Vehicle navigation path planning method based on vehicle-cloud cooperation

Abstract

The application discloses a vehicle-mounted navigation path planning method based on vehicle-cloud cooperation and concretely relates to the technical field of navigation path planning. Through multi-source monitoring and time synchronization, the vehicle end comprehensively captures implicit deviations such as link delay, path version and trajectory deviation. The statistical structure of the sliding time window enhances the sensitivity and stability to critical fluctuations. Rule mapping and weight processing abstract complex time sequence disturbances into link step-out indicators and switching oscillation indicators, and generate out-of-control situation records with causal markers through weighted aggregation, so that a traceable mapping relationship is established between deviation characteristics and control actions. After the situation records are mapped through the action set and encoded through the command queue, the dynamic adjustment of the path validity period, trajectory anchoring and control gain can be executed in a time-ordered and range-controllable manner, avoiding blind switching under high-frequency oscillation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of navigation path planning, and more particularly, to a vehicle-mounted navigation path planning method based on vehicle-cloud cooperation. BACKGROUND

[0002] In the existing vehicle-cloud cooperative vehicle-mounted navigation path planning method, the driving path of the vehicle usually depends on the real-time calculation and update of the cloud on the global road network to ensure that the navigation result can fully utilize the global traffic information. However, in the actual operation process, the instability of the communication link will cause time misalignment between the uploading of the vehicle state information and the issuing of the path by the cloud, that is, "link step loss". Such step loss makes the path information received by the vehicle end often lag behind the real driving state, thereby forming an execution deviation of the "historical optimal path". More complex is that when the system detects that the path does not match the current state, it will automatically switch between the "cloud priority mode" and the "local priority mode" to ensure the basic driving safety. However, due to the fact that the link delay fluctuation is often in a critical state, such mode switching will exhibit high-frequency oscillation. The step loss and the oscillation interact in this process, the former triggers the latter, and the latter constantly amplifies the influence of the former, eventually leading to the frequent refreshing of the path reference, the instability of the trajectory planning, and the forced adjustment of the control amount by the vehicle controller. The superimposed effect of the two not only destroys the continuity of the navigation path, but also seriously reduces the execution stability of the vehicle, which is manifested as the deterioration of the riding experience and the weakening of the safety margin, and becomes the core defect of the current vehicle-cloud cooperative navigation technology that needs to be solved. SUMMARY

[0003] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present application provide a vehicle-mounted navigation path planning method based on vehicle-cloud cooperation to solve the problems raised in the background art.

[0004] To achieve the above object, the present application provides the following technical scheme:

[0005] The vehicle-mounted navigation path planning method based on vehicle-cloud cooperation comprises the following steps:

[0006] Real-time collection of vehicle-cloud cooperative data information is performed by using a multi-source monitoring means, time stamp is used for unified sorting and marking of each source data to form a continuous original data sequence;

[0007] Time synchronization and event label analysis are applied to the original data sequence, a delay difference, a version deviation and a switching type field are attached to each record to form a structured event record, and feature statistics items are obtained in a multi-layer rolling time window to continuously update the feature statistics items to generate a time window feature table by using a sliding time window;

[0008] The window features are converted into delay, version conflict, switching and trajectory deviation anomaly sub-quantities by rule mapping and weight processing, the delay and version conflict sub-quantities are fused to generate a link out-of-sync index, the switching and trajectory deviation sub-quantities are fused to generate a degraded switching oscillation index, the two types of indexes are weighted and aggregated according to timestamps, and an out-of-control situation record with a main cause marker is generated;

[0009] The out-of-control situation record is mapped into a set of actions, including path validity period adjustment, path anchoring, switching holding period setting and control parameter adjustment, and a structured control command queue is generated by action coding;

[0010] The vehicle reads the control command queue to adjust the control gain, path weight and trajectory anchoring state, simultaneously collects the trajectory position, speed, acceleration and switching behavior, associates the collected data with the abnormal sub-quantities and situation record according to timestamps, and forms a diagnostic log through compression coding.

[0011] In a preferred embodiment, the multi-source monitoring means include vehicle-mounted positioning, inertial measurement unit, acceleration sensor, radar and communication link listening.

[0012] In a preferred embodiment, the vehicle-cloud collaborative data information includes vehicle position, speed, acceleration, cloud-end issued path version and timestamp, link round-trip delay, packet loss rate, local trajectory deviation and path switching event.

[0013] In a preferred embodiment, time synchronization and event label analysis are applied to the original data sequence, a delay difference, version deviation and switching type field are added to each record to form a structured event record; and feature statistics are obtained in a multi-layer rolling time window, the feature statistics are continuously updated by a sliding time window to generate a time window feature table, as follows:

[0014] Each record in the original data sequence is corrected to a unified time reference by a clock deviation compensation algorithm according to the collection timestamp and the master clock reference; a unified time reference field is added to each record after correction to form an event record stream after time correction;

[0015] The event record stream after time correction is applied to a rule matching algorithm to identify event types; an event label field is added to each record, and the event trigger condition and source sensor identifier are recorded to generate an event record stream with semantic annotation;

[0016] For each event record, the received delay difference, version deviation and switching type are calculated, classified into standard categories, and added to the event record to form a structured event record;

[0017] According to the length of time series and business requirements, a plurality of overlapping time windows are set, and the structured event records are mapped into corresponding time windows according to timestamps to form time window event sets;

[0018] Feature statistics are obtained in each time window event set;

[0019] Statistical results generate window feature vectors, each vector corresponding to a time window;

[0020] The feature values are updated by moving average method to form a continuous time window feature table.

[0021] In a preferred embodiment, the feature statistics include delay mean and variance, delay peak value count, expired message proportion, version conflict proportion, switching frequency and duration, and standard deviation of trajectory deviation amplitude.

[0022] In a preferred embodiment, the window features are converted into delay, version conflict, switching and trajectory deviation anomaly sub-quantities by rule mapping and weight processing; the delay and version conflict sub-quantities are fused to generate a link out-of-sync indicator, and the switching and trajectory deviation sub-quantities are fused to generate a degraded switching oscillation indicator; the two types of indicators are weighted and aggregated according to timestamps to generate out-of-control situation records with primary cause labels, as follows:

[0023] The window feature table is subjected to adaptive standardization processing; the standardized feature matrix is rule-mapped into continuous anomaly sub-quantities; the anomaly sub-quantity set is injected with confidence and subjected to source-aware weighting; the weighted anomaly sub-quantity sequence is semantically merged into link out-of-sync original indicators and degraded switching oscillation original indicators; the link out-of-sync original indicators and degraded switching oscillation original indicators time series are subjected to time series smoothing and burst suppression processing; the indicator trend flow is decomposed for contribution to determine the primary cause and is serialized according to timestamps; the out-of-control situation records are written into a time index storage one by one.

[0024] In a preferred embodiment, the out-of-control situation records are mapped into a set of actions, including path validity period adjustment, path anchoring, switching holdover period setting and control parameter adjustment; a structured control command queue is generated by action coding, as follows:

[0025] The out-of-control situation records are parsed and expanded one by one to generate an action template list; the candidate action set is parameterized and scenario-mapped item by item; the executable action list is subjected to conflict detection and priority serialization; the action sequence is time-sequenced and fragmented into deliverable command entries; the command entry sequence is subjected to structured coding and integrity annotation; a security consistency check and executability verification are performed immediately after the command queue is written; a monitoring and reply mechanism is embedded for each command in the queue that is permitted to be delivered.

[0026] The technical effects and advantages of the present application are as follows:

[0027] 1、The present application introduces multi-source monitoring, time window statistics, abnormal sub-quantity fusion and situation-to-action mapping into the vehicle-cloud collaborative navigation, constructs a path planning mechanism with closed-loop adaptive adjustment capability, realizes joint suppression of link step loss and mode oscillation, ensures comprehensive capture of implicit deviations such as link delay, path version and trajectory deviation on the vehicle side through multi-source monitoring and time synchronization, and enhances sensitivity and stability to critical fluctuations through the statistical structure of the sliding time window; the rule mapping and weight processing abstract complex time sequence disturbance into link step loss indicators and switching oscillation indicators, and generate out-of-control situation records with causal markers through weighted aggregation, so as to establish a traceable mapping relationship between deviation characteristics and control actions; after the situation records are mapped through the action set and encoded through the command queue, the dynamic adjustment of path validity period, trajectory anchoring and control gain can be executed under the conditions of time order and controllable range, avoiding blind switching under high-frequency oscillation.

[0028] 2、The present application improves the execution stability while maintaining the continuity of path planning, so that the vehicle can still maintain high control smoothness and safety margin in the complex environment of communication link fluctuation and frequent path switching, thereby significantly improving the practicality and reliability of vehicle-cloud collaborative navigation. BRIEF DESCRIPTION OF DRAWINGS

[0029] In order to facilitate the understanding of those skilled in the art, the present application will be further described below with reference to the accompanying drawings;

[0030] Figure 1 The flowchart of the method of the embodiment of the present application. DETAILED DESCRIPTION

[0031] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0032] Embodiment: Figure 1 The vehicle navigation path planning method based on vehicle-cloud collaboration of the present application is given, including the following steps:

[0033] Real-time acquisition of vehicle-cloud collaborative data information (vehicle position, speed, acceleration, cloud-end issued path version and timestamp, link round-trip delay, packet loss rate, local trajectory deviation and path switching event) is performed through multi-source monitoring means (vehicle-mounted positioning, inertial measurement unit, acceleration sensor, radar and communication link monitoring), and the timestamp is used for unified sorting and marking of each source data to form a continuous original data sequence;

[0034] Time synchronization and event tag parsing are applied to the original data sequence, and fields such as latency difference, version deviation, and switch type are added to each record to form a structured event record; feature statistics are obtained within a multi-layered scrolling time window, and the feature statistics are continuously updated through the sliding time window to generate a time window feature table;

[0035] By using rule mapping and weight processing, window features are transformed into abnormal sub-quantities such as delay, version conflict, switching and trajectory deviation. The delay and version conflict sub-quantities are fused to generate a link out-of-step index, and the switching and trajectory deviation sub-quantities are fused to generate a downgrade switching oscillation index. The two types of indicators are weighted and aggregated according to the timestamp to generate an out-of-control situation record with the main cause label.

[0036] The out-of-control situation is recorded and mapped into a set of actions, including path validity period adjustment, path anchoring, switch hold period setting and control parameter adjustment. A structured control command queue is generated through action coding, recording the trigger time, scope of action and parameter values ​​to ensure that the instructions can be executed one by one in chronological order.

[0037] The vehicle reads the control command queue to adjust the control gain, path weight, and trajectory anchoring status. At the same time, it collects trajectory position, speed, acceleration, and switching behavior. The collected data is associated with abnormal sub-quantities and situation records by timestamp and formed a diagnostic log through compression encoding.

[0038] Real-time vehicle-cloud collaborative data information (vehicle location, speed, acceleration, cloud-based path version and timestamp, link round-trip latency, packet loss rate, local trajectory offset, and path switching events) is collected through multi-source monitoring methods (vehicle positioning, inertial measurement unit, accelerometer, radar, and communication link monitoring). The data from each source is then uniformly sorted and labeled using timestamps to form a continuous raw data sequence, as detailed below:

[0039] Set sampling frequencies (examples: GNSS 1–10 Hz, inertial navigation 100–200 Hz, accelerometer 100–200 Hz, radar 10–50 Hz) and data field templates (longitude / latitude / altitude, heading rate, three-axis acceleration, point cloud or target list, etc.) for positioning, inertial navigation, accelerometer, and radar, respectively, and collect RTT, packet loss count, and bandwidth snapshot for communication links; write the sampling parameters into a local sampling table as the operating configuration for subsequent samplers.

[0040] By accessing the vehicle's master clock and using NTP (a standard network protocol used for time synchronization between computer systems via packet-switched networks) as a time synchronization strategy, a high-precision time stamp is added to each sampling record using sensor-driven hardware timestamps; the time source and local clock deviations are recorded in the synchronization log for subsequent clock correction.

[0041] In the driving layer, each sensor output is sent into the local collection buffer in the message format (including sequence number, timestamp, sensor ID, observation field, sampling quality indication), forming a parallel raw sampling stream; at the same time, the cloud-end issued path message and link metadata are recorded in the same structure at the communication monitoring station.

[0042] Sequence number deduplication and CRC / checksum verification are applied to the raw sampling stream to eliminate duplicate and damaged packets, and the receiving time and verification flag are added to each record to generate the original data sequence after driving verification.

[0043] Time synchronization and event label parsing are applied to the original data sequence, and each record is attached with delay difference, version deviation, switching type, etc. to form a structured event record; and feature statistics are obtained within a multi-layer rolling time window, and the feature statistics are continuously updated through a sliding time window to generate a time window feature table, as follows:

[0044] Each record in the original data sequence is corrected to a unified time reference according to the collection timestamp and master clock reference using a clock deviation compensation algorithm (based on linear interpolation or Kalman correction), eliminating sensor clock drift and link delay fluctuations; a unified time reference field is added to each record to form an event record stream after time correction;

[0045] For the time-corrected event record stream, a rule matching algorithm (such as Isolation Forest, cluster residual detection to form a "dynamic rule" for abnormal patterns, and then converted into a label matching condition) is applied to identify event types, including path version update, switching trigger, position offset anomaly, delay peak, etc.; an event label field is added to each record, and the event trigger condition and source sensor identifier are recorded to generate an event record stream with semantic annotation;

[0046] For each event record, the receiving delay difference (current receiving time minus message timestamp) and version deviation (local path version minus cloud-end issued version) are calculated, and the switching type is classified into standard categories (such as mode switching, path correction, emergency switching); these fields are added to the event record to form a structured event record;

[0047] According to the time series length and business requirements, a multi-layer overlapping time window (such as short window 1-2 seconds, medium window 5-10 seconds, long window 30-60 seconds) is set, and the structured event records are mapped into the corresponding time window according to the timestamp to form a time window event set;

[0048] In each time window event set, feature statistics are obtained, including delay mean and variance, delay peak count, expired message proportion, version conflict proportion, switching frequency and duration, and standard deviation of trajectory deviation amplitude;

[0049] The mean delay is used to measure the average level of the overall transmission delay of the communication link, and is used to characterize the overall timeliness of the vehicle receiving instructions from the cloud. The mean delay is obtained by adding the difference between the reception time and the transmission time of all messages, and then dividing by the total number of messages.

[0050] The delay variance is used to measure the volatility of communication link delay and to characterize delay jitter. It is obtained by squared summing the differences between the delay of each message and the mean delay, and then dividing by the total number of messages.

[0051] The latency peak count is used to measure the frequency of extreme high latency occurrences and to reflect whether there is a sudden blockage in the link; the latency peak count is obtained by counting the number of messages whose latency exceeds a preset latency threshold.

[0052] The expired message ratio is used to measure the proportion of received messages that have expired due to exceeding their validity period, and is used to reflect the degree of failure of path instructions; the expired message ratio is obtained by dividing the number of expired messages by the total number of messages.

[0053] The version conflict ratio is used to measure the degree of inconsistency between the local path version and the cloud path version, and is used to characterize the command synchronization mismatch situation; the version conflict ratio is obtained by counting the number of messages with inconsistent version numbers and dividing it by the total number of messages.

[0054] The switching frequency and duration are used to measure how frequently a vehicle switches between local and cloud paths and how long it lasts, thus characterizing system stability. The switching frequency is determined by recording the number of switching events, while the duration of each switch from start to finish is calculated, and the average value is used as the switching duration.

[0055] The standard deviation of the trajectory deviation is used to measure the volatility of the deviation of the vehicle's actual driving trajectory from the reference path, and is used to reflect the stability of control execution. The standard deviation of the deviation is then used as the volatility index of the trajectory deviation by calculating the lateral offset of each position point from the reference path.

[0056] The statistical results generate window feature vectors, with each vector corresponding to a time window;

[0057] The feature vectors of each time window are continuously calculated with a sliding step size, and the feature values ​​are updated by the moving average method, so that the window features corresponding to each timestamp record change smoothly over time, forming a continuous time window feature table.

[0058] The window features are converted into delay, version conflict, switching and trajectory deviation anomaly sub-quantities by rule mapping and weight processing; the delay and version conflict sub-quantities are fused to generate a link out-of-sync index, and the switching and trajectory deviation sub-quantities are fused to generate a degraded switching oscillation index; the two types of indexes are weighted and aggregated according to timestamps to generate a loss-of-control situation record with a main cause label, as follows:

[0059] Adaptive standardization processing is performed on the window feature table: a normalization method based on a recent historical statistical range is used for each field in the window feature table (upper and lower limits are automatically determined using historical quantile limits or exponential weighted mean / variance), and the normalized feature values are written on each record to form a standardized feature matrix with unified dimensions.

[0060] The standardized feature matrix is rule-mapped into continuous anomaly sub-quantities: the mapping table composed of expert rules and fuzzy membership functions is applied to the standardized feature matrix to map the delay mean, delay variance and delay peak count into delay anomaly components, map the expired message proportion and version conflict proportion into version anomaly components, map the switching frequency and duration into switching anomaly components, and map the standard deviation of trajectory deviation amplitude into trajectory deviation anomaly components; during the mapping process, the trigger rule number, fuzzy membership value and mapping weight are recorded for each sub-quantity to form an anomaly sub-quantity set with evidence chain.

[0061] The anomaly sub-quantity set is injected with confidence and source-aware weighting: the confidence factor is calculated by normalizing the sum of the data source freshness (time decay), sensor reliability (historical accuracy) and message integrity (check flag) of each sub-quantity in the anomaly sub-quantity set, and the confidence factor is multiplied into the corresponding sub-quantity value to generate a weighted anomaly sub-quantity sequence, while the confidence meta-information is attached to each weighted record.

[0062] The data source freshness (time decay) is used to measure the "timeliness" or "freshness" of a piece of observation or message relative to the current time, reflecting the speed of information decay over time; the time difference (called "age") between the sending time of the message and the current time is recorded. The age is mapped to a freshness value between 0 and 1 using an exponential function, for example, wherein is the freshness value, is the current time, is the message occurrence time, is the decay coefficient, controlling the speed of freshness decline. The older the message, the closer the freshness value to 0. When the message is just generated (time difference = 0), the freshness is 1.

[0063] The sensor credibility (historical accuracy) is used to measure the frequency of a sensor or data source reporting correct or consistent information in history, that is, historical reliability; by counting the number of times the sensor is verified as "correct" or "consistent with the reference value" within a specified historical window, and the total number of reports. First, calculate the correct number divided by the total number to obtain the historical accuracy.

[0064] The message integrity (check flag) is used to measure the degree of undamaged and sequential continuity of a single message or a window of messages in the transmission and storage process, reflecting the availability and integrity of the message; by performing integrity check on a single message: if the checksum / CRC passes and the message format is complete, the message integrity is marked as 1, otherwise as 0. For messages in a time window, calculate the window integrity rate as the number of messages passing the check divided by the total number of messages in the window.

[0065] The weighted abnormal sub-quantity sequence is merged according to the semantics to form a link step-out original indicator and a degraded switching oscillation original indicator: the delay class and version class weighted sub-quantities are merged to form a link step-out original indicator according to the context adaptive weight, and the switching class and deviation class weighted sub-quantities are merged to form a degraded switching oscillation original indicator according to the task-aware weight; the contribution proportion of each sub-quantity is recorded during merging, and two types of original indicator time series containing original scores and contribution decomposition are generated.

[0066] The contribution proportion of each sub-quantity is obtained by dividing the specific value of each sub-quantity by the sum of the specific values of all sub-quantities;

[0067] The link step-out original indicator and the degraded switching oscillation original indicator time series are subjected to time series smoothing and burst suppression processing: multi-stage smoothing (short-term exponential smoothing for response and long-term median filtering for anti-pulse) and burst suppression logic (peak trigger retention flag and threshold amplitude reduction) are applied to the two types of original indicator time series respectively, and the smoothed value, short-term slope and peak label are recorded at each time point to form a steady-state processed indicator trend stream.

[0068] The indicator trend stream is decomposed for contribution to determine the main cause and is sequenced by time stamp: at each time point, the contribution degree analysis is performed on the indicator trend stream (the contribution proportion recorded during merging is multiplied by the current smoothed value and normalized), the maximum contribution item is selected as the main cause based on the contribution degree threshold, and the main cause label is generated, and the time stamp, link step-out value (obtained by weighted summation of the link step-out original indicator), degraded oscillation value (obtained by weighted summation of the degraded switching oscillation original indicator), main cause label, and sub-quantity contribution vector are sequenced as the out-of-control situation record with main cause label.

[0069] Write each uncontrolled situation record into a time-indexed storage: perform structured encoding (field name, data type, unit) on the generated uncontrolled situation record, time index and traceable rule reference, and write into a time-ordered situation storage for downstream time retrieval and rule backtracking evidence.

[0070] Map the uncontrolled situation record to a set of actions, including path validity period adjustment, path anchoring, switch holdover period setting, and control parameter adjustment; generate a structured control command queue through action encoding, record trigger time, scope of action, and parameter value, to ensure that the instructions can be executed in chronological order, as follows:

[0071] Parse and expand the action template list for each uncontrolled situation record: call the pre-defined action library and adaptive rules according to the main cause label, generate candidate action templates according to "action type, priority, applicable condition, parameter domain", record trigger evidence and rule number, and form a candidate action set with evidence chain.

[0072] Parameterize and scenario map each item in the candidate action set: use mapping table and scaling strategy to convert abstract templates to specific parameter values (e.g. set path validity period to current time plus mapped conservative time length; linearly map anchoring weight according to trajectory deviation amplitude; calculate switch holdover period by weighting switch frequency and system jitter index; scale control parameters according to current vehicle speed and load), and fill in estimated execution window, scope of action description, and unique command identifier for each item to form an executable action list.

[0073] Conflict detection and priority serialization for executable action list: check actions that are mutually exclusive or mutually affected according to priority, scope overlap, resource occupation, and time dependence, use transactional merging strategy or hierarchical degradation strategy to split, parameter adjust or sequence mutually exclusive actions, generate action sequence after conflict resolution, and attach contribution decomposition and rollback token to each sequence item.

[0074] Time sequence and fragment the action sequence into command entry that can be issued: calculate accurate trigger time, validity period, and timeout rollback strategy according to the estimated execution window of each action and system real-time constraints; perform time interpolation and fragmentation on actions that need to be executed smoothly (split a single large action into multiple continuous small commands to achieve smooth transition), and write dependency and execution precondition in each fragment to form a command entry sequence arranged in chronological order.

[0075] Structural coding and integrity annotation of command entry sequence: encode a fixed field set (command ID, sequence number, version number, trigger time, priority, scope of action, parameter list, preconditions, rollback strategy, timeout threshold, check code and signature) for each command, and write the coded commands into the control command queue one by one according to strict time indexing, and assign quality of service markers and transmission priorities.

[0076] Immediately after writing the command queue, perform safety consistency check and executability verification: call the light dynamics constraint solver and safety rule set for each command in the queue, if the command violates the constraints, adjust the parameters or replace them with alternative commands according to the rules and record the adjustment reasons; verified commands are marked as "allowed to issue", and operation logs with detection evidence are generated and written to the trace storage.

[0077] Embed monitoring and feedback mechanism for each command in the queue that is allowed to issue: bind a unique monitoring identifier, timeout window, retry strategy and dependent completion callback to each command, and serialize these runtime metadata together with the command, ensuring that subsequent commands are read, issued, confirmed or rolled back according to the rollback strategy in time sequence.

[0078] The vehicle reads the control command queue to adjust the control gain, path weight and trajectory anchoring state, and collects trajectory position, speed, acceleration and switching behavior; associate the collected data with abnormal sub-quantities and situation records by timestamp, and form a diagnostic log by compression coding, as follows:

[0079] Through vehicle process polling, the control command queue is obtained and locked, the signature and check code of the queue entry are verified, and the current command context containing command ID, sequence number, trigger time and parameter field is formed after verification.

[0080] The current command context is parsed through the parameter mapper and scaled according to the vehicle operating situation, using speed / lateral acceleration / load factors to linearly or nonlinearly scale command parameters, converting abstract actions into specific controller parameter sets and expected execution windows.

[0081] The controller parameter set is decomposed into parameter time slices by the smoothing processor: for large parameter changes, use linear interpolation or cubic spline interpolation to generate high-frequency control parameter sequences, and for each time slice, add a timestamp, command ID and dependency identifier to form a parameter time sequence.

[0082] The parameter time sequence is issued in real time to the actuator scheduler and verified by the light dynamics constraint solver, and the actual issued execution command and verification result are recorded as execution record items (including actual instructions, timestamps, verification flags and expected response windows) after the scheduler passes the verification.

[0083] The execution record item is associated with the sampling of the sensor sampling subsystem during vehicle operation, position, speed, three-axis acceleration, steering angle, braking force and switching event are written into the track execution sample stream one by one with synchronous time stamp, and the execution record item ID and parameter time slice ID are marked on each sample.

[0084] The track execution sample stream is derived by the online calculation unit in real time to track error, control output amplitude, execution delay and execution abnormality identification, and the derived quantity is packaged together with the original sample as an enhanced execution event stream, each event containing a time stamp, a command context reference, a derived metric and an abnormality annotation.

[0085] The enhanced execution event stream is compressed by the encoder using differential residual prediction and adaptive entropy coding; the key fields (command ID, time stamp, abnormality annotation) are lossless encoded, the high-frequency original sample can be down-sampled according to the threshold or selectively retained based on importance, and the encoding result forms a diagnostic log block together with the integrity hash and signature.

[0086] The present application introduces a whole process of multi-source monitoring, time window statistics, abnormal sub-quantity fusion and situation-to-action mapping in vehicle-cloud collaborative navigation, constructs a path planning mechanism with closed-loop adaptive adjustment capability, realizes joint suppression of link step loss and mode oscillation, and guarantees comprehensive capture of implicit deviations such as link delay, path version and trajectory deviation on the vehicle side through multi-source monitoring and time synchronization; the statistical structure of the sliding time window enhances the sensitivity and stability to critical fluctuations; the rule mapping and weight processing abstract the complex time sequence disturbance into link step loss indicators and switching oscillation indicators, and generate out-of-control situation records with causal markers through weighted aggregation, so that a traceable mapping relationship is established between deviation characteristics and control actions; after the situation record is mapped to the action set and encoded to the command queue, the dynamic adjustment of the path validity period, trajectory anchoring and control gain can be executed under the conditions of time order and controllable range, avoiding blind switching under high-frequency oscillation.

[0087] The present application maintains the continuity of path planning while improving the execution stability, so that the vehicle can still maintain high control smoothness and safety margin in a complex environment of communication link fluctuation and frequent path switching, thereby significantly improving the practicality and reliability of vehicle-cloud collaborative navigation.

[0088] The above formulas are dimensionless numerical calculations, the formulas are obtained by software simulation of a large amount of data to obtain a formula of the latest real situation, and the preset parameters in the formula are set by a person skilled in the art according to the actual situation.

[0089] It should be understood that the size of the sequence number of each process described above in various embodiments of the present application does not mean the order of execution, and the execution order of each process should be determined according to its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0090] The above description is only specific embodiments of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of changes or replacements within the technical scope disclosed by the present application, which should be covered in 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 vehicle navigation route planning method based on vehicle-cloud collaboration, characterized in that: Includes the following steps: Vehicle-cloud collaborative data is collected in real time through multi-source monitoring methods, and the data from each source is uniformly sorted and marked using timestamps to form a continuous raw data sequence. Time synchronization and event tag parsing are applied to the original data sequence, and each record is appended with fields for latency difference, version deviation, and switch type to form a structured event record; feature statistics are obtained within a multi-layered scrolling time window, and the feature statistics are continuously updated through the sliding time window to generate a time window feature table; By using rule mapping and weight processing, window features are transformed into abnormal sub-quantities such as delay, version conflict, switching and trajectory deviation. The delay and version conflict sub-quantities are fused to generate a link out-of-step index, and the switching and trajectory deviation sub-quantities are fused to generate a downgrade switching oscillation index. The two types of indicators are weighted and aggregated according to the timestamp to generate an out-of-control situation record with the main cause label. The out-of-control situation is recorded and mapped into a set of actions, including path validity period adjustment, path anchoring, handover hold period setting and control parameter adjustment, and a structured control command queue is generated through action coding; The vehicle reads the control command queue to adjust the control gain, path weight, and trajectory anchoring status. At the same time, it collects trajectory position, speed, acceleration, and switching behavior. The collected data is associated with abnormal sub-quantities and situation records by timestamp and formed a diagnostic log through compression encoding.

2. The vehicle navigation route planning method based on vehicle-cloud collaboration according to claim 1, characterized in that: The multi-source monitoring methods include vehicle positioning, inertial measurement unit, acceleration sensor, radar, and communication link monitoring.

3. The vehicle navigation route planning method based on vehicle-cloud collaboration according to claim 1, characterized in that: The vehicle-cloud collaborative data information includes vehicle location, speed, acceleration, cloud-deployed path version and timestamp, link round-trip latency, packet loss rate, local trajectory offset, and path switching events.

4. The vehicle navigation route planning method based on vehicle-cloud collaboration according to claim 1, characterized in that: Time synchronization and event tag parsing are applied to the original data sequence, and each record is appended with fields for latency difference, version deviation, and switch type to form a structured event record. Feature statistics are then obtained within multi-level scrolling time windows, and these statistics are continuously updated through the sliding time windows to generate a time window feature table, as detailed below: Each record in the original data sequence is corrected to a unified time base using a clock deviation compensation algorithm based on the acquisition timestamp and the master clock reference; a corrected unified time base field is added to each record to form a time-corrected event record stream; For the time-corrected event log stream, a rule matching algorithm is applied to identify the event type; Add an event tag field to each record, and record the event triggering conditions and source sensor identifier to generate an event record stream with semantic annotation; For each event record, calculate the reception delay difference and version deviation, classify the switching type into a standard category, and append it to the event record to form a structured event record; Based on the time series length and business requirements, multiple overlapping time windows are set up to map structured event records to the corresponding time windows according to timestamps, forming a time window event set; Obtain feature statistics within each time window event set; The statistical results generate window feature vectors, with each vector corresponding to a time window; The feature vectors of each time window are continuously calculated with a sliding step size, and the feature values ​​are updated by the moving average method to form a continuous time window feature table.

5. The vehicle navigation route planning method based on vehicle-cloud collaboration according to claim 4, characterized in that: The feature statistics include mean and variance of delay, peak delay count, proportion of expired messages, proportion of version conflicts, switching frequency and duration, and standard deviation of trajectory deviation.

6. The vehicle navigation route planning method based on vehicle-cloud collaboration according to claim 4, characterized in that: By utilizing rule mapping and weighting, window features are transformed into anomalous sub-quantities such as latency, version conflict, switching, and trajectory deviation. The latency and version conflict sub-quantities are fused to generate a link synchronization failure indicator, and the switching and trajectory deviation sub-quantities are fused to generate a degraded switching oscillation indicator. The two types of indicators are then weighted and aggregated according to timestamps to generate a record of the out-of-control situation with a primary cause marker, as detailed below: Adaptive standardization is performed on the window feature table; the standardized feature matrix rules are mapped to continuous anomalous subquantities; confidence is injected into the set of anomalous subquantities and source-aware weighting is performed; the weighted anomalous subquantity sequences are semantically merged into the original index of link out-of-step and the original index of degradation switching oscillation; the time series of the original index of link out-of-step and the original index of degradation switching oscillation are subjected to time series smoothing and burst suppression processing. Decompose the contribution of the indicator trend stream to determine the main cause and serialize it by timestamp; Record each instance of the out-of-control situation and store it in a time-indexed database.

7. The vehicle navigation path planning method based on vehicle-cloud collaboration according to claim 6, characterized in that: The out-of-control situation is recorded and mapped to a set of actions, including path validity period adjustment, path anchoring, handover hold period setting, and control parameter adjustment; a structured control command queue is generated through action coding, as follows: The system analyzes each record of the out-of-control situation and expands it into a list of action templates; it parameterizes and maps each candidate action set to a scenario; it performs conflict detection and priority serialization on the list of executable actions; it temporals the action sequence and segments it into command entries that can be issued; and it performs structured encoding and integrity annotation on the command entry sequence. Immediately after a command is written to the command queue, a security consistency check and executability verification are performed; a monitoring and acknowledgment mechanism is embedded for each authorized command in the queue.

Images