A vehicle layered multi-module fusion precise positioning method under satellite system denial

Abstract

This solution discloses a hierarchical multi-module fusion precise positioning method for vehicles in situations where the global navigation satellite system is denied. It is mainly aimed at achieving precise vehicle positioning in special driving scenarios where the global navigation satellite system is denied, such as underground spaces, long tunnels in mountains, and the lower levels of three-dimensional transportation hubs. By utilizing a modular framework at different levels, the scalability and maintainability of the designed positioning method are enhanced. By fully mobilizing multi-source heterogeneous positioning information and utilizing a vehicle-road collaborative approach, precise vehicle self-positioning is achieved. This solves the problem of vehicles being unable to accurately perceive their own precise position and posture in driving situations where the global satellite navigation system is denied, as well as the "corridor effect" in long sections of roads with consistent features. By constructing a hierarchical, multi-module fusion precise positioning framework, heterogeneous positioning information flows are integrated and utilized to the greatest extent possible, eliminating redundancy and conflict between information flows from different modules.

Description

Technical Field

[0001] The present invention belongs to the technical field of high-precision fusion positioning of intelligent vehicles, and in particular relates to a vehicle layered multi-module fusion precise positioning method under global navigation satellite system denial conditions. Background Art

[0002] With technological advancements, vehicles equipped with assisted driving features are gradually becoming part of everyday life. Intelligent driving not only improves daily commuting safety but also provides drivers with a more relaxed and comfortable driving experience. Currently, mainstream assisted driving technology in the industry focuses on Level 2-3 autonomous driving, with continued advancement towards Level 4 and higher autonomous driving technologies. As intelligent driving technology continues to evolve, accurate and real-time vehicle location acquisition is crucial for ensuring the stable and reliable operation of intelligent driving systems, necessitating continued in-depth research and advancement.

[0003] Existing research shows that if a vehicle's positioning information is lost or drifts, autonomous driving planning, decision-making, and control execution will be affected to varying degrees. This is especially true for highly intelligent vehicles, where the autonomous driving level will regress to a lower level of assisted driving, requiring additional driver intervention. This can significantly reduce driver trust in intelligent driving technology. Furthermore, this temporary driver intervention can cause unexpected directional tremors in the vehicle, further compromising vehicle safety.

[0004] For the current mainstream L2-L3 assisted driving technologies, commonly used vehicle positioning technologies are based on the Global Navigation Satellite System (GNSS) and Real-Time Kinematic (RTK), combined with Inertial Navigation Systems (INS), Light Detection and Ranging (LiDAR), and cameras to achieve precise vehicle positioning. However, in actual driving, GNSS denial occurs in specific vehicle driving scenarios, such as underground spaces (mines, air defense facilities, and parking lots), long mountain tunnels, and the lower levels of multi-level transportation hubs. In these situations, the intelligent driving system cannot obtain navigation satellite signals, and the vehicle positioning system temporarily fails. The current conventional solution is to use the inertial navigation system to record the initial position and velocity at the moment of GNSS denial for positioning. However, due to the integration error of the inertial navigation system, it cannot be used for high-precision vehicle positioning over a long period of time. Therefore, a high-precision vehicle positioning method for GNSS denial is needed to ensure accurate vehicle positioning in all driving scenarios. Summary of the Invention

[0005] The purpose of the present invention is to address the above-mentioned problems and provide a vehicle layered multi-module fusion precise positioning method under satellite system denial conditions. The specific solution is as follows:

[0006] A vehicle layered multi-module fusion precise positioning method under satellite system denial conditions, comprising a vehicle position information perception layer, a time synchronization layer, and a fusion calculation layer connected in sequence;

[0007] The vehicle position information perception layer includes a roadbed positioning module, a detection positioning module, and an inertial positioning module, which are respectively used to obtain roadbed positioning information, detection positioning information, and inertial positioning information;

[0008] The time synchronization layer is used to obtain the standard time from the timing center and unify the timestamps of the roadbed positioning information, detection positioning information and inertial positioning information sent by the vehicle position information perception layer;

[0009] The fusion computing layer includes a fusion computing module and a collaborative request module;

[0010] The collaborative request module is used to provide weighted parameters for the fusion calculation layer according to different scenarios;

[0011] The fusion calculation module is used to perform fusion calculation on the vehicle position information perception layer data after time synchronization based on weighted parameters and output positioning information.

[0012] In the above-mentioned vehicle layered multi-module fusion precise positioning method under satellite system denial conditions, the roadbed positioning module includes a UWB base station, and the roadbed positioning information is obtained through the UWB base station;

[0013] The UWB base station is connected to the timing center via optical fiber transmission standards, and the time synchronization layer is connected to the timing center via the UWB base station to obtain standard time from the timing center. The connection referred to in this solution can refer to a wired connection or a wireless connection depending on the specific situation.

[0014] In the above-mentioned vehicle layered multi-module fusion precise positioning method under satellite system denial conditions, the roadbed positioning module also includes roadside markers to provide feature identification auxiliary positioning for the detection and positioning module, which is mainly used to solve the corridor effect of radar and images.

[0015] In the above-mentioned vehicle layered multi-module fusion precise positioning method under satellite system denial conditions, the detection and positioning module obtains a point cloud map by radar and combines it with the real-time driving image of the camera for fusion and collaborative positioning.

[0016] In the above-mentioned vehicle layered multi-module fusion precise positioning method under satellite system denial conditions, the detection and positioning module includes a 3D lidar, a millimeter-wave radar, a binocular camera, and an infrared camera. The four work together to perceive the surrounding environment. The acquired perception data is matched with a point cloud through a three-dimensional normal distribution transformation. After the distortion of the collected perception data is corrected, it is transmitted to the time synchronization layer.

[0017] In the above-mentioned vehicle layered multi-module fusion precise positioning method under satellite system denial conditions, the inertial positioning module includes an inertial measurement unit and an external wheel speedometer. The data of the inertial measurement unit is dominant, and the external wheel speedometer is used to calibrate the inertial positioning information through matrix transformation and linear interpolation.

[0018] In the above-mentioned vehicle layered multi-module fusion precise positioning method under satellite system denial conditions, the collaborative request module matches the image data of the detection and positioning module and the position information of the roadbed positioning module with the existing scenario rule library, and uses fuzzy reasoning technology based on the weighted rule library and the matched scenario to obtain weighted parameters, and provides the weighted parameters to the fusion calculation module;

[0019] The fusion calculation module determines the dominant positioning module and the auxiliary positioning module through weighted parameters, obtains the preliminary positioning position of the vehicle according to the dominant positioning module, obtains the auxiliary positioning position of the vehicle according to the auxiliary positioning module, judges the deviation between the auxiliary positioning position and the preliminary positioning position, and determines whether to perform weighted correction according to the weighted parameter of the deviation size. If correction is required, the position information is output after correction according to the correction rule. Otherwise, the position information of the dominant positioning module is directly output as the vehicle position information of the system.

[0020] In the above-mentioned vehicle layered multi-module fusion precise positioning method under satellite system denial conditions, the fusion calculation layer is also connected to the vehicle position information perception layer;

[0021] The collaborative request module is also used to coordinate and request the data transmission volume of different modules in the vehicle position information perception layer according to different scenarios and feed it back to the vehicle position information perception layer.

[0022] In the above-mentioned vehicle layered multi-module fusion precise positioning method under satellite system denial conditions, the collaborative request module simultaneously feeds back the weighted parameters to the vehicle position information perception layer;

[0023] After receiving the weighted parameters from the collaboration request layer, the vehicle position information perception layer adjusts the proportion of information collection according to the weighted parameters;

[0024] Furthermore, the time information transmitted by the roadbed UWB is not affected by the weight adjustment.

[0025] In the above-mentioned vehicle layered multi-module fusion precise positioning method under satellite system denial conditions, the roadbed positioning information, detection positioning information and inertial positioning information are all filtered through the adaptive volumetric Kalman filter algorithm and then transmitted to the time synchronization module.

[0026] The advantages of the present invention are:

[0027] A multi-module fusion precise positioning method at the multi-functional level is proposed, mainly for situations where the global navigation satellite system is denied in special driving scenarios such as underground spaces (mines, air defense facilities, parking lots), long tunnels in mountains, and the lower levels of three-dimensional transportation hubs, to achieve precise vehicle body positioning.

[0028] The use of modular frameworks at different levels enhances the scalability and maintainability of the designed positioning method. By fully mobilizing multi-source heterogeneous positioning information and utilizing vehicle-road collaboration, accurate vehicle body self-positioning is achieved.

[0029] This solves the problem of vehicles being unable to accurately perceive their precise position and posture in situations where global satellite navigation systems are denied, as well as the "corridor effect" on long stretches of roads with consistent features. By building a hierarchical, multi-module integrated precision positioning framework, it maximizes the integration and utilization of heterogeneous positioning information flows, eliminating redundancy and conflicts between information flows from different modules.

[0030] By differentiating perception and positioning weights, adjusting the proportion of information flow, reducing the amount of redundant information transmission, and alleviating the fusion computing overhead, it provides a reliable fusion perception and positioning solution for the development of autonomous driving and expands the applicable scenarios. Its hierarchical and modular framework design reduces the cost of maintenance and updates, and provides reliable technical support for the further development of autonomous driving. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 Schematic diagram of the implementation structure of the vehicle layered multi-module fusion precise positioning method under satellite system denial conditions of the present invention;

[0032] Figure 2 The present invention is a flow chart of the vehicle layered multi-module fusion precise positioning method under satellite system denial conditions. DETAILED DESCRIPTION

[0033] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0034] In special driving scenarios, since the global navigation satellite system is in a denied state and satellite positioning information cannot be obtained, in order to ensure the full operation of the intelligent driving system, it is necessary to use multi-source heterogeneous positioning information fusion analysis to form accurate positioning data. This solution provides a vehicle layered multi-module fusion precise positioning method with strong environmental adaptability, high positioning accuracy, and good scenario robustness for the intelligent driving system in the case of global navigation satellite system denial through a layered modular framework structure. This is achieved by establishing a fusion precise positioning architecture that collaborates between each layer and each module, specifically including:

[0035] First, establishing a fusion and precise positioning architecture that collaborates between each layer and each module is the basis for achieving the fusion and precise positioning of vehicles in this solution. To this end, it is necessary to consider the goals and scope of use of each layer and design Figure 1 The integrated precise positioning architecture shown in the figure. The vehicle position information perception layer constantly calibrates the vehicle position when GNSS is operating normally and initializes the vehicle position when GNSS enters a denied state. The collected information from each positioning module in the vehicle position information perception layer is initially filtered using an adaptive cubature Kalman filter algorithm to reduce noise and interference. This data is aggregated and passed to the time synchronization layer. The subgrade UWB in the subgrade positioning module also synchronizes with the timing center via optical fiber to obtain standard time.

[0036] Secondly, the time synchronization module in the time synchronization layer communicates with the roadbed UWB in the roadbed positioning module using the IEEE 1588PTP hardware standard to obtain time information and unify the timestamps of the vehicle location information perception layer data. After completing the timestamp unification of the vehicle location information data, the data information is passed to the fusion computing layer.

[0037] Finally, the collaborative request module matches the environmental identification data against the existing environmental rule base and utilizes fuzzy reasoning technology to derive weighted parameters for each positioning data point. These weighted parameters are then sent simultaneously to the fusion calculation module and the vehicle position information perception layer. The vehicle position information perception layer adjusts the information transmission ratio based on the weighted parameters, selectively reducing the amount of information transmitted, alleviating the fusion calculation pressure, and achieving precise vehicle position fusion positioning using simplified data. The fusion calculation module determines the dominant positioning module based on the weighted parameters and performs fusion positioning using a combination of primary and secondary positioning and weighted difference fusion. This resolves conflicts in heterogeneous vehicle positioning data and produces vehicle positioning data that is dominated by the primary positioning data and calibrated with the auxiliary positioning data.

[0038] Specifically, if Figure 1 and Figure 2 As shown, the vehicle layered multi-module fusion precise positioning system provided by this solution under satellite system denial conditions includes a vehicle position information perception layer, a time synchronization layer and a fusion calculation layer; the vehicle body position information acquisition modules in the vehicle position information perception layer are distributed in parallel and connected to the time synchronization layer; the time synchronization layer is connected to the vehicle position information perception layer and the fusion calculation layer; and the fusion calculation layer is connected to the time synchronization layer.

[0039] The vehicle body position information acquisition module in the vehicle position information perception layer includes an inertial positioning module, a detection positioning module and a roadbed positioning module, which are used to obtain roadbed positioning information, detection positioning information and inertial positioning information respectively. The fusion calculation layer will obtain three vehicle position information obtained in different ways: roadbed positioning information, detection positioning information and inertial positioning information.

[0040] Specifically, the inertial positioning module includes an inertial measurement unit (IMU) and an external wheel speedometer, wherein the IMU is continuously calibrated when the GNSS is in a normal state, keeps recording the position and posture of the vehicle, and continuously records changes in the vehicle position information after the vehicle enters a GNSS denial state. At the same time, the external vehicle wheel speedometer calibrates the vehicle position information recorded by the IMU. Taking the data of the inertial measurement unit as the main body, the IMU module measures the data and obtains the vehicle's motion state and body posture after Kalman filtering, and then obtains the actual position of the body in the form of integration. Since there is an integration error in the integration form, the external wheel speedometer is further used here to calibrate the inertial positioning information in the form of matrix transformation and linear interpolation to obtain inertial positioning information, and transmit the inertial positioning information to the time synchronization layer. The correction method through matrix transformation and linear interpolation is a relatively mature existing technology, and the process will not be described in detail here.

[0041] Specifically, the detection and positioning module uses radar to acquire a point cloud image, combined with real-time driving images from cameras for fusion and collaborative positioning. Specifically, it includes 3D lidar, millimeter-wave radar, binocular cameras, and infrared cameras. The radar component is used to acquire the environmental point cloud, namely the 3D lidar and millimeter-wave radar directly obtain radar point cloud images; the camera component, namely the binocular cameras and infrared cameras, obtain image information, which captures image features around the vehicle.

[0042] The four sensors work together to perceive the surrounding environment, and the acquired sensor data is matched to a point cloud using a 3D normal distribution transformation. Due to the differences in the installation locations of different sensors on the vehicle, the acquired sensor data is first matched to a point cloud using a 3D normal distribution transformation. This distortion-corrected data is then used to generate detection and positioning information, which is then transmitted to the time synchronization layer.

[0043] Specifically, the roadbed positioning module includes roadbed UWB and roadside markers. The roadbed UWB part communicates with the vehicle-mounted UWB positioning tag card. When the vehicle passes through the roadbed UWB coverage area, the vehicle's position is calculated by the Time Difference of Arrival (TDOA) method, and the roadbed positioning information is transmitted to the time synchronization layer. The roadside markers represent the corresponding positions of the road sections and provide perception markers for the detection and positioning module. Their characteristics are unified and change regularly according to the driving distance and position. Its function is to eliminate the "corridor effect" that appears on the roadside of long sections of roads with consistent characteristics in the detection and positioning module, to achieve precise positioning, and to effectively avoid the influence of the "corridor effect" of the corridor sections of the detection and positioning module. The roadbed UWB includes a roadbed optical fiber, which is connected to the timing center for communication through the optical fiber transmission standard to obtain time information.

[0044] Due to the comprehensiveness of the construction, there will generally not be a situation where there is only one kind of positioning information. If there is only one kind of situation, temporary positioning will be carried out by relying on a single positioning method. This is a special case. After obtaining new positioning information, a process is adopted to determine the positioning error → make corrections → re-determine the fusion dominant → fusion positioning method to achieve the recovery of positioning information.

[0045] Specifically, the time synchronization layer includes a time synchronization module, which is connected to the timing center through the roadbed UWB to obtain standard time from the timing center, and unifies the timestamps of the roadbed positioning information, detection positioning information and inertial positioning information based on the standard time.

[0046] The time synchronization layer includes a time synchronization module. The time synchronization module obtains the standard time of the timing center through the roadbed UWB based on the IEEE 1588PTP hardware standard. After obtaining the standard time, it uses Lanczos resampling to unify the frequency of vehicle position information collected by different positioning modules, thereby obtaining vehicle position perception data with consistent time synchronization and sampling frequency.

[0047] The fusion computing layer consists of a vehicle-side fusion computing module and a collaborative request module. The fusion computing layer transmits location information to the fusion computing module and environmental feature data to the collaborative request module. The location information primarily consists of three types of vehicle position information, namely, vehicle position information based on the inertial positioning module, the detection positioning module, and the roadbed positioning module. The environmental feature data primarily consists of image data captured by the camera and the specific location acquired via the roadbed UWB. The location of the roadbed UWB and its context are pre-stored in the system's database. Therefore, the specific location of the roadbed UWB can represent the scene context, and combining this with the UWB location improves scene matching accuracy. The collaborative request module determines the current scene based on the environmental feature data and then matches it against a pre-prepared scene rule base, which primarily includes underground spaces (mines, air defense facilities, parking lots), long mountain tunnels, and the lower levels of roads in three-dimensional transportation hubs. Corresponding features extracted from standard scene photos under different weather conditions are categorized and labeled as scenes A, B, C, etc., for weather conditions 1, 2, 3, and so on. Fast R-CNN matches the feature values ​​of standard scene photos in the scene rule library for identification. The result is the confidence level of the current clip's match to the standard scene. The matched scene may not be a 100% match. For example, the matching degrees of the current scene with three scenes in the library, A, B, and C, are 60%, 30%, and 10%, respectively. Combining the matching degrees of the three standard scenes, the weighted rule library uses a fuzzy subset approach to assign a feasible weight. For example, fuzzy subsets with linguistic variables such as "large," "medium," and "small" are used for different standard scenes. Each fuzzy subset uses a membership function to indicate the degree to which an exact value on the basic domain belongs to that fuzzy subset. The final weighted value is obtained through a weighted summation. The weighted rule library also includes a pre-existing expert database of weighting parameters for different positioning modules corresponding to different scene matching situations.

[0048] The collaborative request module passes the weighted parameters to the vehicle-side fusion calculation module and returns them to the vehicle position information perception layer.

[0049] The vehicle's position information perception layer adjusts the weighting of information transmission based on weighted parameters, selectively reducing the amount of information transmitted. In many cases, after acquiring scene data, the scene remains unchanged for a certain period of time. In this case, the established dominant data will dominate the vehicle's positioning data, while the auxiliary data is only used to correct for any deviations. Considering the relationship between sampling frequency and data transmission volume, the sampling frequency of auxiliary data can be coordinated to be reduced, or the sampling frequency can be reduced directly. This reduces the total amount of auxiliary data transmitted, thereby achieving coordinated and differentiated data transmission in different scenarios.

[0050] The fusion calculation module first determines the dominant positioning module using weighted parameters. In this embodiment, the positioning module with the highest weighted parameter is designated as the dominant positioning module, and the positioning information of the dominant positioning module is used as the positioning feature data. For example, if the detection positioning module is determined as the dominant positioning module, the detection positioning information is used as the dominant positioning feature data. Generally, the weighting is: roadbed positioning module > detection positioning module > inertial positioning module. After initially determining the vehicle's position using the information from the dominant positioning module, the deviation between the auxiliary positioning module and the dominant positioning module is determined. Based on the magnitude of the deviation and the weight difference, a correction is determined as to whether correction is necessary. The specific rules are pre-defined by skilled personnel based on the specific situation. For example, a correction is required when the deviation is greater than 30 cm or the weight difference is less than 30%. If correction is not required, the positioning information of the dominant positioning module is directly output as the vehicle's position. If correction is required, it is performed according to the correction rules. This correction can be performed using linear interpolation to obtain the vehicle's position by applying a small correction coefficient to the vehicle body. The correction coefficient and correction rules can be pre-defined by the technician. Finally, the fused vehicle position information is output to the assisted driving module or the human-machine interface.

[0051] This solution constructs different simultaneous localization and mapping (SLAM) modules, and through the collaborative fusion and calibration of multiple modules, maximizes the integration and utilization of heterogeneous positioning information flows, eliminates the redundancy and conflict between information flows of different modules, and achieves high-precision vehicle body positioning. It has very important scientific significance and extremely broad engineering application value for breaking through the application scope of intelligent driving systems, improving the reliability and safety of intelligent driving systems, improving traffic efficiency in scenarios, and promoting the further application and promotion of autonomous driving technology.

[0052] The specific embodiments described in this embodiment are merely illustrative of the spirit of the present invention. Those skilled in the art may make various modifications, additions, or substitutions to the described specific embodiments without departing from the spirit of the present invention or exceeding the scope defined by the appended claims.

Claims

1. A vehicle layered multi-module fusion precise positioning method under satellite system denial conditions, characterized by: It includes a vehicle location information perception layer, a time synchronization layer, and a fusion calculation layer connected in sequence; The vehicle position information perception layer includes a roadbed positioning module, a detection positioning module, and an inertial positioning module, which are respectively used to obtain roadbed positioning information, detection positioning information, and inertial positioning information; The time synchronization layer is used to obtain the standard time from the timing center and unify the timestamps of the roadbed positioning information, detection positioning information and inertial positioning information sent by the vehicle position information perception layer; The fusion computing layer includes a fusion computing module and a collaborative request module; The collaborative request module is used to provide weighted parameters for the fusion calculation layer according to different scenarios; The fusion calculation module is used to perform fusion calculation on the time-synchronized vehicle position information perception layer data based on weighted parameters and output positioning information; The collaborative request module matches the image data of the detection and positioning module and the position information of the roadbed positioning module with the existing scene rule library, obtains weighted parameters based on the weighted rule library and the matched scenes using fuzzy reasoning technology, and provides the weighted parameters to the fusion calculation module; The fusion calculation module determines the dominant positioning module and the auxiliary positioning module through weighted parameters, obtains the preliminary positioning position of the vehicle according to the dominant positioning module, obtains the auxiliary positioning position of the vehicle according to the auxiliary positioning module, judges the deviation between the auxiliary positioning position and the preliminary positioning position, and determines whether to perform weighted correction according to the deviation size and weighted parameters. If correction is required, the position information is output after correction according to the correction rules. Otherwise, the position information of the dominant positioning module is directly output as the vehicle position information of the system.

2. The vehicle layered multi-module fusion precise positioning method under satellite system denial conditions according to claim 1 is characterized in that: The roadbed positioning module includes a UWB base station, and obtains roadbed positioning information through the UWB base station; The UWB base station is connected to the timing center via an optical fiber transmission standard, and the time synchronization layer is connected to the timing center via the UWB base station to obtain standard time from the timing center.

3. The vehicle layered multi-module fusion precise positioning method under satellite system denial conditions according to claim 2 is characterized in that: The roadbed positioning module also includes roadside markers to provide feature identification auxiliary positioning for the detection and positioning module.

4. The vehicle layered multi-module fusion precise positioning method under satellite system denial conditions according to claim 3 is characterized in that: The detection and positioning module obtains a point cloud image from the radar and performs fusion and collaborative positioning in combination with the real-time driving image from the camera.

5. The vehicle layered multi-module fusion precise positioning method under satellite system denial conditions according to claim 4 is characterized in that: The detection and positioning module includes a 3D laser radar, a millimeter-wave radar, a binocular camera, and an infrared camera. The four work together to perceive the surrounding environment. The acquired perception data is matched with a point cloud through a three-dimensional normal distribution transformation. After the distortion of the collected perception data is corrected, it is transmitted to the time synchronization layer.

6. The vehicle layered multi-module fusion precise positioning method under satellite system denial conditions according to claim 4 is characterized in that: The inertial positioning module includes an inertial measurement unit and an external wheel speedometer. The data of the inertial measurement unit is dominant, and the inertial positioning information is calibrated using the external wheel speedometer through matrix transformation and linear interpolation.

7. The vehicle layered multi-module fusion precise positioning method under satellite system denial conditions according to claim 1 is characterized in that: The fusion calculation layer is also connected to the vehicle position information perception layer; The collaborative request module is also used to coordinate and request the data transmission volume of different modules in the vehicle position information perception layer according to different scenarios and feed it back to the vehicle position information perception layer.

8. The vehicle layered multi-module fusion precise positioning method under satellite system denial conditions according to claim 7 is characterized in that: The collaborative request module also feeds back the weighted parameters to the vehicle position information perception layer; After receiving the weighted parameters from the collaboration request layer, the vehicle position information perception layer adjusts the proportion of information collection according to the weighted parameters; Furthermore, the time information transmitted by the roadbed positioning module is not affected by the weight adjustment.

9. The vehicle layered multi-module fusion precise positioning method under satellite system denial conditions according to claim 4 is characterized in that: The roadbed positioning information, detection positioning information and inertial positioning information are all filtered by the adaptive cubature Kalman filter algorithm and then transmitted to the time synchronization module.

Images