External multi-source fusion positioning method and system based on vehicle speed injection enhancement

Abstract

The invention discloses an external multi-source fusion positioning method and system based on vehicle speed injection enhancement, and the method is operated on an external calculation unit which is decoupled from a vehicle-mounted sensor and a plurality of heterogeneous operation systems, and comprises the steps: carrying out the global time synchronization of multi-source sensing data; executing a fusion algorithm, and performing priority scheduling and protocol conversion on a fusion positioning result; distributing the data streams to corresponding target systems through a cross-system compatible architecture; through a remote maintenance link, the collection quality of algorithm parameters or multi-source sensing data in an external calculation unit is continuously enhanced. According to the invention, a core fusion algorithm is stripped from a dedicated chip and is externally set as an independent computing service, so that the long-term performance and adaptation flexibility of the system are remarkably improved; the cross-system compatible architecture unifies underlying communication, protocol conversion and ecological adaptation, so that the multi-system adaptation and deployment cost is greatly reduced; based on a dynamic vehicle speed weight injection mechanism of GNSS signal quality, the positioning precision of a weak / non-GNSS scene is improved.

Description

Technical Field

[0001] This invention relates to the field of high-precision navigation and positioning technology, specifically to an external multi-source fusion positioning method and system based on vehicle speed injection enhancement. Background Technology

[0002] With the development of intelligent driving technology, the demand for high-precision, high-reliability, and all-scenario-usable positioning capabilities for vehicles is becoming increasingly urgent. Currently, mainstream vehicle positioning solutions mainly rely on the fusion of Global Navigation Satellite Systems (GNSS) and Inertial Measurement Units (IMUs). However, in complex urban environments (such as high-rise blocks, underground parking garages, and tunnels), GNSS signals are easily attenuated or even interrupted due to blockage and reflection, rendering positioning services relying solely on GNSS unusable. Furthermore, relying solely on IMUs for inertial estimation will generate significant cumulative errors (drift) in a short period. Existing implementation schemes have the following significant drawbacks: The core fusion algorithm is usually pre-programmed into the GNSS / IMU integrated navigation chip, making it impossible for the algorithm to be upgraded independently of the hardware. This makes it difficult to adapt to new vehicle models, new scenarios, or optimize algorithm parameters. Smart cockpits generally use multiple heterogeneous operating systems such as Linux, Android, and HarmonyOS. Traditional solutions require the separate development of data interfaces and communication modules for each operating system, resulting in long development cycles, high maintenance costs, and difficulty in ensuring the consistency of positioning data between systems. Existing solutions often use vehicle speed as an auxiliary observation with fixed weight. In critical scenarios such as weak GNSS signals, they fail to dynamically increase the fusion weight of vehicle speed data to fully compensate for the decrease in GNSS observation accuracy, resulting in large fluctuations in positioning output and inaccuracies that cannot meet the requirements of advanced driver assistance systems. Summary of the Invention

[0003] The purpose of this invention is to provide an external multi-source fusion positioning method and system based on vehicle speed injection enhancement, so as to solve the problems mentioned in the background art, such as the rigidity of existing algorithms, high cost of cross-system ecosystem adaptation, and poor performance in weak signal scenarios.

[0004] To achieve the above objectives, the present invention adopts the following technical solution: According to one aspect of the present invention, an external multi-source fusion localization method based on vehicle speed injection enhancement is provided, the method operating on an external computing unit decoupled from onboard sensors and multiple heterogeneous operating systems, comprising: Global time synchronization is performed on multi-source sensor data from the vehicle, wherein the multi-source sensor data includes at least inertial measurement data, satellite positioning data, and vehicle speed data; In the external computing unit, a fusion algorithm is executed to process the multi-source sensor data after global time synchronization and generate a fusion positioning result; the fusion algorithm dynamically adjusts the contribution weight of the vehicle speed data in the fusion calculation according to the real-time quality of the satellite positioning signal. The fused positioning results are prioritized and converted according to the real-time requirements and protocol requirements of the target system to form multiple formatted data streams. The multiple formatted data streams are distributed to the corresponding target systems using a cross-system compatible architecture. By remotely maintaining the link, the algorithm parameters in the external computing unit or the acquisition quality of the multi-source sensor data can be continuously enhanced.

[0005] Based on the aforementioned scheme, the global time synchronization includes establishing and taming a global reference clock source; based on the global reference clock source, marking the inertial measurement data, satellite positioning data, and vehicle speed data with a native timestamp of a unified time reference at the physical time of acquisition; and performing time axis alignment and data reorganization on the multi-source data streams carrying the native timestamps to generate a time-synchronized data sequence.

[0006] Based on the aforementioned scheme, the fusion algorithm determines the scene as a strong satellite positioning signal scene, a weak satellite positioning signal scene, or a scene without a satellite positioning signal based on the intensity parameter of the satellite positioning signal; according to the scene determination result, the observation noise covariance matrix or observation vector of the fusion algorithm is dynamically configured to adjust the fusion weight of the vehicle speed data relative to other data sources.

[0007] Based on the aforementioned scheme, when the scenario is determined to be a weak satellite positioning signal scenario or a scenario without a satellite positioning signal, the fusion weight of the vehicle speed data is increased, making the vehicle speed data the core observation source for suppressing the integral drift of inertial measurement data.

[0008] Based on the aforementioned scheme, the priority scheduling and protocol conversion includes configuring and distributing priorities for different target systems, wherein the priorities are determined based on the real-time level and functional safety criticality of the target systems; and calling the corresponding protocol conversion engine in descending order of the priorities to convert the fused positioning results into a specified data format that meets the requirements of each target system.

[0009] Based on the aforementioned scheme, the protocol conversion engine supports at least one or more of the following data formats for conversion: NMEA-0183 format, SOME / IP protocol format, Protocol Buffers encoding format, and vehicle CAN bus message format.

[0010] Based on the aforementioned scheme, distribution is carried out through the cross-system compatible architecture, including encapsulating the native communication mechanisms of different operating systems into a unified application programming interface through the underlying communication abstraction layer; and performing final encapsulation and permission adaptation in accordance with the target system's ecological specifications through the protocol and data compatibility layer before the formatted data stream is injected into the communication link.

[0011] Based on the aforementioned scheme, the cross-system compatible architecture also includes transmission quality monitoring and primary / backup link switching, which automatically switches to the backup communication link to ensure the continuity of data distribution when the quality of the primary distribution link is detected to be deteriorated.

[0012] Based on the aforementioned scheme, the remote maintenance link includes injecting auxiliary ephemeris data into the satellite positioning module through Assisted Global Navigation Satellite System technology to shorten its positioning startup time; and remotely sending and securely updating the fusion algorithm parameters or sensor calibration parameters to the external computing unit through over-the-air download technology.

[0013] According to another aspect of the present invention, an external multi-source fusion positioning system based on vehicle speed injection enhancement is provided. The system includes: a data acquisition module, an external computing unit, a communication interface module, and a remote maintenance interface. The data acquisition module is used to acquire multi-source sensor data from the vehicle, including interfaces for receiving satellite positioning data, inertial data, and vehicle speed data. The external computing unit is communicatively connected to the data acquisition module and is used to perform global time synchronization, fusion computing, and distribution scheduling. The communication interface module is used to establish communication with multiple heterogeneous target systems through a cross-system compatible architecture; The remote maintenance interface is used to connect to a remote maintenance link to continuously enhance the algorithm parameters in the external computing unit or the acquisition quality of the multi-source sensor data.

[0014] As can be seen from the above technical solutions, this invention has at least the following advantages and positive effects compared with existing technologies: By decoupling the core fusion algorithm from a dedicated chip and making it an independent computing service, the hardware constraint on the algorithm is broken, significantly improving the long-term performance and adaptability flexibility of the system. The cross-system compatible architecture unifies underlying communication, protocol conversion, and ecosystem adaptation, greatly reducing the cost of multi-system adaptation and deployment. Through a dynamic vehicle speed weight injection mechanism based on GNSS signal quality, in scenarios with weak or no satellite signals, vehicle speed data is automatically elevated to a core observation source, effectively suppressing IMU integral drift and improving positioning accuracy in weak / no GNSS scenarios. Through the AGNSS-OTA integrated enhancement closed loop, the system can remotely shorten the positioning cold start time and optimize algorithm parameters and sensor calibration parameters in a targeted manner, building remote intelligent maintenance capabilities. Most performance tuning and fault repair work can be completed in the cloud without requiring vehicles to return to the factory.

[0015] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit the invention. Attached Figure Description

[0016] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention. It is obvious that the drawings described below are merely some embodiments of the invention, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort. In the drawings: Figure 1 This is a schematic diagram of an external multi-source fusion localization method based on vehicle speed injection enhancement according to the present invention; Figure 2 This is a schematic diagram of an external multi-source fusion positioning system based on vehicle speed injection enhancement according to the present invention. Detailed Implementation

[0017] To more clearly illustrate the purpose, technical solutions, and advantages of the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein. On the contrary, these embodiments are provided so that the present invention will be more comprehensive and complete, and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0018] Furthermore, the described features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. Numerous specific details are provided in the following description to give a full understanding of embodiments of the invention. However, those skilled in the art will recognize that the technical solutions of the invention can be practiced without one or more of the specific details, or other methods, components, apparatuses, steps, etc., can be employed. In other instances, well-known methods, apparatuses, implementations, or operations are not shown or described in detail to avoid obscuring various aspects of the invention.

[0019] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0020] The flowcharts shown in the accompanying drawings are merely illustrative and do not necessarily include all content and operations / steps, nor do they necessarily have to be performed in the described order. For example, some operations / steps can be broken down, while others can be combined or partially combined; therefore, the actual execution order may change depending on the specific circumstances.

[0021] The present invention will now be described in detail with reference to specific embodiments.

[0022] Example 1

[0023] like Figure 1 As shown, this embodiment provides an external multi-source fusion positioning method based on vehicle speed injection enhancement. This method runs on an external computing unit. This unit is physically and logically decoupled from the vehicle's original sensor modules (such as GNSS / IMU) and the multiple heterogeneous operating systems (such as Linux, Android, HarmonyOS) installed in the vehicle. The external computing unit is an independent hardware module, which can be an edge computing box or a dedicated computing partition integrated into a domain controller. It is equipped with at least: an interface for connecting to the vehicle's sensor network and an interface for communicating with the various operating systems in the vehicle. The external computing unit has a built-in high-performance processor and sufficient memory, possessing the computing power to run complex fusion algorithms. It has an independent operating system or real-time kernel deployed on it, serving as a unified computing platform to support the algorithm services described in this invention. By adopting this external decoupled architecture, this invention achieves the decoupling of the core positioning algorithm from specific chips, thereby supporting independent iteration and upgrades of the algorithm. The specific steps of the method are as follows: S1: Perform global time synchronization on multi-source sensor data from the vehicle, wherein the multi-source sensor data includes at least inertial measurement data, satellite positioning data, and vehicle speed data.

[0024] This process unifies the time reference among multi-source sensor data from different physical interfaces of the vehicle, which have heterogeneous characteristics, to provide spatiotemporally consistent input data for subsequent fusion computing. First, a reference clock source is established and tamed. Based on the global reference clock source, the inertial measurement data, satellite positioning data, and vehicle speed data are marked with a native timestamp of a unified time reference at the physical time of acquisition. The multi-source data streams carrying the native timestamps are then time-axis aligned and data reorganized to generate a time-synchronized data sequence.

[0025] Specifically, a highly stable local crystal oscillator, such as a temperature-compensated crystal oscillator (TCXO), is used as the master clock source, with a frequency stability preferably not exceeding ±0.1 ppm. The pulse-per-second (1PPS) signal output from the Global Navigation Satellite System (GNSS) receiver is received as an external time reference. Through feedback control of a phase-locked loop (PLL) or a time interval counter, the phase and frequency deviations of the master clock source relative to the 1PPS signal are continuously measured and compensated, thereby generating a global reference clock that combines long-term stability (from GNSS) with short-term accuracy and continuity (from the local crystal oscillator). The frequency and phase information of this reference clock are distributed and synchronized to all data acquisition and processing nodes through the clock synchronization protocol of a Time-Sensitive Network (TSN).

[0026] Each sensor uses a local clock synchronized to the global reference clock to mark a high-precision native timestamp at the physical moment of data generation, including inertial measurement data, satellite positioning data, and dual-source vehicle speed data. Inertial measurement data, raw acceleration and angular velocity data read from a six-axis inertial measurement unit (IMU) at a fixed frequency (e.g., 50Hz or 100Hz), has a timestamp embedded based on the synchronized local clock at the moment of a hardware interrupt triggered by the IMU sensor data ready signal, or at the start of the serial communication for data reading. T IMU Satellite positioning data is used to extract the UTC time information of the message from the data frame output by the GNSS receiver. Using the global reference clock count value corresponding to the rising edge of the 1PPS signal as a reference, the UTC time is mapped to the global reference clock time domain through time interpolation to generate a timestamp. T GNSS The timestamp is then bound to the positioning and navigation information calculated within the current GNSS observation cycle. This positioning and navigation information includes spatial location (longitude, latitude, altitude), ground velocity vector, and parameters used to evaluate the current calculation quality, such as the horizontal accuracy factor (HDOP), the number of effective satellites, and the statistical values ​​of the signal-to-noise ratio (SNR) for each satellite. Dual-source vehicle speed data includes bus speed and pulse speed. For bus speed, a CAN controller supporting hardware timestamps automatically records the global reference clock value as a timestamp when the start bit of the speed message frame arrives. T CANThe vehicle speed value decoded from the message V CAN Binding; Pulse vehicle speed, the global reference clock count value when the wheel speed pulse edge is triggered is recorded by a hardware capture circuit to obtain a timestamp sequence. T PULSE Then the pulse vehicle speed is calculated. V PULSE .

[0027] The aforementioned multi-stream data with native timestamps undergoes time axis alignment and data reorganization (pre-processing before fusion) in an external computing unit to generate a time-synchronized data sequence. This data sequence is based on a unified virtual time axis. T base [ k The data frames are arranged in a manner that each frame contains at least: a high-precision time index, aligned angular velocity and acceleration data from the inertial measurement unit (IMU), aligned position and velocity data from the global navigation satellite system (GNSS), and processed velocity data from dual-source vehicle speeds.

[0028] Specifically, using a global reference clock as a reference, a virtual time series is established in memory that increments at a fixed period (e.g., 1 millisecond). T base [ k This serves as a unified virtual timeline. The native timestamp of each data unit ( T IMU , T GNSS , T CAN , T PULSE Mapped to the virtual timeline T base The closest point; in this process, the fixed delay parameters of each link obtained in the calibration phase are used to compensate for the timestamp, so that it more accurately represents the time of occurrence of the physical event; the fixed delay parameters of each link are obtained in the calibration phase, including sending test signals with known timestamps to the data acquisition module and receiving its return signals, and determining the fixed delay of the link by calculating the round-trip time or comparing the timestamps. Since the output frequencies of each data source are different, in T base [ k The upper part of the data source may lack actual sampling points; an interpolation algorithm is used to generate synthetic data points for the missing data sources. For example, for low-frequency GNSS location data, kinematic model interpolation can be performed based on the high-frequency motion information of the IMU data to generate synthetic data points corresponding to the IMU time. T base [ k Aligned virtual GNSS position observations; for dual-source vehicle speeds, inT base [ k [Time can be used] V CAN and V PULSE The optimal vehicle speed estimate is obtained by using linear interpolation or a timestamp-based weighted average. V veh [ k There are no restrictions on the interpolation algorithm; spline interpolation or filter-based prediction algorithms can also be used.

[0029] Preferably, consistency checks between data sources are performed periodically (e.g., comparing IMU integral speed with vehicle speed observations); if an anomaly is detected, the health status of the corresponding data source is updated (e.g., normal / degraded / failed). For data sources that fail temporarily, their last valid parameters are maintained and a virtual data stream is generated based on a reference clock, while their status is marked as "degraded".

[0030] Preferably, synchronization health monitoring and adaptive tuning are implemented; the maximum deviation between the timestamp of each data source and the reference clock is calculated and recorded. If the deviation exceeds a set threshold (e.g., 10μs) for N consecutive frames (e.g., 3 frames), a synchronization accuracy alarm is triggered, and this status is recorded in the unified log as part of the system health indicators. Output is... T base [ k Time-series arranged, multi-source aligned, standardized data frames; each frame contains: a time index. T base [ k Health status identifiers of each data source, aligned data fields from IMU, GNSS, vehicle speed, and other sources, and their corresponding data quality factors.

[0031] S2: In the external computing unit, a fusion algorithm is executed to process the multi-source sensor data after global time synchronization and generate a fusion positioning result; the fusion algorithm dynamically adjusts the contribution weight of the vehicle speed data in the fusion calculation according to the real-time quality of the satellite positioning signal.

[0032] In this embodiment, the fusion algorithm uses Extended Kalman Filter (EKF) as the core fusion framework. EKF is a recursive optimal estimation algorithm suitable for nonlinear systems. Its core lies in using a prediction-update loop to combine the system dynamics model (based on IMU state transitions) and the multi-source observation model (based on observation equations of GNSS, vehicle speed, etc.) to make the optimal estimate of the true state of the system in the sense of minimum mean square error. This algorithm framework has the ability to process asynchronous, multi-rate, and multi-source observation data, and can provide a quantitative description of the uncertainty of the estimation results (through the error covariance matrix).

[0033] Specifically, a state vector X is defined, which contains the following estimated states: three-dimensional position (such as longitude λ, latitude φ, altitude h), three-dimensional velocity (eastward velocity v E , northward velocity v N , upward velocity v U ), and the error states of the inertial measurement unit (IMU) (such as the biases of the accelerometer and gyroscope); the dimension of the state vector can be extended according to the accuracy requirements. The data quality factors attached to the GNSS data from step S1 are analyzed in real time, including but not limited to the horizontal dilution of precision (HDOP), the number of available satellites (N sat ), and the average signal-to-noise ratio (SNR avg ). Based on a preset threshold of the intensity parameter of the satellite positioning signal, the current environment is determined: a strong satellite positioning signal scenario, a weak satellite positioning signal scenario, and a no satellite positioning signal scenario. When the comprehensive index of the satellite positioning signal quality is excellent, it is determined as a strong satellite positioning signal scenario, and exemplary conditions are HDOP ≤ 2.0, N sat ≥ 6, and SNR avg ≥ 35 dB-Hz; when the satellite positioning signal is available but the accuracy significantly decreases, it is determined as a weak satellite positioning signal scenario, and exemplary conditions are insufficient number of available satellites (such as 4 ≤ N sat < 6) and accompanied by deterioration of the dilution of precision (such as HDOP > 2.0) or reduction of the signal-to-noise ratio (such as SNR avg < 35 dB-Hz); when the satellite positioning signal fails, it is determined as a no satellite positioning signal scenario, and exemplary conditions are too few available satellites (such as N sat < 4) or the GNSS receiver outputs an invalid flag.

[0034] Based on the real-time scene determination results, the observation noise covariance matrix R(k) or observation vector Z(k) of the fusion algorithm is dynamically configured to substantially adjust the fusion weight of vehicle speed data relative to other data sources. The core principle is that the higher the quality of the satellite positioning signal, the greater its weight in the observation update, achieved by setting a smaller variance in R(k). Conversely, when the quality of the satellite positioning signal decreases, the weight of vehicle speed observations is automatically increased to suppress drift caused by IMU inertial calculations. In scenarios with strong satellite positioning signals, high weights are assigned to GNSS position observations (i.e., their observation noise variance is set to be small), while lower auxiliary weights are assigned to vehicle speed observations. The fusion update uses high-precision GNSS data as the core to correct IMU recursion errors. In scenarios with weak satellite positioning signals, the weight of GNSS observations is significantly reduced (e.g., its noise variance in R(k) is increased), while the weights of the eastward and northward speed observations obtained from the preprocessed vehicle speed data are significantly increased, making them the main observation sources constraining IMU speed drift. In scenarios without satellite positioning signals, GNSS observations are removed from the observation vector or given infinite noise variance. The fusion algorithm uses vehicle speed as the external observation for dead reckoning. Periodic vehicle speed observations are used to correct the cumulative position and velocity errors generated by IMU dead reckoning.

[0035] Wherein, the observation vector Z(k) is the set of all external measurement data used by the fusion algorithm in the update step at time k; Z(k) is a column vector, each element of which corresponds to an independent observation value. In this embodiment, Z(k) dynamically contains observation information from different sensors; for example, latitude and longitude and altitude provided by the GNSS receiver; velocity components provided by the GNSS receiver; eastward and northward velocities after preprocessing the dual-source vehicle speed data and converting it to the navigation coordinate system; the observation vector is related to the system's state vector X(k) through the observation equation, in the form: Z(k) = H(k) * X(k) + V(k), where H(k) is the observation matrix and V(k) is the observation noise.

[0036] The observation noise covariance matrix R(k) is a key parameter matrix in the EKF framework used to quantitatively characterize the uncertainty or noise level of each observation in the observation vector Z(k). R(k) is modeled as a diagonal matrix or has off-diagonal elements to represent the correlation between observation noises; its diagonal elements... R ii ( k That is, the corresponding observed value. Z i ( k The noise variance of R(k) directly affects the calculation of the Kalman gain K(k) in the Kalman filter, thus determining the strength of the correction of the observation information to the state estimate. In Kalman filter theory, for an observation, its noise variance... Rii The smaller the value, the more accurate and reliable the observation, and the greater the weight assigned to it in the algorithm's calculations, and vice versa. Therefore, dynamically adjusting the contribution weights of each data source is directly equivalent to dynamically configuring the values ​​of the corresponding elements in the R(k) matrix in the algorithm implementation. If a data source is assigned a high weight, then its corresponding... R ii Set it to a small value; if a data source is assigned a low weight, then set its corresponding... R ii Set it to a large value; in extreme cases, ignore a data source: you can set its corresponding... R ii Setting it to a theoretically infinite value is mathematically equivalent to removing the observation from the observation vector.

[0037] Furthermore, multi-source data fusion calculations are performed; in each fusion cycle, the Extended Kalman Filter (EKF) executes a prediction-update loop, as follows: It should be noted that, in this description, state estimation and its error covariance The subscripts follow the standard notation of Kalman filtering: a represents the time corresponding to the state, and b represents the cutoff time of the observation data on which the estimate is based. For example, This represents the optimal estimate at time k-1 based on all information up to time k-1.

[0038] 1) State prediction: Utilizing the optimal state estimate from the previous fusion cycle and the incremental data (angular velocity ω) of the synchronized inertial measurement unit (IMU) during the current cycle. m acceleration a m The state prediction value at the current moment is obtained by recursively arranging the equations of inertial navigation mechanics. and its corresponding prediction error covariance matrix This process is characterized by a nonlinear state transition function. describe: ; ; Where uk-1 is the control input, i.e., the angular velocity and acceleration measurements of the IMU; Fk-1 is the state transition matrix, which is a function... exist The Jacobian matrix at point Qk-1 represents the propagation relationship of state error; Qk-1 is the process noise covariance matrix, representing the measurement noise of the IMU accelerometer and gyroscope, zero-bias instability and other sensor errors, as well as the errors introduced by the simplification of the inertial navigation mechanical model.

[0039] 2) Observation Update: The predicted state is corrected using multi-source observation data. Specifically, the result S is determined based on the real-time scene. k Dynamically select and assemble observation data to form the observation vector Z at the current moment. k ; ; in, , GNSS latitude and longitude , For GNSS eastward and northward speeds; , For the vehicle speed after preprocessing and vehicle heading angle The eastward and northward velocity components calculated (from the state vector or by IMU): Call the observation noise covariance matrix R that matches the current scene. k By configuring differentiated R for different data sources k The diagonal elements enable dynamic allocation of fusion weights across data sources in different scenarios. The Kalman gain matrix K is calculated to balance the weights between predicted and observed values. k : ; in, Let be the observation matrix, and be the observation function. The Jacobian matrix at the predicted state establishes a linearized relationship between the state space and the observation space. The observation function h(·) maps the state vector to the observation space; for example, for GNSS position observations, h(·) directly extracts the position component from the state vector; for observations based on vehicle speed V... veh The converted velocity observation, h(·), involves extracting the heading angle ψ from the state vector and performing trigonometric operations. Then, the actual observed value Z is used. k Compared with the observed estimates based on the predicted state The residuals (also known as innovations) between the predicted states are used to make optimal corrections. ; ; Based on the above calculations, the results are as follows: This represents the optimal state estimate at the current moment after multi-source data fusion, containing high-precision position, velocity, and attitude information. Error covariance matrix. This quantitatively reflects the uncertainty of the estimated value; this process is repeated cyclically to achieve continuous, adaptive, high-precision positioning output.

[0040] The final output state estimate of each fusion cycle This is converted into a standardized fusion positioning result; this result contains core navigation parameters extracted from the state vector, such as position in the geodetic coordinate system (longitude). λ ,latitude ,altitude h ) and the velocity component in the northeast-northeast coordinate system (eastward velocity) v E Northbound speed v N Horizontal speed v U Preferably, the fused positioning result further includes a confidence index characterizing the reliability of the current positioning result; this index is determined by analyzing the error covariance matrix. This is obtained through analytical calculation; for example, an output uncertainty metric, such as horizontal position accuracy, can be calculated based on... The covariance submatrix corresponding to the mid-position state is used to calculate the circular probability error (CEP) or error ellipse parameters (including the semi-major axis) in the horizontal direction. a semi-short shaft b and elliptical direction angle α The confidence level index, along with speed accuracy and the standard deviation of each speed component, serves as a quality label, output synchronously with the core navigation parameters. Downstream intelligent driving systems, navigation applications, or data recording units can use this quality label to assess the reliability of the positioning data and thus execute corresponding strategies.

[0041] Preferably, while outputting the fusion result, the fusion process itself is monitored, and the algorithm is optimized based on the monitoring information. The statistical properties of the innovation sequence (observation residuals) and the error covariance matrix of the Extended Kalman Filter (EKF) are calculated and monitored in real time. If an abnormal increase in innovation or rapid divergence in covariance is detected, it is determined that the fusion algorithm may be inaccurate, and an "algorithm health warning" event is immediately generated. Optionally, an algorithm reset or degradation strategy (such as increasing process noise) may be triggered. The estimation accuracy of the current fusion result (from...) The data (exported) and algorithm confidence (health score) are used as part of the core data and output to subsequent steps and system management logs.

[0042] S3: The fused positioning results are prioritized and converted according to the real-time requirements and protocol requirements of the target system to form multiple formatted data streams.

[0043] The system receives the standardized fusion positioning results from step S2 and performs intelligent priority scheduling and protocol conversion based on the real-time requirements, data format requirements, and application scenarios of multiple heterogeneous target systems in the vehicle environment to form multiple independent and adapted data streams for subsequent cross-system distribution.

[0044] Specifically, a requirements configuration table is maintained for each target system (client system) that needs to receive positioning data. Each client system must complete registration during initialization or runtime. Registration information includes: a system identifier to uniquely identify the client, such as Domain_Controller (autonomous driving domain controller), CarPlay_Interface (CarPlay interface), HM_Autopilot (HarmonyOS autonomous driving), etc.; data content requirements, specifying the required navigation parameters, such as position, speed, heading, confidence index, etc.; protocol and data format requirements, specifying compatible data encapsulation formats, such as NMEA-0183, SOME / IP, Protobuf, custom binary streams, or CAN frames defined by a specific DBC file, etc.; real-time requirements and update frequency, defining the real-time category (e.g., hard real-time, soft real-time) and the expected data update cycle (e.g., 100Hz, 10Hz, 1Hz); and transmission interface, specifying its communication interface or address, such as Ethernet socket address, CAN bus channel, USB endpoint, etc. Based on the real-time level and functional safety criticality of the target system, a distribution priority is assigned to each registered target system; this priority is used to schedule limited system resources in the data distribution link. For example, the following priority order can be followed: Autonomous Driving Domain Controller (hard real-time, safety critical) > Advanced Driver Assistance Systems (ADAS) > In-vehicle Connectivity Systems (such as CarPlay) > Map Navigation Applications > Traditional In-vehicle Information Display. The priority strategy can be dynamically adjusted via configuration files or OTA updates to adapt to different vehicle configurations or feature updates.

[0045] Furthermore, data scheduling and queue management are performed based on priority. Priority-based scheduling queues or parallel pipelines are maintained to manage data to be distributed. In each distribution cycle, the latest fused positioning result output from step S2 is written to the shared data buffer; based on the registration requirements of each client system, the required specific navigation data fields are extracted from the shared buffer to generate intermediate data packets for that system; each generated intermediate data packet is placed into the corresponding priority scheduling queue according to the distribution priority of the corresponding client system, with data packets in high-priority queues being processed first.

[0046] Intermediate data packets to be processed are retrieved from each scheduling queue in order of priority, and the corresponding protocol conversion engine is called to dynamically encapsulate the fused positioning results into a specified data format that meets the requirements of each target system. The protocol conversion engine incorporates several protocol generators, including: an NMEA-0183 generator for generating NMEA-compliant ASCII strings for target systems such as CarPlay, including $GNGGA and $GNRMC statements containing positioning information and $PASHR statements compliant with Apple INS certification requirements. The generator strictly adheres to the field order, checksum calculation, and frame interval specifications of the corresponding protocols; a SOME / IP serializer for providing serialized data compliant with the SOME / IP protocol stack specifications for AUTOSAR Adaptive or specific ADAS systems, including service identifiers, method identifiers, and structured payloads; a Protobuf encoder for generating binary streams based on Protocol Buffers encoding for target systems such as HarmonyOS. The encoder efficiently encodes navigation parameters according to a predefined .proto message structure; and a CAN frame encapsulator for generating CAN messages compliant with specific CAN database (DBC) definitions for systems communicating via the CAN bus (such as traditional motors), mapping speed, position, and other information to specified CAN bus parameters. The signal bits for the ID and data field; a custom binary stream wrapper, used to generate binary streams with a private frame structure, including frame headers, payloads, and cyclic redundancy checks, for systems with extremely high transmission efficiency requirements, such as autonomous driving domain controllers. During the conversion process, data downsampling or buffering can be performed before batch transmission based on the update frequency requirements registered by the target system.

[0047] A distribution deadline is set for each high-priority data distribution task. The scheduler monitors the task status to ensure that the protocol encapsulation and transmission process of the task starts before the deadline. For medium- and low-priority data distribution tasks, fair scheduling strategies such as time-slice round-robin can be used during scheduling cycles that do not require meeting the deadline to ensure that these systems can obtain processing resources periodically and avoid long-term blocking. If an anomaly occurs (such as transmission timeout, verification failure, protocol mismatch), the error is recorded and a decision is made based on the type and duration of the anomaly to skip the current distribution to the system, attempt to retransmit, or trigger a system alarm, ensuring that the distribution of high-priority systems is not affected by the failure of low-priority systems.

[0048] After completing the scheduling, protocol conversion, and timing control described above, multiple independent, formatted data streams are output. Each data stream contains complete data packets conforming to the protocol specifications of a specific target system, which can be directly acquired and sent by step S4 (cross-system compatible distribution step).

[0049] S4: Distribute the multiple formatted data streams to the corresponding target systems through a cross-system compatible architecture.

[0050] Multiple formatted and scheduled data streams are distributed to target applications running on heterogeneous operating systems through a layered adaptation framework, which includes an underlying communication abstraction layer, a protocol and data compatibility layer, and a system scheduling and application adaptation layer.

[0051] Specifically, the underlying communication abstraction layer encapsulates the native communication mechanisms of different operating systems into a unified application programming interface. It provides standardized functions for connecting, sending, receiving, and disconnecting; the interface takes the target system identifier as a parameter; upon system startup, it automatically identifies the currently running operating system type and version and dynamically loads the corresponding communication driver module. For example, for Linux systems, the adapter encapsulates Socket communication, supports binding Socket threads to specific CPU cores for real-time tasks, and employs real-time scheduling strategies such as SCHED_FIFO. For Android systems, the adapter encapsulates the Binder mechanism; thread safety is ensured through the Handler mechanism. For HarmonyOS systems, the adapter encapsulates a distributed soft bus, implementing device discovery and communication through DeviceManager. This layer uniformly implements connection keep-alive (heartbeat mechanism), automatic reconnection, timeout retransmission, and end-to-end data verification (such as CRC32) to ensure the reliability of the underlying link.

[0052] The protocol and data compatibility layer performs final encapsulation and permission adaptation to conform to the target system's ecosystem specifications before injecting the formatted data stream into the communication link. Based on the target system type, it performs final protocol encapsulation on the data packets formatted in step S3; for example, it adds a specific header conforming to Apple MFi certification requirements to the CarPlay NMEA data stream; and it encapsulates a distributed service header on the HarmonyOS Protobuf data stream. Before sending data to the target system, it automatically checks and requests the required system permissions and handles the callback logic for granting and denying permissions.

[0053] The system scheduling and application adaptation layer optimizes data distribution strategies based on the characteristics of the target system, ensuring that data can be consumed legally and efficiently. High priority is given to Linux real-time tasks; for Android applications, components such as AsyncTask or LiveData are used to avoid blocking the main thread; and task priority and load are balanced for HarmonyOS. Data is provided in a manner that complies with the target system's security specifications, such as through Android's Services or HarmonyOS's Abilities.

[0054] The cross-system compatible architecture also includes transmission quality monitoring and primary / backup link switching. This automatically switches to a backup communication link to ensure data distribution continuity when the primary distribution link quality deteriorates. Specifically, it monitors the end-to-end latency, packet loss rate, and heartbeat response status of each data link in real time, adding a sequence number and transmission timestamp to each data frame. Two independent communication paths (e.g., Ethernet primary, 4G / 5G network backup) are pre-configured for each target system. When the primary link quality continuously deteriorates (e.g., latency continuously exceeds thresholds, packet loss rate is too high, or heartbeat is lost), a switching decision is automatically triggered, including pausing the primary link, enabling the backup link, retransmitting cached data, and switching the data stream. The entire process is transparent to the application, and the switching time is controlled within the hundreds of milliseconds. After the primary link recovers, it supports smooth reverting to the backup link to avoid data interruption. Preferably, the quantitative triggering conditions for primary and backup link switching can be preset (e.g., 5 consecutive heartbeat timeouts > 200ms, or delays > 100ms for 1 second); once triggered, the switching logic is executed automatically, and the switching event, reason and new link identifier are recorded in the diagnostic log.

[0055] The adapted link will push data to the target application and feed back the distribution status information (such as success, failure, currently used link, monitoring metrics) to the system's management module for diagnosis, logging and status display.

[0056] S5: Continuously enhance the algorithm parameters in the external computing unit or the acquisition quality of the multi-source sensor data through a remote maintenance link.

[0057] By maintaining the remote link, the quality of algorithm parameters or multi-source sensor data acquisition in the external computing unit is continuously enhanced. This is achieved through two collaborative mechanisms: Assisted Global Navigation Satellite System (AGNSS) enhancement, which injects auxiliary ephemeris and clock data into the satellite positioning module through AGNSS technology to significantly shorten its cold start time and ensure that reliable GNSS observations can be obtained quickly in step S1; and Over-the-Air (OTA) technology upgrade, which remotely sends securely certified update packages to the external computing unit through the OTA channel to achieve secure updates and iterations of fusion algorithm parameters, sensor calibration parameters, or the algorithm logic itself.

[0058] Specifically, in AGNSS-assisted enhancement, when the vehicle is ignited or the GNSS module is cold-started, the external computing unit obtains auxiliary ephemeris and clock data of the current time and location from the cloud-based AGNSS server via a cellular network (such as 4G / 5G) or in-vehicle Ethernet, and injects it into the GNSS module through interfaces such as serial ports (UART). This allows the GNSS module to utilize this auxiliary information, greatly reducing the time required for satellite signal acquisition, ephemeris decoding, and initial positioning, and significantly shortening the cold start time (TTFF). This enables step S1 to quickly obtain reliable, high-precision GNSS timestamps and location observations in the early stages of system startup, especially when the vehicle leaves signal-obstructed areas such as underground parking garages, allowing the system to quickly enter a fusion state of strong / weak GNSS scenarios. AGNSS assistance can be automatically triggered by vehicle events (such as ignition) or actively requested by the monitoring function after detecting that GNSS has been unavailable for an extended period.

[0059] For OTA remote parameter optimization, the cloud platform receives and aggregates operational data from a large fleet of vehicles, including not only positioning results but also complete operational context, such as: positioning error distribution characteristics under different geographical and environmental scenarios, long-term performance statistics and drift characteristics of various types of sensors (IMU, GNSS modules), and the actual effectiveness of the core parameters of the fusion algorithm under the current configuration (such as the effect comparison of different weight strategies, the convergence speed and stability of filters). Based on this, two types of targeted upgrade packages are generated and distributed: 1. Parameter update package, which contains optimized values ​​of existing configuration parameters, which are applied to multiple levels, such as the optimization of the vehicle speed injection weight mapping table in step S2 (e.g., fine-tuning the vehicle speed weight from 0.5 to 0.55 in weak GNSS scenarios for a specific vehicle model), the correction of sensor fixed delay calibration values ​​in step S1 (e.g., updating the new calibration value of the transmission delay of a certain batch of IMUs), and the optimization of the baseline value of the EKF process noise covariance matrix Q or observation noise covariance matrix R in step S2. 2. Algorithm enhancement package, which includes improved versions of the core logic of the fusion algorithm (such as improved interpolation algorithms, new error compensation models), or new functional modules (such as more intelligent scene recognition and adaptive switching strategies).

[0060] Upgrades are performed via a secure OTA channel. The cloud management platform sends a digitally signed upgrade task to the target vehicle. The vehicle-side system management agent (a resident process) receives the task and uses differential upgrade technology to download only the changed parts to save bandwidth. The management agent verifies the digital signature and integrity of the upgrade package in a secure isolation area (e.g., hash verification). After successful verification, the upgrade package is temporarily stored in a backup non-volatile storage partition. When the system is idle and preset safety conditions are met (e.g., the vehicle is off and the battery / power status allows), the management agent deploys the new parameters or algorithm library to the runtime environment, preferably using A / B partitioning or containerization technology to achieve seamless and atomic switching between old and new versions, ensuring uninterrupted location services during the switching process. After the upgrade takes effect, key performance indicators are automatically collected during subsequent operation and compared with the baseline before the upgrade. Verification data can be fed back to the cloud through the independent diagnostic channel of the vehicle network module, forming an optimized closed loop from deployment to verification. If serious anomalies are detected after the upgrade (e.g., key performance indicators remain unmet or services crash), the management agent can automatically roll back to the previous stable version according to preset policies, maximizing vehicle functional safety and availability.

[0061] Example 2

[0062] like Figure 2 As shown in the figure, this embodiment exemplarily presents an external multi-source fusion positioning system based on vehicle speed injection enhancement, including a data acquisition module, an external computing unit, a communication interface module, and a remote maintenance interface; The data acquisition module is used to acquire multi-source sensor data from the vehicle, including interfaces for receiving GNSS (Global Navigation Satellite System) data, inertial measurement unit (IMU) data, and vehicle speed data. Specifically, the data acquisition module is the interface layer between the system and the vehicle's physical sensing environment. It reliably acquires multi-source, heterogeneous raw sensor data, handles initial signal conditioning and format parsing, and transmits the raw data stream to an external computing unit. The GNSS interface is used to receive data from the Global Navigation Satellite System. This interface can be a physical serial communication interface (such as UART) for connecting a standalone GNSS receiver module, or a vehicle bus interface (such as CAN or Ethernet) for subscribing to pre-calculated GNSS position / velocity information from other domain controllers in the vehicle (such as the infotainment domain controller). The inertial measurement unit (IMU) interface, typically a high-speed serial interface (such as SPI or I2C), is used to connect a six-axis or nine-axis IMU to acquire raw angular velocity and acceleration data of the vehicle. The vehicle speed data interface mainly refers to the controller area network interface, which is used to periodically read vehicle speed messages from the vehicle's CAN bus; optionally, it includes a pulse capture interface for directly acquiring pulse signals from wheel speed sensors as a dual-source vehicle speed verification and supplement.

[0063] The external computing unit, communicating with the data acquisition module, is used to perform global time synchronization, fusion computing, and distribution scheduling. It is an independent computing platform physically and logically decoupled from the vehicle's existing sensors and onboard operating system. It includes at least a high-performance multi-core processor (such as an ARM Cortex-A series), sufficient memory, and necessary coprocessors (such as an FPGA for high-precision timestamp processing). It acquires data through the data acquisition module and outputs results through the communication interface module. Its external and decoupled nature allows the algorithms and services within it to be developed, tested, deployed, and upgraded independently of the vehicle's electronic and electrical architecture, forming the basis for the system's freedom in algorithm evolution and cross-platform compatibility. The external computing unit carries and executes: global time synchronization service (corresponding to method step S1), achieving microsecond-level time alignment of multi-source data; multi-source fusion algorithm service (corresponding to method step S2), performing high-precision positioning calculations based on extended Kalman filtering with dynamic vehicle speed injection weight adjustment; and multi-path distribution scheduling service (corresponding to method step S3), managing the priority scheduling and protocol conversion of data streams.

[0064] The communication interface module is used to establish communication with multiple heterogeneous target systems through a cross-system compatible architecture. This module implements the cross-system compatible architecture, specifically including: a low-level communication abstraction layer, which encapsulates heterogeneous communication mechanisms such as Linux Socket, Android Binder, and HarmonyOS's distributed soft bus into a unified API through software drivers; a protocol and data compatibility layer, which incorporates multiple protocol stacks (such as NMEA-0183, SOME / IP, and Protobuf) to convert the scheduled data stream into the format required by the target system; and a system scheduling and application adaptation layer, which optimizes the distribution strategy based on the characteristics of the target operating system (such as Linux's real-time requirements and Android's application lifecycle). This module ensures that standardized positioning data generated by the external computing unit can be efficiently, reliably, and compliantly distributed to different target systems within the vehicle.

[0065] The remote maintenance interface connects to the remote maintenance link, continuously enhancing the acquisition quality of algorithm parameters or multi-source sensor data from external computing units. It encrypts and uploads system operating status, performance indicators, and fault logs to the cloud platform, providing a basis for remote data analysis and optimization decisions. It securely receives two key enhancements from the cloud: AGNSS auxiliary data, receiving ephemeris, clock, and other auxiliary information and forwarding it to the GNSS module to significantly shorten positioning startup time and improve data source quality; and OTA upgrade packages, receiving digitally signed and encrypted algorithm parameter packages, calibration parameter packages, or complete firmware upgrade packages, which are then verified and installed by the system's secure management agent, enabling remote algorithm iteration and dynamic parameter optimization.

[0066] The data acquisition module supplies raw data to the external computing unit; the results processed by the external computing unit are distributed to each vehicle-side system through the communication interface module; at the same time, the remote maintenance interface collaborates with the cloud to continuously optimize the input quality of the data acquisition module and the processing logic of the external computing unit.

[0067] Other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of the invention are indicated by the claims. It should be understood that the invention is not limited to the precise structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. An external multi-source fusion localization method based on vehicle speed injection enhancement, characterized in that, The method operates on an external computing unit decoupled from onboard sensors and multiple heterogeneous operating systems, including: Global time synchronization is performed on multi-source sensor data from the vehicle, wherein the multi-source sensor data includes at least inertial measurement data, satellite positioning data, and vehicle speed data; In the external computing unit, a fusion algorithm is executed to process the multi-source sensor data after global time synchronization and generate a fusion positioning result; the fusion algorithm dynamically adjusts the contribution weight of the vehicle speed data in the fusion calculation according to the real-time quality of the satellite positioning signal. The fused positioning results are prioritized and converted according to the real-time requirements and protocol requirements of the target system to form multiple formatted data streams. The multiple formatted data streams are distributed to the corresponding target systems using a cross-system compatible architecture. By remotely maintaining the link, the algorithm parameters in the external computing unit or the acquisition quality of the multi-source sensor data can be continuously enhanced.

2. The external multi-source fusion localization method based on vehicle speed injection enhancement according to claim 1, characterized in that, The global time synchronization includes establishing and taming a global reference clock source; based on the global reference clock source, marking the inertial measurement data, satellite positioning data, and vehicle speed data with a native timestamp of a unified time reference at the physical time of acquisition; and performing time axis alignment and data reorganization on the multi-source data streams carrying the native timestamps to generate a time-synchronized data sequence.

3. The external multi-source fusion localization method based on vehicle speed injection enhancement according to claim 1, characterized in that, The fusion algorithm determines the scene as a strong satellite positioning signal scene, a weak satellite positioning signal scene, or a scene without a satellite positioning signal based on the intensity parameter of the satellite positioning signal. According to the scene determination result, the observation noise covariance matrix or observation vector of the fusion algorithm is dynamically configured to adjust the fusion weight of the vehicle speed data relative to other data sources.

4. The external multi-source fusion localization method based on vehicle speed injection enhancement according to claim 3, characterized in that, When the scenario is determined to be a weak satellite positioning signal scenario or a scenario without a satellite positioning signal, the fusion weight of the vehicle speed data is increased, so that the vehicle speed data becomes the core observation source to suppress the integral drift of the inertial measurement data.

5. The external multi-source fusion localization method based on vehicle speed injection enhancement according to claim 1, characterized in that, The priority scheduling and protocol conversion process includes configuring and distributing priorities for different target systems, wherein the priorities are determined based on the real-time level and functional safety criticality of the target systems; and calling the corresponding protocol conversion engine in descending order of priority to convert the fused positioning results into a specified data format that meets the requirements of each target system.

6. The external multi-source fusion localization method based on vehicle speed injection enhancement according to claim 5, characterized in that, The protocol conversion engine supports conversion of at least one or more of the following data formats: NMEA-0183 format, SOME / IP protocol format, Protocol Buffers encoding format, and vehicle CAN bus message format.

7. The external multi-source fusion localization method based on vehicle speed injection enhancement according to claim 1, characterized in that, Distribution is carried out through the cross-system compatible architecture, including encapsulating the native communication mechanisms of different operating systems into a unified application programming interface through the underlying communication abstraction layer; Through the protocol and data compatibility layer, the formatted data stream is injected into the communication link, and final encapsulation and permission adaptation are performed in accordance with the target system's ecological specifications.

8. The external multi-source fusion localization method based on vehicle speed injection enhancement according to claim 7, characterized in that, The cross-system compatible architecture also includes transmission quality monitoring and primary / backup link switching, which automatically switches to a backup communication link to ensure the continuity of data distribution when the quality of the primary distribution link is detected to be deteriorated.

9. The external multi-source fusion localization method based on vehicle speed injection enhancement according to claim 1, characterized in that, The remote maintenance link includes injecting auxiliary ephemeris data into the satellite positioning module through Assisted Global Navigation Satellite System technology to shorten its positioning startup time; and remotely sending and securely updating the fusion algorithm parameters or sensor calibration parameters to the external computing unit through over-the-air download technology.

10. An external multi-source fusion positioning system based on vehicle speed injection enhancement, used to implement the method as described in any one of claims 1-9, characterized in that, Includes a data acquisition module, an external computing unit, a communication interface module, and a remote maintenance interface; The data acquisition module is used to acquire multi-source sensor data from the vehicle, including interfaces for receiving satellite positioning data, inertial data, and vehicle speed data. The external computing unit is communicatively connected to the data acquisition module and is used to perform global time synchronization, fusion computing, and distribution scheduling. The communication interface module is used to establish communication with multiple heterogeneous target systems through a cross-system compatible architecture; The remote maintenance interface is used to connect to a remote maintenance link to continuously enhance the algorithm parameters in the external computing unit or the acquisition quality of the multi-source sensor data.

Images