analysis model is an analysis of the state at the next moment based on the navigation data at the current moment. In the analysis process, the discrete node graph and the adaptive factor graph are combined, and the combined adaptive factor makes the selection range of the adaptive factor larger. The adaptive factor is optimized and selected, and an adaptive combined navigation analysis algorithm is constructed based on the optimized adaptive factor. This solves the problem of adaptive factor selection when an abnormal state occurs, and improves the adaptability to abnormal states. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is an overall analysis flow chart of the present invention;

[0038] Figure 2 It is a module diagram of the present invention. DETAILED DESCRIPTION

[0039] The above and other technical contents, features and effects of the present invention are described in detail below with reference to the attached Figure 1 To the attached Figure 2The detailed description of the embodiments will clearly show that the embodiments of the present application and the features of the embodiments can be combined with each other, and the terms used in the specification have the meanings commonly understood by those skilled in the art of the present invention.

[0040] The adaptive algorithm in the prior art is an adaptive combination algorithm based on different analysis focuses. When the data mutates, an adaptive unscented Kalman filter combined navigation algorithm based on a particle swarm is used to build a model based on the vehicle's posture and motion state, and then use the Kalman filter to estimate the posture state. The adaptive filtering makes the analysis process more reliable in the current observation, thereby compensating for the random model error. According to the construction of the adaptive factor, it is divided into forgetting filtering, fading memory filtering, and adaptive filtering, and an adaptive factor needs to be constructed.

[0041] The adaptive combined navigation strategy corresponds to the operating state. When the operating state changes, different parameter items and parameter values ​​are changed, and the most suitable adaptive factor at the next moment is different.

[0042] When the state of the vehicle changes, the most adaptable adaptive factor occurs. The most adaptable node of the most adaptable factor is analyzed according to the node diagram. After the most adaptable factor is determined, a filtering analysis model corresponding to the adaptive factor is established to perform adaptive filtering. The influence of the adaptive factor is considered within the parameter evaluation range, and the vehicle is subjected to adaptive navigation filtering analysis to determine the adaptive combined navigation strategy.

[0043] The adaptive process of the combined navigation system is to use adaptive decision-making strategies, according to the inertial navigation information and single positioning algorithms such as GPS, to calculate the weight distribution of the two through adaptive algorithms to achieve intelligent combination. Various adaptive algorithms may be used in this process, such as the least mean square error algorithm (LMS algorithm) and the recursive least squares algorithm (RLS algorithm). These algorithms dynamically adjust and optimize the combination mode and weight of each subsystem according to the real-time performance and environmental conditions of the system. The positioning accuracy of the adaptive intelligent combined navigation algorithm is relatively simple in complex environments and signal interference conditions. It has good anti-seismic and anti-interference capabilities through adaptation, and can operate stably under different speeds and vibration conditions, allowing automatic switching between navigation subsystem working modes, thereby further improving the system's working reliability.

[0044] An adaptive integrated navigation system applied to a vehicle, comprising a data acquisition module, a data storage module, a state analysis module, a model optimization module, a factor optimization module, a filtering analysis module, and a combination analysis module;

[0045] The data acquisition module acquires navigation data during vehicle operation through vehicle sensors, pre-processes the navigation data, and stores it in the data storage module;

[0046] The traditional adaptive combined navigation algorithm has the problem of slow response and inability to respond quickly to sudden dangerous road conditions. The non-integrity constraint is applied to the inertial navigation system combination, which can effectively suppress the error divergence of the combined navigation system during signal interruption. The noise setting is based on a fixed empirical value. However, due to the complex and changeable trajectory of the vehicle, its motion state cannot fully meet the premise assumptions. The noise given by experience cannot accurately reflect the actual motion of the vehicle. A noise adaptive method based on the vehicle state is constructed.

[0047] The state analysis module establishes a state prediction analysis model according to the navigation data, performs prediction analysis on the running state of the vehicle at the next moment, obtains the predicted state analysis result of the vehicle, and constructs the state pool of the vehicle according to the state analysis result corresponding to the actual state of the vehicle at the historical moment;

[0048] Calculate the correlation between the predicted state analysis result and the state analysis result in the state pool, determine whether a driving state that is not in the state pool appears, when the predicted driving state appears in the state pool, calculate the state deviation between the driving state in the predicted state analysis result and the driving state at the current moment, when the state deviation exceeds the determination threshold, the predicted driving state and the optimal adaptive combined navigation algorithm corresponding to the current driving state are different, the model optimization module calls the optimal adaptive combined navigation algorithm corresponding to the predicted driving state, and performs adaptive combined navigation optimization according to the optimal adaptive combined navigation algorithm;

[0049] In the embodiment, the determination threshold is determined according to the difference of the driving state. When the state deviation does not exceed the determination threshold, the predicted driving state is similar to the current driving state. Similar driving states involve the same data indicators of navigation data, but the change intensity of the corresponding values ​​of the data indicators is different. The adaptive combined navigation algorithm at the next moment is optimized on the basis of the adaptive combined navigation algorithm corresponding to the current moment, and there is no need to select and optimize the adaptive combined navigation algorithm again.

[0050] When the predicted driving state is a driving state that has not appeared in the state pool, the similarity between the predicted driving state and the historical driving state in the state pool is calculated, and the similarity relative to different historical driving states is mapped onto a discrete node graph;

[0051] The factor optimization module analyzes the adaptive factor graph corresponding to the adaptive combined navigation algorithm in the state pool in combination with the predicted driving state, combines the discrete node graph with the adaptive factor graph, optimizes and analyzes the node relationship of the adaptive factor, and selects the optimal adaptive factor;

[0052] Different vehicle operation status fluctuation factors have different corresponding most suitable adaptive factors. An experience pool of vehicle operation status information is constructed. After training data samples, new driving status that appears is analyzed from the experience pool.

[0053] Adaptive navigation and positioning must have a reasonable and easy-to-calculate adaptive factor. A reasonable adaptive factor should be able to balance the weight ratio of the dynamic model information and the observation information, and be able to control the impact of the dynamic model error on the navigation parameter solution.

[0054] Different adaptive factors correspond to different measurement covariance matrices. After the optimal adaptive factor is selected, the filtering analysis module constructs the corresponding filtering analysis model based on the selected optimal adaptive factor, and then uses the filtering analysis model to perform filtering analysis on the adaptive combined navigation process at the next moment. In the filtering analysis process, the adaptive factor is used to optimize the estimation process.

[0055] The combined analysis module performs adaptive combined navigation control on the combined navigation according to the error check result of the filtering analysis module.

[0056] The driving state of a vehicle is the result of the interaction between the vehicle operating environment and the vehicle's own operating system. The parameters of the combined navigation coefficients have an important impact on the navigation performance, such as filter parameters and sensor fusion weights. By designing an adaptive parameter optimization algorithm, the system parameters are automatically adjusted according to real-time data to improve navigation performance. At the same time, the system should have a certain self-learning ability and be able to optimize its own performance based on data and experience. However, the adaptive factors corresponding to the empirical analysis cannot cover all vehicle operating states. When data that is not in the experience pool appears, self-optimization cannot be achieved. Under the same driving operation, the driving state of the vehicle is the result of a comprehensive analysis of the navigation data. The navigation data of the vehicle is used as the input data of the dynamic analysis model to construct a state prediction analysis model for the driving state, as follows:

[0057] e. Convert navigation data into observation data and construct an optimization objective function. The optimized objective minimizes the overall error.

[0058]

[0059] Among them, J(T) is the error function, T is the vehicle's posture matrix, z i For point p i The observed data, the measurement equation is is the observation error, N is the number of observation data;

[0060] f. Let X k is the state matrix at the kth moment, u k is external input, w kis the noise, n k+1 is the landmark measurement pixel noise, A is the increment coefficient, and the prediction of the state at the next moment is:

[0061] X k+1 =f(X k ,u k , w k+1 )=X k exp(Ω(X k ,u k )+n k+1 )+AX k ;

[0062] g. Equation of state ξ k is the position change matrix, the length of the time period between Ω moments, X k and u k It is known that k In input mode, X k The state changes to state X k+1 ;Observation equation:

[0063]

[0064] Where X k represents the state matrix at time k, represents the input value at time k, wk represents the Gaussian state noise at time k, and its corresponding covariance matrix is ​​Q k , v k represents the Gaussian measurement noise at time k, and its corresponding covariance matrix is ​​R k , z i is the measured pixel coordinate of the i-th landmark in the camera, is the measured pixel noise of the ith landmark.

[0065] Determine the driving state X at the next moment based on data analysis of the observed data k+1 The current driving state is jointly determined by different parameters of the current navigation data. As the driving environment changes, the correlation between different parameters changes with time. The current navigation data is used as input to obtain the predicted state analysis result at the next moment. The predicted state analysis result includes the predicted driving state and the corresponding prediction sequence.

[0066] The driving state is presented as a result, which is obtained by analyzing different navigation data. The difference between the corresponding navigation data when the driving state is the same is recorded as the state deviation. The observation sequence x={x i ,i=1,2,3,...,M},x iis the observed value, M is the number of observation items, and the difference set {Δ i =x i -y i , i = 1, 2, 3, ..., M}, the two observation sequences to be compared are recorded as two phase points in the M-dimensional phase space, x and y = {x i , i = 1, 2, 3, ..., M} is recorded as C(r), and the correlation function C(r) is calculated

[0067]

[0068] Where r is the given critical distance, θ is the unit function, C(r) is the number of phase points in the phase space whose distance is less than r, and 2 The proportion of the pairing in the data is analyzed by the correlation function to obtain the correlation dimension d, Md is the deviation dimension, The state deviation is then judged, and the model optimization module optimizes the adaptive combined navigation algorithm according to the judgment result.

[0069]

[0070] When the correlation function C(r) is in r→0, it has the following relationship with r:

[0071]

[0072] d is the correlation dimension. As the value of r changes, the correlation dimension of the m0-dimensional phase space is obtained.

[0073]

[0074] The model optimization module uses different adaptive combined navigation algorithms to perform adaptive combined analysis on the prediction sequence to obtain corresponding adaptive combined navigation analysis results. The model optimization module analyzes the adaptability between different models and observation data to obtain an adaptability vector P, and then constructs a transformation relationship between data changes and models through the vector P, thereby selecting the optimal model parameters, and then selecting the optimal adaptive combined navigation analysis algorithm based on the optimal model parameters.

[0075] When a driving state that has not appeared in the state pool appears, the optimal adaptive factor of the predicted driving state needs to be determined, and the similarity between the predicted driving state and the historical driving state in the state pool is analyzed. Then, the observation items related to the predicted sequence in the state pool are extracted with the similarity as the proportional parameter. The extracted observation items are combined with the predicted sequence to obtain a new adaptive observation sequence. The adaptive combined navigation algorithm in the state pool is then used to perform adaptive analysis on the adaptive observation sequence respectively. The adaptive factors corresponding to different algorithms are mapped to the node space to obtain a discrete node graph.

[0076] Different adaptive combined navigation algorithms in the state pool analyze the observation sequence of historical navigation data to obtain the corresponding adaptive factor graph, and analyze the discrete node graph according to the new adaptive observation sequence. The adaptive factor graph is then combined with the discrete node graph, and the adaptive factor graph is range-checked through the combination. Global search optimization is performed on all nodes to select the optimal adaptive factor.

[0077] The purpose of building a joint tree is to realize message passing in the joint tree and find the conditional probability of a random variable. After the joint tree is formed, the conditional probability table in the Bayesian network is converted into the joint tree, and the index parameters of all nodes are calculated using the formula By initializing each clique node and obtaining a joint tree that satisfies global consistency through message passing, we can find the probability distribution of any random variable in the original Bayesian network, select any clique node containing the random variable, and find the probability distribution through marginalization.

[0078] The relationship between the factor graph probability model and the optimal navigation solution, including the analysis of the sensor error model and factor nodes, and the derivation of the optimization algorithm based on the factor graph, provides a theoretical basis for the subsequent multi-sensor information fusion algorithm. At the same time, the plug-and-play function of the fusion navigation system is realized by utilizing the separability of the factor nodes. Then, the inertial navigation system and the integrated navigation system are analyzed and studied, the Kalman filter principle is analyzed using the Bayesian inference method, and the error model of the inertial navigation is parsed. The linear state equation and observation equation of the multi-sensor navigation system are arranged, and the error model function based on the factor graph is constructed.

[0079] The discrete node graph is combined with the adaptive factor graph. The specific process is as follows:

[0080] Step 1: construct a Moral graph, insert different nodes in the discrete node graph into the adaptive factor graph, connect them with the adaptive factor graph according to the distribution position, connect the parent nodes of the same node in the adaptive factor graph to each other, and remove the truncated head of each connection;

[0081] Step 2: Triangulate the graph. For a ring containing 4 or more nodes, add an undirected edge to connect two non-adjacent nodes in the ring to complete the triangulation of the Moral graph.

[0082] Step 3: Distinguish the clique nodes and determine the unity points in the triangulated graph. Each unity point is a subgraph of the undirected graph.

[0083] Step 4: Establish a union tree. The established union tree must include all cluster nodes, and the intersection is a subgraph of the undirected graph that connects the two cluster nodes.

[0084] According to the distribution probability of the joint nodes, the nodes in the discrete node graph and the adaptive factor graph are fused to obtain new adaptive factor distribution nodes. The adaptability of different adaptive factors is evaluated, and then the particle swarm optimization method is used to optimize the node relationship of the adaptive factor. The adaptive factor with the optimal position is searched globally, the adaptive factor is optimized, and the optimal adaptive factor is selected.

[0085] The navigation data collected by the data acquisition module includes inertial navigation data, vehicle status data and driving environment data. The vehicle's adaptive combined navigation algorithm performs adaptive analysis based on the navigation data collected at different times, and builds a vehicle status pool based on the data corresponding to the vehicle's historical driving status during the adaptive analysis process. The status pool includes all driving statuses that have occurred.

[0086] When the present invention is used specifically, the system includes a data acquisition module, a data storage module, a state analysis module, a model optimization module, a factor optimization module, a filtering analysis module, and a combination analysis module; the data acquisition module obtains navigation data during vehicle operation through vehicle sensors, pre-processes the navigation data, and stores it in the data storage module; the state analysis module then establishes a state prediction analysis model based on the navigation data, predicts and analyzes the vehicle's operating state at the next moment, and obtains the vehicle's predicted state analysis result, constructs a vehicle state pool based on the state analysis result corresponding to the vehicle's actual state at a historical moment, and judges the predicted driving state. When the predicted driving state appears in the historical state pool, According to the differences between different observation sequences under the same driving state, the optimal adaptive combined navigation analysis algorithm is selected to optimize the model to achieve accurate analysis of the adaptive process at the next moment. When facing a complex formal environment, the predicted driving state is difficult to determine. The model optimization module optimizes the driving state that has appeared in the state pool. The same driving state means that the overall driving state of the vehicle is the same, which is a result state. Maintaining the same driving state requires the joint action of different observation data and sub-navigation systems. There are differences in the data collected by sensors of different vehicles at different times, but the driving state may be the same, so the corresponding adaptive combined navigation analysis algorithms are also different. Different adaptive The influence process of factors on parameter estimation is different. When the driving state is the same, the model optimization module of the present invention optimizes and analyzes the adaptive combined navigation model, which greatly improves the accuracy of the adaptive combined navigation. When a new driving state that is not in the state pool appears, it means that new data that has not been monitored has appeared and has an impact on the driving state. The number of observation items in the prediction sequence is less than the observation items applied to the driving state image. The internal factor optimization module of the present invention uses similarity analysis to retrieve other observation items and combine them with the prediction sequence, thereby expanding the analyzed data items. When the driving state changes, the change of the observation sequence, the state prediction analysis model is an analysis of the state of the next moment based on the navigation data at the current moment. After analyzing the process In the process, through the combination of discrete node graph and adaptive factor graph, the combined adaptive factor makes the selection range of adaptive factor larger, the adaptive factor is optimized and selected, and the adaptive combined navigation analysis algorithm is constructed according to the optimized adaptive factor, which solves the problem of adaptive factor selection when abnormal state occurs and improves the adaptability to abnormal state. The filtering analysis module constructs the corresponding filtering analysis model according to the optimal adaptive factor, and then uses the filtering analysis model to perform filtering analysis on the adaptive combined navigation process at the next moment. In the filtering analysis process, the adaptive factor is used to optimize the estimation process; the combined analysis module performs adaptive combined navigation control on the combined navigation according to the error verification result of the filtering analysis module.

[0087] The above is a further detailed description of the present invention in combination with specific embodiments, but it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

Claims

1. An adaptive integrated navigation system applied to a vehicle, characterized in that: It includes data acquisition module, data storage module, state analysis module, model optimization module, factor optimization module, filter analysis module and combination analysis module; The data acquisition module acquires navigation data during vehicle operation through vehicle sensors, pre-processes the navigation data, and stores it in the data storage module; The state analysis module establishes a state prediction analysis model according to the navigation data, performs prediction analysis on the running state of the vehicle at the next moment, obtains the predicted state analysis result of the vehicle, and constructs the state pool of the vehicle according to the state analysis result corresponding to the actual state of the vehicle at the historical moment; Calculate the correlation between the predicted state analysis result and the state analysis result in the state pool, determine whether a driving state that is not in the state pool appears, when the predicted driving state appears in the state pool, calculate the state deviation between the driving state in the predicted state analysis result and the driving state at the current moment, when the state deviation exceeds the determination threshold, the predicted driving state and the optimal adaptive combined navigation algorithm corresponding to the current driving state are different, the model optimization module calls the optimal adaptive combined navigation algorithm corresponding to the predicted driving state, and performs adaptive combined navigation optimization according to the optimal adaptive combined navigation algorithm; When the predicted driving state is a driving state that has not appeared in the state pool, the similarity between the predicted driving state and the historical driving state in the state pool is calculated, and the similarity relative to different historical driving states is mapped onto a discrete node graph; The factor optimization module analyzes the adaptive factor graph corresponding to the adaptive combined navigation algorithm in the state pool in combination with the predicted driving state, combines the discrete node graph with the adaptive factor graph, optimizes and analyzes the node relationship of the adaptive factor, and selects the optimal adaptive factor; Different adaptive factors correspond to different measurement covariance matrices. After the optimal adaptive factor is selected, the filtering analysis module constructs the corresponding filtering analysis model based on the selected optimal adaptive factor, and then uses the filtering analysis model to perform filtering analysis on the adaptive combined navigation process at the next moment. In the filtering analysis process, the adaptive factor is used to optimize the estimation process. The combined analysis module performs adaptive combined navigation control on the combined navigation according to the error check result of the filtering analysis module.

2. The adaptive integrated navigation system for vehicles according to claim 1, characterized in that: The driving state is presented as a result, which is obtained by analyzing different navigation data. The difference between the corresponding navigation data when the driving state is the same is recorded as the state deviation. The observation sequence x={x i ,i=1,2,3,...,M},x i is the observed value, M is the number of observation items, and the difference set {Δ i =x i -y i , i=1,2,3,...,M}, the two observation sequences to be compared are recorded as two phase points in the M-dimensional phase space, x and y={x i , i=1,2,3,...,M} is recorded as C(r), and the correlation function C(r) is calculated Where r is the given critical distance, θ is the unit function, C(r) is the number of phase points in the phase space whose distance is less than r, and 2 The proportion of the pairing in the data is analyzed by the correlation function to obtain the correlation dimension d, Md is the deviation dimension, The state deviation is then judged, and the model optimization module optimizes the adaptive combined navigation algorithm according to the judgment result.

3. The adaptive integrated navigation system for vehicles according to claim 1, characterized in that: The model optimization module uses different adaptive combined navigation algorithms to perform adaptive combined analysis on the prediction sequence to obtain corresponding adaptive combined navigation analysis results. The model optimization module analyzes the adaptability between different models and observation data to obtain an adaptability vector P, and then constructs a transformation relationship between data changes and models through the vector P, thereby selecting the optimal model parameters, and then selecting the optimal adaptive combined navigation analysis algorithm based on the optimal model parameters.

4. The adaptive integrated navigation system for vehicles according to claim 1, characterized in that: When a driving state that has not appeared in the state pool appears, the optimal adaptive factor of the predicted driving state needs to be determined, and the similarity between the predicted driving state and the historical driving state in the state pool is analyzed. Then, the observation items related to the predicted sequence in the state pool are extracted with the similarity as the proportional parameter. The extracted observation items are combined with the predicted sequence to obtain a new adaptive observation sequence. The adaptive combined navigation algorithm in the state pool is then used to perform adaptive analysis on the adaptive observation sequence respectively. The adaptive factors corresponding to different algorithms are mapped to the node space to obtain a discrete node graph.

5. The adaptive integrated navigation system for vehicles according to claim 1, characterized in that: Different adaptive combined navigation algorithms in the state pool analyze the observation sequence of historical navigation data to obtain the corresponding adaptive factor graph, and analyze the discrete node graph according to the new adaptive observation sequence. The adaptive factor graph is then combined with the discrete node graph, and the adaptive factor graph is range-checked through the combination. Global search optimization is performed on all nodes to select the optimal adaptive factor.

6. The adaptive integrated navigation system for vehicles according to claim 5, characterized in that: The discrete node graph is combined with the adaptive factor graph. The specific process is as follows: Step 1: construct a Moral graph, insert different nodes in the discrete node graph into the adaptive factor graph, connect them in the adaptive factor graph according to the distribution position, connect the parent nodes of the same node in the adaptive factor graph to each other, and remove the truncated head of each connection; Step 2: Triangulate the graph. For a ring containing 4 or more nodes, add an undirected edge to connect two non-adjacent nodes in the ring to complete the triangulation of the Moral graph. Step 3: Distinguish the clique nodes and determine the unity points in the triangulated graph. Each unity point is a subgraph of the undirected graph. Step 4: Establish a union tree. The established union tree must include all cluster nodes, and the intersection is a subgraph of the undirected graph that connects the two cluster nodes.

7. The adaptive integrated navigation system for vehicles according to claim 1, characterized in that: According to the distribution probability of the joint nodes, the nodes in the discrete node graph and the adaptive factor graph are fused to obtain new adaptive factor distribution nodes. The adaptability of different adaptive factors is evaluated, and then the particle swarm optimization method is used to optimize the node relationship of the adaptive factor. The adaptive factor with the optimal position is searched globally, the adaptive factor is optimized, and the optimal adaptive factor is selected.

8. The adaptive integrated navigation system for vehicles according to claim 7, characterized in that: The navigation data collected by the data acquisition module includes inertial navigation data, vehicle status data and driving environment data. The vehicle's adaptive combined navigation algorithm performs adaptive analysis based on the navigation data collected at different times, and builds a vehicle status pool based on the data corresponding to the vehicle's historical driving status during the adaptive analysis process. The status pool includes all driving statuses that have occurred.

Images