Parking relocation memory method and device, electronic equipment and storage medium

Abstract

The invention provides a memory parking repositioning method and device, electronic equipment and a storage medium, and in the method, a continuous path is decomposed into a plurality of topological path segments, so that a complex continuous space positioning problem is converted into a discrete path segment sequence matching problem, and the positioning accuracy is improved. Complexity and calculation amount of subsequent positioning search are greatly reduced, cost is low, in addition, when it is recognized that the vehicle enters the next topological path section from one topological path section, the database is inquired to obtain the starting point pose of the new topological path section, the starting point pose of the new topological path section is directly used for resetting the pose of the vehicle subjected to track plotting, and therefore the accuracy of track plotting of the vehicle is improved. According to the method, when a vehicle runs in a topological path section from zero, the transverse distance from the vehicle to the reference path of the current topological path section is measured according to a preset interval, and the pose of the vehicle is subjected to filtering correction, so that the effects of regular zero clearing of a large error and real-time suppression of a small error are achieved; and high precision of full-path relocation is ensured.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of automobiles, in particular to a memory parking repositioning method and device, electronic equipment and a storage medium. BACKGROUND

[0002] Memory parking (HPP) is one of the key scenarios for the landing of automatic driving technology, and repositioning is a core supporting technology to ensure its reliable operation. Through the closed-loop logic of learning-memory-reproduction, the high-frequency demand problem of automatic driving in closed / semi-closed scenarios (GNSS denial scenarios) is solved.

[0003] Repositioning refers to the process of determining the precise position of a vehicle in a map through perception or other technical means when reproducing a memory path, and its importance is comparable to that of a "compass for automatic driving".

[0004] The current repositioning method mainly includes the following three kinds: the first is through the way of visual SLAM, the advantage is strong environmental understanding ability, the shortcoming is that the global search calculation is complex (O(n³)), the matching rate is low in degenerative scene, and the feature points are reduced by 80% in low light (<50 lux) scene; the second is through the way of laser SLAM, the advantages are millimeter level precision, all-weather, the shortcomings are initial pose dependence (<3m), high cost (>1500 dollars), and in the scene of glass curtain wall reflection, the point cloud distortion is ±3.2m; the third is the way of inertial navigation (DR), the advantages are autonomous work and low cost, and the shortcoming is that the error accumulates quickly (1.5m / minute), and after 60 seconds of lock loss, the error reaches 25m.

[0005] In summary, the traditional repositioning method cannot meet the needs of low cost, high precision and strong robustness at the same time. SUMMARY

[0006] Therefore, the purpose of the present application is to provide a memory parking repositioning method, device, electronic equipment and storage medium to alleviate the technical problem that the traditional repositioning method cannot meet the needs of low cost, high precision and strong robustness at the same time.

[0007] In a first aspect, an embodiment of the present application provides a memory parking repositioning method, comprising: When a vehicle records a memory path, a turning action is detected in real time, and a pose at the time when the turning action is detected is taken as a turning point, the turning point is taken as a boundary, and then a continuous path is divided into a plurality of topological path segments, and a database of a topological path segment sequence is constructed; When the vehicle travels along the memory path, inertial measurement unit-based dead reckoning is performed to obtain the pose of the vehicle, and the following layered pose correction is performed on the pose of the vehicle: Inter-segment global correction: when it is identified that the vehicle enters a next topological path segment from a current topological path segment, triggering a global correction, querying the database to obtain a start pose of the next topological path segment, and updating the pose of the vehicle with the start pose of the next topological path segment; In-segment local correction: when the vehicle is driving in the interior of a topological path segment, triggering a local correction at a preset interval, and filtering and correcting the pose of the vehicle by measuring a lateral distance of the vehicle to a reference path of the current topological path segment.

[0008] Further, detecting a turning action in real time, and taking the pose at the time when the turning action is detected as a turning point, comprising: A calculation formula of an angle change of vehicle turning calculating the angle change of vehicle turning, wherein, denotes the angle change of vehicle turning, denotes an angular rate output by a gyroscope, denotes a corresponding zero offset value of the gyroscope; if the angle change of vehicle turning is greater than a preset angle threshold value and a duration is greater than a preset duration threshold value, it is determined that the vehicle has a turning action, and the pose at the time when the turning action is detected is taken as a turning point.

[0009] Further, when it is identified that the vehicle enters a next topological path segment from a current topological path segment, comprising: when the turning action is detected again, it is determined that the vehicle enters the next topological path segment from the current topological path segment.

[0010] Further, filtering and correcting the pose of the vehicle by measuring a lateral distance of the vehicle to a reference path of the current topological path segment, comprising: taking a system state of a dead reckoning as a prediction value, taking the lateral distance as an observation, and filtering and correcting the system state by a Kalman filtering algorithm to obtain a corrected system state, wherein the system state comprises the pose of the vehicle.

[0011] Further, an observation equation of the lateral distance is wherein, denotes the lateral distance, denotes an observation matrix, , denotes a heading angle of the current topological path segment, denotes the corrected system state, denotes observation noise.

[0012] Further, the preset interval comprises a preset distance interval or a preset time interval.

[0013] Further, the system state comprises: planar coordinates of the vehicle, a heading angle, a gyroscope zero offset and an accelerometer zero offset. The database comprises: an identification of each topological path segment, a start pose of each topological path segment and a length of each topological path segment.

[0014] In a second aspect, the embodiments of the present application further provide a device for memorizing a parking repositioning, comprising: The constructing unit is configured to, when the vehicle is recording a memory path, detect a turning action in real time, take a pose at the time when the turning action is detected as a turning point, take the turning point as a boundary, and further decompose a continuous path into a plurality of topological path segments to construct a database of a topological path segment sequence. The hierarchical pose correction unit is configured to, when the vehicle is driving along the memory path, perform a hierarchical pose correction on the pose of the vehicle based on an inertial measurement unit, and obtain the pose of the vehicle. Inter-segment global correction: when it is identified that the vehicle enters a next topological path segment from a current topological path segment, a global correction is triggered, a start pose of the next topological path segment is obtained by querying the database, and the pose of the vehicle is updated by using the start pose of the next topological path segment. In-segment local correction: when the vehicle is driving in the interior of a topological path segment, a local correction is triggered at a preset interval, a lateral distance of the vehicle to a reference path of the current topological path segment is measured, and the pose of the vehicle is filtered and corrected.

[0015] In a third aspect, the embodiments of the present application further provide an electronic device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method in any one of the first aspect.

[0016] In a fourth aspect, the embodiments of the present application further provide a computer readable storage medium, which stores a computer program, wherein the computer program is executable on a processor to execute the method in any one of the first aspect.

[0017] In the embodiment of the present application, a memory parking relocation method is provided, comprising: when a vehicle records a memory path, a turning action is detected in real time, and a pose at the time when the turning action is detected is taken as a turning point, the turning point is taken as a boundary, and then a continuous path is divided into a plurality of topological path segments, so as to construct a database of a topological path segment sequence; when the vehicle travels along the memory path, inertial measurement unit is used for dead reckoning to obtain a pose of the vehicle, and the following layered pose correction is performed on the pose of the vehicle: inter-segment global correction: when it is identified that the vehicle enters a next topological path segment from a topological path segment, global correction is triggered, a starting pose of the next topological path segment is obtained by querying the database, and the starting pose of the next topological path segment is used to update the pose of the vehicle; intra-segment local correction: when the vehicle travels in the interior of a topological path segment, local correction is triggered at a preset interval, and the pose of the vehicle is filtered and corrected by measuring a lateral distance of the vehicle to a reference path of the topological path segment in which the vehicle is currently located. As can be seen from the above description, in the memory parking relocation method of the present application, the continuous path is divided into a plurality of topological path segments, so that the complex continuous space positioning problem is converted into a discrete path segment sequence matching problem, the complexity and calculation amount of subsequent positioning search are greatly reduced, the cost is low, and in addition, when it is identified that the vehicle enters a next topological path segment from a topological path segment, a starting pose of the next topological path segment is obtained by querying the database, and the starting pose of the next topological path segment is directly used to reset the pose of the vehicle obtained by dead reckoning, so that "starting from zero" is realized, and when the vehicle travels in the interior of a topological path segment, the pose of the vehicle is filtered and corrected by measuring a lateral distance of the vehicle to a reference path of the topological path segment in which the vehicle is currently located at a preset interval, so that the effect of periodically clearing small errors and real-time suppressing large errors is realized, and the low-frequency absolute correction and high-frequency relative correction are used together to guarantee the high accuracy and good robustness of the whole path relocation, and the technical problem that the traditional relocation method cannot simultaneously meet the requirements of low cost, high accuracy and strong robustness is solved. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the specific embodiments of the present application or the technical solutions in the prior art, the following will briefly introduce the drawings needed to be used in the specific embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without any creative effort.

[0019] Figure 1 A flowchart of a memory parking relocation method provided in the embodiment of the present application is shown in the figure. Figure 2 A schematic diagram of generating a topological path segment provided in the embodiment of the present application is shown in the figure. Figure 3A schematic diagram of the hierarchical correction strategy provided by the embodiment of the present application is shown in the figure; Figure 4 A schematic diagram of the global correction effect provided by the embodiment of the present application is shown in the figure; Figure 5 A schematic diagram of the long straight path contrast result provided by the embodiment of the present application is shown in the figure; Figure 6 A schematic diagram of the curved path scene contrast result provided by the embodiment of the present application is shown in the figure; Figure 7 A schematic diagram of a memory parking repositioning device provided by the embodiment of the present application is shown in the figure; Figure 8 A schematic diagram of an electronic device provided by the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0020] The technical solutions of the present application will be described in detail below with reference to the embodiments. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts fall within the scope of the present application.

[0021] The traditional repositioning method cannot meet the requirements of low cost, high precision and strong robustness at the same time.

[0022] Therefore, in the memory parking repositioning method of the present application, the continuous path is divided into a plurality of topological path segments, so that the complex continuous spatial positioning problem is converted into a discrete path segment sequence matching problem, greatly reducing the complexity and calculation amount of subsequent positioning search, and the cost is low. In addition, when the vehicle is identified to enter the next topological path segment from a topological path segment, the start pose of the new topological path segment is queried from the database, and the pose of the vehicle calculated by the track is directly reset to the start pose of the new topological path segment, so as to realize "starting from zero". When the vehicle is driving inside a topological path segment, the pose of the vehicle is filtered and corrected by measuring the lateral distance of the vehicle to the reference path of the current topological path segment at a preset interval, that is, the effect of large error periodic zero clearing and small error real-time suppression is realized. The low-frequency absolute correction and high-frequency relative correction are used together to ensure the high precision and good robustness of the whole path repositioning.

[0023] In order to facilitate the understanding of the present embodiment, first, a memory parking repositioning method disclosed by the embodiment of the present application is described in detail.

[0024] Embodiment one: According to an embodiment of the present application, an embodiment of a method for memory parking repositioning is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described herein can be executed in an order different from that shown.

[0025] Figure 1 is a flowchart of a method for memory parking repositioning according to an embodiment of the present application, as shown, the method comprises the following steps: Figure 1 Step S102, when recording the memory path, the vehicle detects the turning action in real time, and takes the pose at the time of detecting the turning action as the turning point, and further decomposes the continuous path into a plurality of topological path segments, and a database of the topological path segment sequence is constructed; Step S104, when the vehicle travels along the memory path, the inertial measurement unit is used for dead reckoning to obtain the pose of the vehicle, and the following layered pose correction is performed on the pose of the vehicle: Inter-segment global correction: when it is identified that the vehicle enters the next topological path segment from one topological path segment, the global correction is triggered, the starting pose of the new topological path segment is obtained by querying the database, and the starting pose of the new topological path segment is used to update the pose of the vehicle; In-segment local correction: when the vehicle travels inside one topological path segment, the local correction is triggered at a preset interval, the pose of the vehicle is filtered and corrected by measuring the lateral distance of the vehicle to the reference path of the current topological path segment.

[0026] Specifically, the present application proposes a hybrid positioning strategy of "segmented topological path matching + limited correction", which realizes the conversion of the absolute positioning problem into the path sequence matching problem, the suppression of the DR (Dead Reckoning) error divergence through topological constraints, and the use of in-segment correction to maintain local accuracy. In the mapping stage, the continuous path is decomposed into a topological path segment sequence with turning actions as boundaries, and in the positioning stage, layered pose correction is performed, including inter-segment global correction (triggered when the topological segment is switched) and in-segment local correction (triggered at a preset interval). The absolute positioning problem is converted into the problem of judging whether the current driving path and the mapping path are isomorphic. The above-mentioned reference path is the memory path recorded during the recording of the memory path.

[0027] The essence of DR (Dead Reckoning) is integration. It calculates the displacement and orientation of the vehicle through sensor data (acceleration, angular velocity) such as inertial measurement unit (IMU).

[0028] ​The tiny bias and random noise of the sensor will be amplified during the integration process, leading to the monotonic and unbounded increase of the estimated position error over time, which is called "error divergence". The longer the vehicle runs, the more inaccurate the estimated position will be.

[0029] "Topology" refers to the connection and structure of the vehicle's driving path (e.g., sequence, turning points). "Topology constraint" refers to the prior knowledge that the vehicle's actual driving path must be consistent with the pre-recorded memory path in terms of topology. In simple terms, it is a powerful rule: "the vehicle can only follow the route drawn by the memory path, and cannot run randomly". This rule provides a strong anchor for suppressing DR error.

[0030] Topology constraint suppresses error divergence through two mechanisms: inter-segment global correction and intra-segment local correction.

[0031] a) Inter-segment global correction: error reset using "key points" in the topology structure (i.e., the start / turning points of the topology path segment (the start point is the turning point)); Correspondence: This directly corresponds to the most core and effective part of "suppressing DR error divergence through topology constraint".

[0032] Working principle: Topology path segments are composed of "straight line segments" divided by "turning points". These turning points are the key points in the topology structure. During repositioning, the system determines that the vehicle will enter the next topology path segment through pattern recognition (e.g., detecting that the vehicle is turning).

[0033] Constraint application: At this time, the topology constraint takes effect - "the vehicle must start from the start point of the next topology path segment". The system then queries the memory database and directly calls the precise coordinates (x, y, θ) of the start point of the next topology path segment to forcibly reset the vehicle's pose estimated by the DR, which has accumulated a large error.

[0034] Effect: Inter-segment global correction is to forcibly "reset to zero" the DR accumulated error at the boundary (turning point) of each topology path segment, thereby preventing the continuous divergence of error across topology path segments. This is a discrete, forced, and global error suppression.

[0035] b) Intra-segment local correction: error fine-tuning within the topology structure "segment"; Correspondence: This is a mechanism to assist in maintaining topology constraints within the segment.

[0036] Working principle: Within a topology path segment (usually a long straight line), DR error is still accumulating. Topology constraint is reflected in "the vehicle should drive near this straight line reference path".

[0037] Application of constraints: the system continuously measures the lateral distance of the vehicle from this straight reference path via sensors (e.g. cameras). If the DR-reckoned trajectory starts to deviate, this observation will capture the anomaly.

[0038] Effect: this lateral distance observation is used to continuously and slightly correct the DR pose estimate via algorithms such as Kalman filtering, ensuring that the vehicle trajectory is tightly "glued" to the reference path. This suppresses the rate and magnitude of error divergence within individual segments, and is a continuous, smooth, local error suppression. It provides a safeguard against the DR error becoming too large by the time it reaches the next topological path segment boundary. Using a simple and reliable observation (the lateral distance of the vehicle from the reference path, i.e.

[0039] "Suppression of DR error divergence via topological constraints" is a system-level design philosophy. It mainly achieves a qualitative leap (periodic zeroing) via inter-segment global correction, while intra-segment local correction is a quantitative optimization (continuous fine-tuning) based on this. The two complement each other, and together transform the DR system from an "open loop" system where errors will diverge indefinitely, to a "closed loop" system where errors are limited to fluctuation around the topological path, thus achieving high-precision repositioning.

[0040] ​In this embodiment of the invention, a method for memory-based parking relocation is provided, comprising: when a vehicle is recording a memory path, real-time detection of turning actions is performed, and the pose at the time of detection of the turning action is used as the turning point, the turning point is used as the boundary, and then the continuous path is decomposed into multiple topological path segments to construct a database of topological path segment sequences; when the vehicle is traveling along the memory path, trajectory calculation is performed based on the inertial measurement unit to obtain the vehicle's pose, and the following hierarchical pose correction is performed on the vehicle's pose: inter-segment global correction: when it is detected that the vehicle enters the next topological path segment from one topological path segment, global correction is triggered, the database is queried to obtain the starting pose of the new topological path segment, and the vehicle's pose is updated using the starting pose of the new topological path segment; intra-segment local correction: when the vehicle is traveling inside a topological path segment, local correction is triggered at preset intervals, and the vehicle's pose is filtered and corrected by measuring the lateral distance from the vehicle to the reference path of the current topological path segment. As described above, the memory parking repositioning method of the present invention decomposes a continuous path into multiple topological path segments. This transforms the complex continuous spatial positioning problem into a discrete path segment sequence matching problem, greatly reducing the complexity and computational load of subsequent positioning searches, resulting in low cost. Furthermore, when a vehicle is detected entering a new topological path segment from one topological path segment, the starting pose of the new topological path segment is retrieved from the database, and the vehicle's pose calculated from the trajectory is directly reset using the starting pose of the new topological path segment, achieving "starting from zero." When the vehicle is traveling within a topological path segment, the vehicle's pose is filtered and corrected by measuring the lateral distance from the vehicle to the reference path of the current topological path segment at preset intervals. This achieves the effect of periodically clearing large errors and suppressing small errors in real time. The use of low-frequency absolute correction and high-frequency relative correction together ensures high accuracy and robustness of the full-path repositioning, alleviating the technical problem that traditional repositioning methods cannot simultaneously meet the requirements of low cost, high accuracy, and strong robustness.

[0041] The above provides a brief overview of the memory parking repositioning method of the present invention. The specific details involved are described in detail below.

[0042] In an optional embodiment of the present invention, the turning action is detected in real time, and the pose at the time of detection of the turning action is taken as the turning point, specifically including the following steps: (1) Calculation formula based on the change in vehicle steering angle Calculate the change in the vehicle's steering angle, where, Indicates the change in the vehicle's steering angle. This represents the angular rate output by the gyroscope. This indicates the zero bias value corresponding to the gyroscope; (2) If the angle change of the vehicle steering is greater than a preset angle threshold, and the duration is greater than a preset duration threshold, it is determined that the vehicle has a turning action, and the pose at the time of detecting the turning action is taken as the turning point.

[0043] Specifically, in the mapping stage, the topological path segment generation process is as follows (as shown in Figure 2 ): Real-time detection of turning actions during vehicle travel, with the turning point as the boundary of the topological path segment; Turning detection criteria: the angle change of the vehicle steering (i.e., the steering angle rate integral value) > 45° (for more than 0.5s), and the angle change of the vehicle steering is calculated as follows: wherein, represents the angular rate output by the gyroscope, represents the zero offset value corresponding to the gyroscope.

[0044] The path between two turning actions is defined as a straight line segment, and the straight line segment distance is calculated according to the following formula: wherein, represents the straight line segment distance, represents the pulse change amount in the update period, represents the distance corresponding to each pulse; (Pulse change amount): Source: usually from the encoder installed on the wheel. The encoder will generate a fixed number of pulse signals every time the vehicle travels a fixed distance (e.g., one wheel rotation).

[0045] Physical meaning: represents the total number of pulses generated by the encoder in one data update period. It is directly proportional to the distance traveled by the vehicle in that time period.

[0046] (Calibration coefficient, distance corresponding to each pulse): Definition: This is a calibration constant that represents the actual physical distance corresponding to each pulse (e.g., 0.01 meters per pulse).

[0047] How to determine: obtained through experimental calibration. For example, let the vehicle travel exactly 100 meters, and record the total number of pulses . .

[0048] L (straight line segment distance): Result: multiplying the pulse count by the calibration coefficient gives the actual physical length of the straight line path traveled by the vehicle between two turning points.

[0049] The straight line distance L, which plays a crucial role in the topology relocation scheme, is the core attribute of the topology segment.

[0050] Mapping phase: define the "identity" attribute of the topology segment; During mapping, the system calculates the length L of the straight line segment using the above straight line distance calculation formula after detecting the turning points (such as P1 and P2). Then, the key information of a topology path segment is stored in the database, usually including: starting point coordinates (the accurate pose of the turning point P1), length (L), and heading angle (the approximate direction of the straight line segment). In this way, each topology path segment is no longer an abstract line segment, but a real entity with measurable geometric attributes.

[0051] Relocation phase: used for matching verification and auxiliary correction; Segment matching verification: when the vehicle is driving in the relocation phase, it may determine that it has entered a certain topology path segment through DR calculation and turning detection. At this time, the system can check: "does the driving distance calculated by the DR from the starting point of the segment match the length L of the segment in the database?" If there is a serious discrepancy, a segment matching error may have occurred, which needs to be re-matched. This increases the robustness of the system.

[0052] Triggering local correction within the segment: "triggering local correction within the segment at preset intervals". If the "preset interval is a preset distance interval", the reference for the "preset distance interval" is the total length L of the topology segment (i.e. the topology path segment). For example, the system can be set to perform lateral observation and Kalman filter correction when driving to L / 4, L / 2, 3L / 4, etc. L provides an accurate distance reference scale for correction.

[0053] Auxiliary DR calculation: the length L itself can be used as a distance observation value to constrain the DR trajectory calculation, further suppressing the cumulative error. In terms of technology, the difference (residual error) between the theoretical length L (true value) of the topology segment and the cumulative driving distance (observation value) calculated by the DR within the segment is used to correct the system error of the DR (such as the odometer coefficient and sensor zero offset).

[0054] In an optional embodiment of the present application, when the vehicle is identified to enter the next topology path segment from a topology path segment, it includes: When the turning action is detected again, it is determined that the vehicle enters the next topology path segment from a topology path segment.

[0055] In an optional embodiment of the present application, the pose of the vehicle is filtered and corrected by measuring the lateral distance of the vehicle to the reference path of the current topology path segment, including: The system state of the dead reckoning is taken as a prediction value, and a lateral distance is taken as an observation value, and the system state is filtered and corrected through a Kalman filtering algorithm to obtain a corrected system state, wherein the system state includes a pose of the vehicle.

[0056] Specifically, an observation equation of the lateral distance is wherein, represents the lateral distance, represents an observation matrix, , represents a heading angle of a current topological path segment, represents the corrected system state, represents an observation noise.

[0057] Specifically, a hierarchical correction strategy is as shown in Figure 3 . Inter-segment global correction: when a new topological path segment is identified (that is, the vehicle enters a next topological path segment from a topological path segment, that is, a turning action is detected again), an initial pose of the new topological path segment is obtained from a database to reset a current absolute pose (that is, the pose of the vehicle obtained through the dead reckoning), and a previous cumulative error is eliminated.

[0058] In-segment local pose correction: in some application scenarios, a path segment length is more than 100 meters, or even longer, and an error is at least 1 meter according to a one percent index of the dead reckoning, which exceeds a visual matching correction capability. Therefore, an in-segment local correction strategy is introduced, and an optimization model is as shown below. wherein, represents a vehicle plane coordinate, represents a vehicle heading angle, respectively represent a gyroscope and an accelerometer zero offset, represents a process noise, and an initial value is taken as in the system.

[0059] In a vehicle body coordinate system, a lateral distance represents a vertical distance from the vehicle to a reference path, and a specific observation equation is as shown below. wherein, an observation Jacobii matrix is wherein, represents a locked topological path segment heading angle, is an observation noise, and 0.1 m is taken here.

[0060] The process of the in-segment local correction is further introduced below. Step one: DR state prediction Action: During vehicle driving, sensors such as IMU and wheel speedometer are working continuously. DR algorithm predicts the state vector of vehicle at time k by integrating the data (angular velocity, acceleration) from these sensors.

[0061] State vector (xk) : usually contains vehicle's position, velocity, attitude and sensor error parameters.

[0062] where, xk represents vehicle's global coordinate, ψk represents heading angle (vehicle's front direction), gk represents gyroscope bias, ak represents accelerometer bias.

[0063] Output: DR outputs a predicted state vector However, due to sensor bias and noise, this predicted value will accumulate errors over time.

[0064] Step 2: Lateral distance observation Trigger condition: The system triggers observation and correction once every preset distance interval (e.g. every 5 meters) or preset time interval.

[0065] Action: Calculate the vertical (lateral) distance between the currently calculated position and the topological path segment saved in the memory path. This value is recorded as the observation value .

[0066] Importance: is an absolute, time-independent observation quantity. If the DR calculation is accurate, the vehicle should always be driving on the reference path, should be constantly 0. Any ≠ 0 directly reveals the error in DR calculation.

[0067] Step 3: Kalman filter fusion and correction This is the core algorithm step of the entire correction process. The Kalman filter receives the predicted value of DR and the observation value of the sensor, and performs optimal fusion.

[0068] Calculate Kalman gain (Kk): Gain Kk is a weight matrix that determines whether more trust should be placed in the prediction of DR or in the observation of the sensor.

[0069] If the observation is very reliable (e.g. clear camera), the gain increases, and more trust is placed in the observation; if the short-term accuracy of DR is high, more trust is placed in the prediction.

[0070] Update state estimation (xk+1) : ​: Filter calculates the residual (Innovation) between predicted and observed values: Residual = Observed - Predicted -H×

[0071] Where H (i.e. the observation matrix) maps the state vector to the predicted observation space (i.e. what the lateral distance should be based on the DR state prediction). The residual is assigned back to the state vector using the Kalman gain for correction:

[0072] = +Kk×(Residual), resulting in an updated, optimal state estimate . This new state estimate incorporates both the dynamic information from the DR and the calibration information from the absolute observations. Step Four: Error Compensation and Output

[0073] Action: The error estimates (e.g. gyroscope bias ) in the updated state vector are fed back to the DR system. Purpose: Closed-loop control. The DR system uses these corrected error parameters for the next round of trajectory prediction, improving the accuracy of the prediction model at the source rather than just post-processing the output.

[0074] Output: The final corrected high-precision pose (x, y, θ) is sent to the vehicle control system for precise path tracking.

[0075] Step Five: Loop Execution

[0076] Action: Repeat steps one to four until the vehicle reaches the end of the current topology segment (the next turning point). Mode: This is a closed-loop control process of "predict → observe → correct → predict again".

[0077] Effect: Through this periodic "fine-tuning", the cumulative error of the DR is effectively suppressed within a small range, preventing it from freely diverging. Even on a long straight road of hundreds of meters, the positioning accuracy remains extremely high.

[0078]

[0079] is an extremely ingenious design, and its dimension implies that the state vector (i.e. the system state) is 5-dimensional, for example , whose values , , , represents lateral distance only strongly related to the global position (x, y) of the vehicle and this relation has a geometric sine / cosine relation to the heading angle of the reference path There is a geometric sine / cosine relation; [0, 0, 0] represents lateral distance to the heading angle in the state vector , gyro bias , accelerometer bias There is no direct linear relation. This means that the states cannot be directly corrected, their estimation mainly relies on the system model and other states.

[0080] In an optional embodiment of the present application, the preset interval comprises: a preset distance interval or a preset time interval.

[0081] In an optional embodiment of the present application, the system state comprises: a planar coordinate of the vehicle, a heading angle, a gyro bias and an accelerometer bias. The database comprises: an identification of each topological path segment, a starting point pose of each topological path segment and a length of each topological path segment.

[0082] The global correction effect of the present application is shown in Figure 4 .

[0083] The comprehensive performance quantitative comparison is shown in the following table: Table 1 Performance quantitative comparison

[0084] Performance analysis of long straight road scene (80m): Test conditions: no GNSS, initial error 0m, vehicle speed 7km / h; the comparison results are shown in Figure 5 .

[0085] Performance analysis of continuous S-bend scene: Test conditions: 3 continuous turns (60°→90°→120°), path length 120m, the comparison results are shown in Figure 6 .

[0086] The quantitative comparison results of the turning point improvement are shown in the following table:

[0087] Conclusion: The above experiments show that the error divergence suppression strategy of resetting the dead reckoning error at the topological segment boundary and performing lateral correction at a fixed distance interval is effective.

[0088] The invention points of the present application are as follows: Invention point one: path topological segmentation modeling method based on turning action: ​Technical problem solved: Traditional solutions attempt to locate once in the global coordinate system, complex calculation and easy to be affected by cumulative error.

[0089] Technical means: In the mapping stage, the vehicle turning action (steering angle speed integral > 45°) is used as a natural boundary to divide the continuous driving path into a series of "topological segments" dominated by straight lines. Each segment is abstracted as a basic unit with attributes such as starting point coordinates, length, heading, etc.

[0090] Technical effect: The complex continuous space positioning problem is transformed into a discrete path segment sequence matching problem, which greatly reduces the complexity and calculation amount of subsequent positioning search.

[0091] Invention point two: hierarchical correction mechanism combining global between segments and local within segments: Technical problem solved: Low-cost inertial navigation (DR) error will accumulate over time, and a single correction method cannot balance accuracy and frequency.

[0092] Technical means: Global correction between segments: At the boundary of topological segment switching (turning point), the cumulative error of DR is reset by directly using the accurate pose of the starting point of the new segment, achieving "starting from zero".

[0093] Local correction within the segment: Within the segment, at fixed distance intervals, the lateral distance between the vehicle and the current segment reference path is measured as an observation value, and Kalman filtering and other algorithms are used to correct the DR pose in a small range, suppressing error divergence.

[0094] Technical effect: It realizes the effect of "large error periodic zeroing and small error real-time suppression", and uses low-frequency absolute correction and high-frequency relative correction to jointly ensure the high accuracy of the whole path.

[0095] Invention point three: lightweight intra-segment correction model based on lateral distance observation: Technical problem solved: In long straight road scenes, DR error accumulates a large amount, exceeding the effective matching range of sensors such as vision.

[0096] Technical means: In intra-segment correction, instead of relying on complex global feature matching, only the lateral distance between the vehicle and the reference path, a simple and easy-to-get observation, is used. By defining the observation equation and Jacobian matrix, it is deeply integrated with the DR system to achieve efficient filtering.

[0097] Technical effect: It greatly reduces the requirements for sensors, and does not require expensive lidar or high-performance vision systems, but only requires low-cost sensors (such as camera ranging) to achieve effective correction, perfectly meeting the requirements of vehicle-mounted systems for low cost and high robustness.

[0098] Example 2: This invention also provides a memory parking repositioning device, which is mainly used to execute the memory parking repositioning method provided in Embodiment 1 of this invention. The memory parking repositioning device provided in this invention will be described in detail below.

[0099] Figure 7 This is a schematic diagram of a memory parking repositioning device according to an embodiment of the present invention, as shown below. Figure 7 As shown, the device mainly includes: a construction unit 10 and a layered pose correction unit 20, wherein: The building unit is used to detect turning actions in real time when the vehicle is recording and memorizing the path. The pose at the time of detection of the turning action is used as the turning point, and the turning point is used as the boundary. Then, the continuous path is decomposed into multiple topological path segments, and a database of topological path segment sequences is built. The layered pose correction unit is used to calculate the vehicle's pose based on the inertial measurement unit when the vehicle is traveling along the memory path, and then performs the following layered pose correction on the vehicle's pose: Inter-segment global correction: When a vehicle is detected to be moving from one topological path segment to the next, global correction is triggered. The database is queried to obtain the starting pose of the new topological path segment, and the vehicle's pose is updated using the starting pose of the new topological path segment. Intra-segment local correction: When a vehicle travels within a topological path segment, local correction is triggered at preset intervals. The vehicle's pose is filtered and corrected by measuring the lateral distance from the vehicle to the reference path of the current topological path segment.

[0100] In the embodiment of the present application, a device for memory parking repositioning is provided, comprising: when recording a memory path, a turning action is detected in real time, and a pose at the time when the turning action is detected is taken as a turning point, the turning point is taken as a boundary, and then a continuous path is divided into a plurality of topological path segments, and a database of a topological path segment sequence is constructed; when a vehicle travels along the memory path, inertial measurement unit is used for dead reckoning to obtain a pose of the vehicle, and the following layered pose correction is performed on the pose of the vehicle: inter-segment global correction: when it is identified that the vehicle enters a next topological path segment from a current topological path segment, global correction is triggered, a starting pose of the next topological path segment is obtained by querying the database, and the pose of the vehicle is updated by using the starting pose of the next topological path segment; intra-segment local correction: when the vehicle travels in the interior of a topological path segment, local correction is triggered at a preset interval, and the pose of the vehicle is filtered and corrected by measuring a lateral distance of the vehicle to a reference path of the current topological path segment. As can be seen from the above description, in the device for memory parking repositioning, the continuous path is divided into a plurality of topological path segments, so that the complex continuous space positioning problem is converted into a discrete path segment sequence matching problem, the complexity and calculation amount of subsequent positioning search are greatly reduced, the cost is low, and in addition, when it is identified that the vehicle enters a next topological path segment from a current topological path segment, a starting pose of the next topological path segment is obtained by querying the database, and the pose of the vehicle obtained by dead reckoning is directly reset by using the starting pose of the next topological path segment, so that the vehicle starts from zero, and when the vehicle travels in the interior of a topological path segment, the pose of the vehicle is filtered and corrected by measuring a lateral distance of the vehicle to a reference path of the current topological path segment at a preset interval, that is, the effect of periodically clearing a large error and real-time suppressing a small error is realized, and the low-frequency absolute correction and high-frequency relative correction are used together to guarantee the high accuracy and good robustness of the whole path repositioning, and the technical problem that the traditional repositioning method cannot simultaneously meet the requirements of low cost, high accuracy and strong robustness is solved.

[0101] Optionally, the constructing unit is further configured to calculate the angle change of the vehicle turning according to an angle change calculation formula The angle change of the vehicle turning is calculated, wherein, The angle change of the vehicle turning is calculated, wherein, The angular rate output by the gyroscope is represented as The zero offset value corresponding to the gyroscope is represented as; if the angle change of the vehicle turning is greater than a preset angle threshold value and the duration is greater than a preset duration threshold value, it is determined that the vehicle has performed a turning action, and the pose at the time when the turning action is detected is taken as a turning point.

[0102] Optionally, the layered pose correction unit is further configured to: when the turning action is detected again, it is determined that the vehicle enters a next topological path segment from a current topological path segment.

[0103] Optionally, the hierarchical pose correction unit is further configured to: take the system state of the dead reckoning as a predicted value, take the lateral distance as an observation, and correct the system state by a Kalman filtering algorithm to obtain a corrected system state, wherein the system state comprises a pose of the vehicle.

[0104] Optionally, an observation equation of the lateral distance is wherein, represents the lateral distance, represents an observation matrix, , represents a heading angle of a current topological path segment, represents the corrected system state, represents observation noise.

[0105] Optionally, the preset interval comprises a preset distance interval or a preset time interval.

[0106] Optionally, the system state comprises a planar coordinate, a heading angle, a gyroscope zero offset and an accelerometer zero offset of the vehicle, and the database comprises an identifier of each topological path segment, a start pose of each topological path segment and a length of each topological path segment.

[0107] The device provided by the embodiment of the application has the same implementation principle and technical effects as the foregoing method embodiment, and for brevity of description, the part not mentioned in the device embodiment can be referred to the corresponding content in the foregoing method embodiment.

[0108] As shown in Figure 8 The embodiment of the application provides an electronic device 600, which comprises a processor 601, a memory 602 and a bus, the memory 602 stores machine readable instructions executable by the processor 601, when the electronic device is running, the processor 601 and the memory 602 communicate through the bus, and the processor 601 executes the machine readable instructions to execute the steps of the method for memorizing parking relocation.

[0109] Specifically, the memory 602 and the processor 601 can be general memory and processor, which are not specifically limited here, when the processor 601 runs the computer program stored in the memory 602, the method for memorizing parking relocation can be executed.

[0110] The processor 601 can be an integrated circuit chip having a processing capability of signals. In the implementation process, each step of the above method can be completed by the integrated logic circuit of hardware or the instruction in the form of software in the processor 601. The processor 601 described above can be a general processor, including a central processing unit (CPU), a network processor (NP), etc.; can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component. Each method, step and logic block diagram disclosed in the embodiments of the present application can be implemented or executed. The general processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in combination with the embodiments of the present application can be directly embodied as a hardware code processor for execution, or a combination of hardware and software modules in the code processor for execution. The software module can be located in a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium in the art. The storage medium is located in the memory 602, and the processor 601 reads the information in the memory 602, and combines the hardware to complete the steps of the above method.

[0111] Corresponding to the above-mentioned memory parking relocation method, the embodiments of the present application also provide a computer readable storage medium, the computer readable storage medium stores machine executable instructions, when the processor calls and runs the computer executable instructions, the computer executable instructions make the processor run the steps of the above-mentioned memory parking relocation method.

[0112] The device for memory parking relocation provided by the embodiments of the present application can be specific hardware on the equipment or software or firmware installed on the equipment, etc. The device provided by the embodiments of the present application has the same implementation principle and generated technical effects as the foregoing method embodiments. For the sake of brevity, the part of the device embodiment not mentioned in the foregoing method embodiment can be referred to the corresponding content in the foregoing method embodiment. Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system, device and unit described above can be referred to the corresponding process in the foregoing method embodiment, which will not be described here.

[0113] In the embodiments of the present application, it should be understood that the disclosed apparatus and method can be implemented in other manners. The embodiments described above are merely specific implementation manners of the present application, and for example, the division of the units is only a logical function division, and there can be another division manner in actual implementation; for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections between different units, or the among different units, can be indirect couplings or communication connections through some interfaces, communication interfaces, or a combination of other forms, which can be electric, mechanical, or in other forms.

[0114] For example, the flowcharts and block diagrams in the drawings show the possible implementation architectures, functions and operations of the apparatus, method and computer program product according to the embodiments of the present application. In this regard, each block in the flowcharts or block diagrams can represent a module, a program segment or a part of code, which contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions noted in the blocks can occur in different orders from that shown in the drawings. For example, two consecutive blocks can actually be executed substantially in parallel, and sometimes they can be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and the combination of blocks in the block diagrams and / or flowcharts, can be implemented by a dedicated hardware-based system that performs the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0115] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e., they can be located in one place, or can be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the present embodiment.

[0116] In addition, each functional unit in the embodiments of the present application can be integrated into one processing unit, or each unit can exist physically as a separate unit, or two or more units can be integrated into one unit.

[0117] If the functions are implemented in the form of software function units and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application essentially or the parts that contribute to the prior art or parts of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for causing an electronic device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the memory parking relocation method described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various program code storage media.

[0118] It should be noted that similar reference numbers and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. In addition, the terms "first", "second", "third" and the like are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0119] Finally, it should be noted that the above-described embodiments are only specific embodiments of the present application, which are used to illustrate the technical solutions of the present application, but not to limit them. The protection scope of the present application is not limited thereto. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can modify or easily think of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed by the present application, or make equivalent replacements to some of the technical features. These modifications, changes or replacements do not cause the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application. They should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for memory-based parking relocation, characterized in that, include: When the vehicle is recording and memorizing the path, it detects turning actions in real time, and uses the pose at the time of detection of the turning action as the turning point. The turning point is used as the boundary, and then the continuous path is decomposed into multiple topological path segments to build a database of topological path segment sequences. When the vehicle travels along the memory path, the vehicle's pose is obtained by calculating the trajectory based on the inertial measurement unit, and the following layered pose correction is performed on the vehicle's pose: Inter-segment global correction: When it is detected that the vehicle enters the next topological path segment from one topological path segment, global correction is triggered. The database is queried to obtain the starting pose of the new topological path segment, and the pose of the vehicle is updated using the starting pose of the new topological path segment. Intra-segment local correction: When the vehicle travels within a topological path segment, local correction is triggered at preset intervals. The vehicle's pose is filtered and corrected by measuring the lateral distance from the vehicle to the reference path of the current topological path segment.

2. The method according to claim 1, characterized in that, Real-time detection of turning motions, and using the pose at the time of detection of the turning motion as the turning point, including: Calculation formula based on changes in vehicle steering angle Calculate the change in the vehicle's steering angle, where, This indicates the change in the vehicle's steering angle. This represents the angular rate output by the gyroscope. This indicates the zero bias value corresponding to the gyroscope; If the change in the vehicle's steering angle is greater than a preset angle threshold and the duration is greater than a preset duration threshold, then it is determined that the vehicle has performed a turning action, and the position at the time the turning action is detected is taken as the turning point.

3. The method according to claim 1, characterized in that, When it is detected that the vehicle is moving from one topology path segment to the next, the following is included: When a turning action is detected again, it is determined that the vehicle enters the next topological path segment from one topological path segment.

4. The method according to claim 1, characterized in that, The vehicle's pose is filtered and corrected by measuring the lateral distance from the vehicle to the reference path of the current topological path segment, including: Using the system state calculated from the flight path as the predicted value and the lateral distance as the observed value, the system state is filtered and corrected using a Kalman filter algorithm to obtain the corrected system state, wherein the system state includes the vehicle's pose.

5. The method according to claim 4, characterized in that, The equation for the observation of the lateral distance is: ,in, This indicates the lateral distance. Represents the observation matrix. , This indicates the heading angle of the current topological path segment. This indicates the corrected system state. This indicates observation noise.

6. The method according to claim 1, characterized in that, The preset interval includes: preset distance interval or preset time interval.

7. The method according to claim 4, characterized in that, The system status includes: the vehicle's planar coordinates, heading angle, gyroscope zero bias, and accelerometer zero bias; The database includes: the identifier of each topological path segment, the starting pose of each topological path segment, and the length of each topological path segment.

8. A device for memory-based parking repositioning, characterized in that, include: The construction unit is used to detect turning actions in real time when the vehicle is recording and memorizing the path, and to use the pose when the turning action is detected as the turning point, and use the turning point as the boundary to decompose the continuous path into multiple topological path segments, thereby constructing a database of topological path segment sequences. The layered pose correction unit is used to calculate the vehicle's pose based on the trajectory calculation performed by the inertial measurement unit when the vehicle travels along the memory path, and to perform the following layered pose correction on the vehicle's pose: Inter-segment global correction: When it is detected that the vehicle enters the next topological path segment from one topological path segment, global correction is triggered. The database is queried to obtain the starting pose of the new topological path segment, and the pose of the vehicle is updated using the starting pose of the new topological path segment. Intra-segment local correction: When the vehicle travels within a topological path segment, local correction is triggered at preset intervals. The vehicle's pose is filtered and corrected by measuring the lateral distance from the vehicle to the reference path of the current topological path segment.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program executable on the processor, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program thereon, characterized in that, The computer program is executed by the processor to perform the method of any one of claims 1 to 7.

Images