Multi-target vehicle navigation system and method based on vehicle networking road condition and cloud map data

Abstract

The application discloses a multi-target vehicle-mounted navigation system and method based on vehicle networking road conditions and cloud map data, and particularly relates to the technical field of vehicle-mounted navigation, wherein real-time traffic flow, road events and environmental perception information of vehicle networking are integrated, and cloud map and historical evolution data are combined to generate a global road condition situation with time sequence; in a multi-dimensional path search space, evolution reasoning based on conflict weight can generate a Pareto front solution set covering different optimization targets, obtain redundant inflation features and transfer them to a vehicle end; a vehicle end real-time monitoring processor occupies, task queue accumulation and delay fluctuation, extracts resource exhaustion information, and constructs a navigation decision delay degradation model through interactive mapping of the redundant inflation features and the resource exhaustion information to generate a quantifiable delay degradation index; and the path screening depth and priority are dynamically adjusted according to the delay degradation index.

Description

Technical Field

[0001] This invention relates to the field of vehicle navigation technology, and more specifically, to a multi-target vehicle navigation system and method based on vehicle network traffic conditions and cloud map data. Background Technology

[0002] In existing multi-objective in-vehicle navigation systems based on vehicle-to-everything (V2X) road conditions and cloud-based map data, the cloud typically generates a set of Pareto front paths (Pareto refers to the concept of Pareto optimality; the Pareto front path set represents the set of all solutions that cannot be dominated by other solutions) based on real-time road conditions, historical statistics, and user preferences to provide diverse navigation options. However, when there is a strong conflict between the objectives to be optimized, the number of Pareto front solutions generated by the cloud may increase significantly, resulting in solution space inflation. During vehicle-side execution, because computing resources such as CPUs / GPUs need to support autonomous driving, perception, and interaction modules simultaneously, limited computing power makes it difficult to fully evaluate and quickly filter a large number of candidate paths in a short time, leading to navigation decision processing delays. There is a close coupling between solution space inflation and limited local computing resources: the former increases the vehicle-side computing load, while the latter limits the real-time processing capability of candidate paths, thus exacerbating the response lag of the navigation module. Ultimately, the vehicle may be unable to select or execute the truly optimal path provided by the cloud, and is forced to adopt a simplified or approximate path scheme, resulting in insufficient guarantee of multi-objective optimization effectiveness and real-time performance. Summary of the Invention

[0003] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a multi-target vehicle navigation system and method based on vehicle network road conditions and cloud map data, so as to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] The multi-target in-vehicle navigation method based on vehicle network traffic conditions and cloud map data includes the following steps:

[0006] By integrating vehicle-to-everything (V2X) traffic flow, road events, and environmental perception information, and combining cloud maps with historical evolution data, a global road condition status with time series is generated.

[0007] Analyze the global road conditions in the spatiotemporal dimension, extract multi-objective factors, and generate a multi-dimensional path search space by combining road topology;

[0008] Evolutionary reasoning is performed on the multidimensional path search space based on conflict weights to generate a Pareto front solution set;

[0009] The redundancy and complexity of the Pareto front solution set are analyzed to form a redundancy expansion feature, which is then transmitted to the vehicle-side computing engine.

[0010] Real-time tracking of vehicle-side processor usage, task queue backlog, and latency fluctuations; extraction of resource depletion information.

[0011] By interactively mapping the redundancy expansion characteristics with resource depletion information, a vehicle-side navigation decision delay degradation model is constructed, generating a vehicle-side navigation decision delay degradation index.

[0012] The screening depth and priority are dynamically adjusted based on the vehicle-side navigation decision delay degradation index to suppress the computational burden caused by redundant solutions and output an executable path.

[0013] In a preferred embodiment, vehicle-to-everything (V2X) traffic flow, road events, and environmental perception information are integrated, and combined with cloud maps and historical evolution data to generate a global road condition status with time series, as follows:

[0014] Collect vehicle network traffic flow, road events, roadside and vehicle-mounted environmental perception, as well as cloud maps and historical evolution records, and merge them into a raw multi-source time-series stream with source identification;

[0015] The original multi-source time-series stream is aligned with a time base and compensated for delay to obtain a time-series corrected stream with a unified timestamp.

[0016] Spatial semantic mapping is performed on the time series correction stream, and the location information is connected with the cloud map topology to map it into a semantic spatiotemporal segment stream based on road segment / intersection / grid;

[0017] Spatiotemporal interpolation and short-term reconstruction are performed on missing or anomalous segments of the weighted spatiotemporal fragment stream. Continuity is restored by using adjacent segment backfilling, historical template matching and pattern regression to obtain a coherent time series matrix.

[0018] The coherent temporal matrix and road event information are fused in parallel within a sliding window to form a labeled temporal situation with event intensity and impact range annotations;

[0019] The labeled time series situation is subjected to spatiotemporal compression and index extraction at multiple scales. The key sequences of traffic rate, traffic density and event delay are calculated, and an indexable multi-scale time series index set is output.

[0020] The multi-scale time series index set applies short-term forecasting and uncertainty assessment strategies to generate forecast time series and confidence intervals within future windows;

[0021] The predicted time series and confidence interval are associated with and versioned with the original map topology to form a global traffic situation package containing timestamps, spatial indexes, confidence levels, event labels, and impact ranges.

[0022] In a preferred embodiment, the global road condition situation is analyzed in the spatiotemporal dimension, multi-objective factors are extracted, and a multi-dimensional path search space is generated by combining road topology, as detailed below:

[0023] Using the global road condition situation package as input, the situation package is spatiotemporally sliced ​​according to a preset time scale and spatial granularity to generate a set of labeled road segment time units, and an event label and original data confidence field are added to each unit;

[0024] Spectral decomposition and abrupt change detection are performed on the time unit set of road segments to extract the original factors of traffic rate, congestion wave propagation rate, event delay, traffic variance and confidence standard deviation, and output a vector set that constitutes multi-objective factors;

[0025] Normalize and weight the multi-objective factor vector set to obtain the objective factor matrix;

[0026] A temporally weighted road network tensor is constructed by mapping the target factor matrix to road topology elements one-to-one and aligning them with time.

[0027] The temporally weighted road network tensor is subjected to segment combination and attribute accumulation on the time-expanded network according to connectivity rules, generating an ordered sequence of path attribute vectors;

[0028] By performing diversity-preserving clustering and target dominance analysis on the path attribute vector sequence in the vector space, a candidate multidimensional path search space representation is obtained.

[0029] In a preferred embodiment, the redundancy and complexity of the Pareto front solution set are analyzed to form a redundancy expansion feature, which is then transmitted to the vehicle-side computing engine, as follows:

[0030] Receive the Pareto front solution set and the target vector and confidence information of each path, and construct a high-dimensional vector space representation of the solution set according to the target dimension;

[0031] Density estimation is performed on the high-dimensional vector space, and the local clustering density of each solution in the solution set is calculated by the nearest neighbor distance method to generate the spatial distribution feature vector of the solution.

[0032] Based on the mutual information between target vectors, the target conflict degree of each solution is calculated, and a conflict feature description of the solution is generated.

[0033] The information entropy of the solution set is calculated on the target dimension, and the information entropy value is used as an indicator of the structural compactness of the solution set;

[0034] The candidate set size, cluster density, target conflict degree, and information entropy are integrated into a multidimensional redundant expansion feature vector.

[0035] The multidimensional redundancy expansion features are transmitted to the vehicle-side computing engine through efficient differential compression and a publish-subscribe mechanism.

[0036] In a preferred embodiment, the vehicle-side processor usage, task queue backlog, and latency fluctuations are tracked in real time to extract resource depletion information, as detailed below:

[0037] A lightweight acquisition thread is established on the vehicle side to periodically acquire processor utilization, thread utilization, and core frequency, along with operating temperature and power consumption status, forming an initial computing power monitoring stream.

[0038] Continuous monitoring is performed on the vehicle-side task queue to record the length of tasks to be processed, the priority of task types, the distribution of processing time, and changes in queue backlog, generating a task load time stream.

[0039] The collected computing power monitoring stream and task load time series stream are synchronized and aligned with the time base to form a resource status matrix under a unified timestamp;

[0040] Extract resource depletion information from the resource status matrix, including transient fluctuations in processor utilization, rate of change in queue length, bursts in task latency, and combined stress values ​​of temperature and power consumption.

[0041] In a preferred embodiment, the redundant expansion characteristics are interactively mapped with resource depletion information to construct a vehicle-side navigation decision delay degradation model and generate a vehicle-side navigation decision delay degradation index. The formula based on the vehicle-side navigation decision delay degradation model is as follows: ,in The vehicle-side navigation decision delay degradation index, For the redundancy expansion characteristics and resource depletion information, the first Dimensionless values ​​of each indicator. For the first The weighting coefficients of each indicator, and Greater than 0, .

[0042] In a preferred embodiment, a multi-target vehicle navigation system based on vehicle network road conditions and cloud map data includes a global road condition generation module, a multi-target factor extraction and path space construction module, a multi-target evolution reasoning module, a redundancy expansion feature analysis module, a vehicle-side resource status monitoring module, a delay degradation modeling and index generation module, and a dynamic filtering and priority adjustment module.

[0043] The global traffic situation generation module is used to integrate vehicle network traffic flow, road events and environmental perception information, and combine cloud maps and historical evolution data to generate a global traffic situation with time series.

[0044] The multi-objective factor extraction and path space construction module is used to analyze the global road condition situation in the spatiotemporal dimension, extract multi-objective factors, and generate a multi-dimensional path search space by combining road topology.

[0045] The multi-objective evolutionary reasoning module is used to perform evolutionary reasoning on the multi-dimensional path search space based on conflict weights, and generate Pareto front solution sets;

[0046] The redundancy expansion feature analysis module is used to analyze the redundancy and complexity of the Pareto front solution set, form redundancy expansion features, and transmit them to the vehicle-side computing engine.

[0047] The vehicle-side resource status monitoring module is used to track the vehicle-side processor usage, task queue backlog, and latency fluctuations in real time, and to extract resource depletion information.

[0048] The delay degradation modeling and index generation module is used to interactively map the redundancy expansion characteristics with resource depletion information, construct the vehicle navigation decision delay degradation model, and generate the vehicle navigation decision delay degradation index.

[0049] The dynamic filtering and priority adjustment module is used to dynamically adjust the filtering depth and priority based on the vehicle-side navigation decision delay degradation index, suppress the computational burden caused by redundant solutions, and output an executable path.

[0050] The technical effects and advantages of this invention are as follows:

[0051] 1. This invention integrates real-time traffic flow, road events, and environmental perception information from the Internet of Vehicles (IoV) and combines cloud maps with historical evolution data to generate a global road condition situation with time series, enabling detailed characterization and dynamic analysis of multi-objective path factors. In the multi-dimensional path search space, evolutionary reasoning based on conflict weights can generate Pareto front solution sets covering different optimization objectives. Simultaneously, it comprehensively analyzes the redundancy, complexity, cluster density of candidate paths, target conflict degree, and information entropy of the solution set, forming a redundancy expansion feature and transmitting it to the vehicle. The vehicle monitors processor usage, task queue accumulation, and latency fluctuations in real time, extracting resource depletion information. Through the interactive mapping between the redundancy expansion feature and resource depletion information, a navigation decision latency degradation model is constructed, generating a quantifiable latency degradation index. Based on the latency degradation index, the path selection depth and priority are dynamically adjusted to effectively suppress redundant solutions while ensuring the priority processing of high-confidence critical paths, thereby outputting executable paths that balance multi-objective optimization and real-time performance.

[0052] 2. This invention can significantly reduce the impact of solution space expansion on vehicle-side computing load, improve path selection and decision response speed, ensure the real-time performance and optimality of the navigation module in multi-target severe conflict scenarios, enhance vehicle-side resource utilization efficiency, realize intelligent, dynamic and traceable path selection, and comprehensively improve the performance and reliability of multi-target vehicle navigation systems. Attached Figure Description

[0053] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings;

[0054] Figure 1 This is a flowchart of the method in Embodiment 1 of the present invention;

[0055] Figure 2 This is a flowchart of the system in Embodiment 2 of the present invention. Detailed Implementation

[0056] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0057] Example 1: Figure 1 The present invention provides a multi-target in-vehicle navigation method based on vehicle network traffic conditions and cloud map data, comprising the following steps:

[0058] By integrating vehicle-to-everything (V2X) traffic flow, road events, and environmental perception information, and combining cloud maps with historical evolution data, a global road condition status with time series is generated.

[0059] Analyze the global road conditions in the spatiotemporal dimension, extract multi-objective factors, and generate a multi-dimensional path search space by combining road topology;

[0060] Evolutionary reasoning is performed on the multidimensional path search space based on conflict weights to generate a Pareto front solution set;

[0061] The redundancy and complexity of the Pareto front solution set are analyzed to form a redundancy expansion feature, which is then transmitted to the vehicle-side computing engine.

[0062] Real-time tracking of vehicle-side processor usage, task queue backlog, and latency fluctuations; extraction of resource depletion information.

[0063] By interactively mapping the redundancy expansion characteristics with resource depletion information, a vehicle-side navigation decision delay degradation model is constructed, generating a vehicle-side navigation decision delay degradation index.

[0064] The screening depth and priority are dynamically adjusted based on the vehicle-side navigation decision delay degradation index to suppress the computational burden caused by redundant solutions and output an executable path.

[0065] By integrating vehicle-to-everything (V2X) traffic flow, road events, and environmental perception information, and combining cloud-based maps with historical evolution data, a global road condition status with time series is generated, as follows:

[0066] Collect vehicle-to-everything (V2X) traffic flow, road events, roadside and vehicle-mounted environmental perception, cloud maps and historical evolution records, and merge them into a raw multi-source time-series stream with source identification.

[0067] The original multi-source time-series stream is aligned with a time base and compensated for delay. An adaptive time window is used to align asynchronously arriving data, resulting in a time-series corrected stream with a unified timestamp.

[0068] Spatial semantic mapping is performed on the time series correction stream, and the location information is connected with the cloud map topology to map it into a semantic spatiotemporal segment stream based on road segments / intersections / grids, with spatial indexes attached;

[0069] The specific steps of the spatial semantic mapping are as follows:

[0070] The system reads vehicle positioning data and environmental coordinate information from the time-series correction stream, establishes a coordinate docking relationship with the cloud map topology, and generates an initial spatial alignment stream.

[0071] The initial spatial alignment flow is matched with road nodes (road segment centerline, intersection geometry point, region grid boundary), and spatial binding is completed based on nearest neighbor and topological constraints to obtain candidate spatial fragment flow.

[0072] The candidate spatial fragment stream performs topology consistency checks, removes fragments that do not match the road topology, corrects out-of-bounds and drift points, and outputs a structured road-related fragment stream.

[0073] The road-related fragment stream is overlaid with road semantic tags (functional level, speed limit attribute, number of lanes, signal control method, etc.) to expand it into a multi-attribute semantic spatiotemporal fragment stream.

[0074] Semantic spatiotemporal fragment streams are segmented and aggregated according to three granularities: road segment, intersection, and grid, forming a hierarchical set of semantic fragments.

[0075] Assign spatial indexes and timestamp indexes to each segment in the hierarchical semantic segment set to establish a dual index system, enabling fast cross-temporal and spatial retrieval and tracking, and obtaining the final semantic spatiotemporal segment stream.

[0076] The semantic spatiotemporal segment stream is subjected to quality assessment and confidence assignment. Based on arrival rate, sensor accuracy and historical consistency, packet loss, offset and anomaly points are identified, and a weighted spatiotemporal segment stream with confidence weight is generated.

[0077] The specific steps for quality assessment and confidence assignment of semantic spatiotemporal segment streams are as follows:

[0078] Receive semantic spatiotemporal segment streams and split them into several source substreams according to the source identifier, ensuring that each substream carries the source ID, timestamp, and location index.

[0079] The arrival rate and delay distribution of each sub-stream are statistically analyzed, the percentage of packets arriving and the average delay within a given time window are calculated, and an arrival quality description vector is generated.

[0080] Based on the sensor source information, the corresponding static accuracy profile (calibration error, historical drift rate) is retrieved, and the difference between the static accuracy profile and the real-time observation data (actual observation calibration error and drift rate) is compared as a measure of sensor reliability.

[0081] Using historical time series of the same locus as a baseline, a historical consistency test is performed: the deviation of the current observation from the historical distribution, the mutation frequency and the pattern similarity are compared, and abnormal patterns of long-term shift or short-term mutation are identified and the consistency score is marked.

[0082] The arrival quality description vector, sensor reliability metric, and historical consistency score are fused with configurable weights, and the initial confidence score for each subflow is obtained by regularization and normalization transformation.

[0083] Anomaly detection and confidence recalibration are performed on the confidence score: When the score shows inconsistency across sources or significant conflict with neighboring spatiotemporal segments, the misjudgment is reduced by neighborhood weighting or back to the historical baseline, and the corrected confidence value is output.

[0084] The corrected confidence values ​​are mapped to spatiotemporal segment weights and an uncertainty interval (confidence interval) is attached.

[0085] The downstream weighted fusion and filtering are performed using weighted spatiotemporal segments as input, and the output is a weighted spatiotemporal segment stream with confidence weights. The source traceability information and confidence score evolution record are retained for each segment for subsequent interpolation, prediction and decision-making.

[0086] Spatiotemporal interpolation and short-term reconstruction are performed on missing or abnormal segments of the weighted spatiotemporal fragment stream. Continuity is restored by using adjacent segment backfilling, historical template matching and pattern regression, and a repaired coherent time series matrix is ​​generated.

[0087] The coherent temporal matrix and road event information are fused in parallel within a sliding window. Through event time window aggregation, influence radius propagation, and causal relationship confirmation, a labeled temporal situation with event intensity and influence range annotations is formed.

[0088] The specific steps for forming the labeled time-series situation are as follows:

[0089] Slide a time window across the coherent time series matrix with a fixed step size, extract the corresponding road segment time series segments according to the start and end times of the window, and output the road segment time series subset within the window.

[0090] Within the same time window, road event records are filtered by timestamp, and event time sets are formed according to the event occurrence time and type. The event locations are then spatially indexed, and a subset of events within the window is output.

[0091] Perform event intensity quantization on a subset of events within the window according to event type, obtain the event duration of different event types as the original event intensity value, and output event entries with intensity labels.

[0092] The time-series subset of road segments within the window is expanded into flow propagation channels according to flow direction and topological adjacency. Directed influence paths are constructed based on flow direction and velocity gradient to form a set of propagation paths that carry the impact of the event.

[0093] The impact attenuation calculation is performed along the propagation path. The original intensity of the event is attenuated by the three factors of distance attenuation, time delay attenuation and traffic dilution. The event perception intensity contribution value at each road segment is obtained (by weighted summation of the original event intensity value, distance correction value, time delay correction value and traffic correction value). The road segment-level event perception contribution matrix is ​​output.

[0094] The causal association test is performed on the road segment-level event perception contribution matrix and the time-series subset of road segments within the window in a time-synchronized manner. The time lag correlation, Granger causality, or simplified causal graph test is used to confirm the significant association between the occurrence of the event and the time-series anomaly of the road segment, and the causal confidence level is marked.

[0095] For road segments with significant causal confidence, a weighted summation is performed based on the contribution value of event perception intensity and causal confidence to generate a candidate set of event intensity score and influence radius for each road segment within the time window; the candidate set is then pruned according to topological connectivity and threshold to obtain the final influence range.

[0096] The event intensity score and impact range of each road segment are integrated with the original time series indicators, and the road segment sequence values ​​at the corresponding time are inserted or overwritten according to the time series to form a time series record row and column with event intensity and impact range labeling.

[0097] The labeled time series records are spliced ​​together in window order and smoothed as needed to output the final labeled time series status with event intensity and impact range labels, which can be used for subsequent path trade-off and impact propagation analysis.

[0098] The labeled time-series situation is subjected to spatiotemporal compression and index extraction at multiple scales (road segment / sub-region / administrative region), and key sequences such as traffic rate, traffic density, and event delay are calculated to output an indexable multi-scale time-series index set.

[0099] The multi-scale time series index set applies short-term forecasting and uncertainty assessment strategies, and uses lightweight time series forecasting and Bayesian or confidence update methods to generate forecast time series and confidence intervals within the future window.

[0100] The predicted time series and confidence interval are associated with and versioned with the original map topology to form a global traffic situation package containing timestamps, spatial indexes, confidence levels, event labels, and impact ranges.

[0101] The global traffic situation package is distributed to the edge and vehicle terminals through differential updates and subscription mechanisms, forming a time-series global traffic situation stream that can be streamed and subscribed to, for use in subsequent multi-objective path generation and online decision-making.

[0102] Within the spatiotemporal dimension, the global road condition situation is analyzed, multi-objective factors are extracted, and a multi-dimensional path search space is generated by combining road topology, as detailed below:

[0103] Using the global traffic situation package as input, the situation package is spatiotemporally sliced ​​according to a preset time scale and spatial granularity to generate a set of road segment time units identified as (road segment / intersection / grid, time window), and an event label and original data confidence field are added to each unit;

[0104] The spectral decomposition and abrupt change detection are performed on the time unit set of road segments to extract original factors such as traffic rate, congestion wave propagation rate, event delay, traffic variance and confidence standard deviation, and output a vector set that constitutes multi-objective factors.

[0105] Normalize and weight the multi-objective factor vector set to obtain the objective factor matrix;

[0106] The target factor matrix is ​​mapped one-to-one with road topology elements (edges, nodes, lane connectivity) and aligned with time to construct a temporally weighted road network tensor, so that each edge has a complete target attribute vector in each time window.

[0107] The temporally weighted road network tensor is subjected to segment combination and attribute accumulation on the time-expanded network according to connectivity rules, generating an ordered sequence of path attribute vectors;

[0108] The path attribute vector sequence is subjected to diversity-preserving clustering and target dominance analysis in the vector space. Redundant solutions are compressed according to the approximate dominance pruning strategy and the representativeness preservation strategy, and a candidate multidimensional path search space representation is generated (Figure / Tensor Structure).

[0109] Evolutionary reasoning is performed on the multidimensional path search space based on conflict weights to generate the Pareto front solution set, specifically including the following steps:

[0110] Step A1: Using the multidimensional path search space and conflict weight matrix as input, construct an initial solution set representation according to the time window, map the path attribute vector sequence to the representation vector of the population individual and write it into the candidate pool.

[0111] Step A2: Perform parallel target evaluation on each representation vector in the candidate pool, calculate the weighted target vector of each target under the effect of conflict weight, and record the evaluation results and source confidence together as an individual fitness description.

[0112] Step A3: Based on the non-dominated relationship, perform preliminary non-dominated stratification on the parallel evaluation results, label the non-dominated level and local crowding index of each individual, and form the initial frontier layer and candidate replacement layer;

[0113] Step A4: Parent individuals are extracted from the initial frontier layer and the reserve layer according to the adaptive selection pressure. The selection pressure is determined by the conflict weight matrix and the individual confidence, so that the diversity under high conflict objectives is preserved and individuals with low confidence are deweighted.

[0114] Step A5: Perform crossover and mutation operations based on path connectivity on the extracted parent generation. Crossover fuses parent generation features in units of time-coherent segments, and mutation uses local time window offset as the operator to generate child generation solutions and write them into the candidate pool.

[0115] Step A6: Perform local refinement on the newly generated offspring: Use short-range heuristic search to explore improved fragments in the neighborhood of the original path, and compare the improved results with the original generation. Add the improved fragments with high confidence and better target performance to the pool.

[0116] Step A7: Merge the parent generation, child generation and replacement layer according to non-dominated level and crowding degree, and apply the representative retention strategy: sample and retain representative solutions in high-density clusters according to the diversity priority criterion, eliminate approximately redundant solutions, and at the same time apply weak retention to boundary solutions to maintain the solution set boundary.

[0117] Step A8: Perform uncertainty propagation in the merged solution set: accumulate the confidence of each individual through the path components to obtain the confidence interval of each path on each target, and adjust the sorting and retention probability of the solutions accordingly.

[0118] Step A9: Repeat steps A4 to A8, and perform iterative control based on convergence criteria (stability of non-dominated layer, crowding change threshold, or iterative budget). Stop evolution when convergence is achieved.

[0119] Step A10: Extract the non-dominated optimal layer from the final merged solution set as the Pareto front solution set, and add annotation information to each solution in the front: target vector, confidence interval, local density index, representative cluster identifier and generation timestamp.

[0120] The redundancy and complexity of the Pareto front solution set are analyzed to form a redundancy expansion feature, which is then transmitted to the vehicle-side computing engine, as follows:

[0121] Receive the Pareto front solution set and the target vector and confidence information of each path, and construct a high-dimensional vector space representation of the solution set according to the target dimension;

[0122] Density estimation is performed on the high-dimensional vector space, and the local clustering density of each solution in the solution set is calculated by the nearest neighbor distance method to generate the spatial distribution feature vector of the solution.

[0123] Based on the mutual information between target vectors, the target conflict degree of each solution is calculated, and a conflict feature description of the solution is generated.

[0124] The degree of target conflict in the solution is quantified by the mutual information value between target vectors;

[0125] The information entropy of the solution set is calculated on the target dimension to measure the effective diversity and representativeness of the solution set, quantify the proportion of paths that truly provide decision-making information, and use the information entropy value as an indicator of the structural tightness of the solution set.

[0126] The candidate set size, cluster density, target conflict degree, and information entropy are integrated into a multidimensional redundant expansion feature vector.

[0127] The multidimensional redundancy expansion features are transmitted to the vehicle-side computing engine through efficient differential compression and a publish-subscribe mechanism.

[0128] Real-time tracking of vehicle-side processor usage, task queue backlog, and latency fluctuations; extraction of resource exhaustion information, as detailed below:

[0129] A lightweight acquisition thread is established on the vehicle side to periodically acquire processor utilization, thread utilization, and core frequency, along with operating temperature and power consumption status, forming an initial computing power monitoring stream.

[0130] Continuous monitoring is performed on the vehicle-side task queue to record the length of tasks to be processed, the priority of task types, the distribution of processing time, and changes in queue backlog, generating a task load time stream.

[0131] The collected computing power monitoring stream and task load time series stream are synchronized and aligned with the time base to form a resource status matrix under a unified timestamp;

[0132] Extract resource depletion information from the resource status matrix, including transient fluctuations in processor utilization, rate of change of queue length, bursts in task latency, and combined stress values ​​of temperature and power consumption.

[0133] The transient fluctuation value of processor utilization is used to measure the severity of fluctuations in computing load over a short period of time. It characterizes the dispersion of processor utilization (CPU or GPU) relative to its short-term mean within a short time window, reflecting the risk of sudden load changes. The transient fluctuation value is obtained by first calculating the arithmetic mean of the processor utilization samples within a sliding time window of length W; then calculating the arithmetic mean (variance) of the squared differences between the samples and the mean; taking the square root of the variance yields the standard deviation; and dividing the standard deviation by the window mean gives the transient fluctuation value of processor utilization.

[0134] The queue length change rate is used to measure the speed at which the task queue grows or shrinks, reflecting the rate of change in the number of tasks to be processed, and capturing queue accumulation trends and sudden growth. The instantaneous change rate is obtained by taking the difference between the queue length at time point t and the queue length at time point t−Δt within an interval Δt between two adjacent time points. The exponentially weighted moving average (EWMA) is applied to the instantaneous change rate sequence within a time window of length W, and the output is used as the current queue length change rate.

[0135] The task delay burst amplitude is used to measure the peak amplitude that occurs in task processing delay, that is, the deviation of the high end of the task response time distribution from the center position within the observation window, representing the impact of the burst delay event on real-time performance. By obtaining a task delay sample set within a time window of length W, the median of the sample is first calculated as the baseline; then the 95th percentile (or 99th percentile, depending on the sensitivity to peaks) of the sample set is calculated; the difference between the high percentile and the median is the task delay burst amplitude.

[0136] The temperature-power consumption joint stress value is used to measure the combined depletion tendency of heat and power consumption on the processor's sustained performance. It is obtained by normalizing the current temperature value according to the ratio between it and the safe upper limit (or historical peak value), obtaining the temperature normalization ratio; normalizing the current power consumption according to the ratio between it and the nominal maximum power consumption, obtaining the power consumption normalization ratio; and taking a weighted average of the temperature normalization ratio and the power consumption normalization ratio according to a pre-set weight to obtain the temperature-power consumption joint stress value.

[0137] By standardizing real-time resource depletion information, an indicator stream can be output that can be directly used for navigation decision delay degradation modeling, ensuring that the vehicle-side can quantify operability constraints and guide route selection under computing power constraints.

[0138] By interactively mapping the redundancy expansion characteristics with resource depletion information, a vehicle-side navigation decision delay degradation model is constructed, generating a vehicle-side navigation decision delay degradation index.

[0139] The process involves interactively mapping redundancy expansion characteristics with resource depletion information to construct a vehicle-side navigation decision delay degradation model and generate a vehicle-side navigation decision delay degradation index. The formula underlying the vehicle-side navigation decision delay degradation model is as follows: ,in The vehicle-side navigation decision delay degradation index, For the redundancy expansion characteristics and resource depletion information, the first Dimensionless values ​​of each indicator. For the first The weighting coefficients of each indicator, and Greater than 0, ;

[0140] It should be noted that the above formulas are all dimensionless calculations. Commonly used methods for removing dimensions include Min-Max normalization and Z-Score standardization, which will not be elaborated here. The settings can be adjusted according to the actual situation. For example, the expert weighting method can be adopted, which involves inviting experts in relevant fields to determine the weight coefficients of each indicator through professional opinion surveys and comprehensive evaluations.

[0141] Based on the vehicle-side navigation decision latency degradation index, the screening depth and priority are dynamically adjusted to suppress the computational burden caused by redundant solutions, and an executable path is output as follows:

[0142] The vehicle-side navigation decision delay degradation index is mapped to the screening depth and priority level to generate a graded screening baseline.

[0143] In an optional example, the vehicle-side navigation decision latency degradation index is mapped to filtering depth and priority levels, for example, setting the threshold sequence as {0.2, 0.5, 0.8}:

[0144] When the vehicle navigation decision delay degradation index is less than 0.2, it is determined to be in the normal range, and the maximum filtering depth and the lowest priority level are maintained accordingly.

[0145] When the vehicle navigation decision delay degradation index is between 0.2 and 0.5, it is determined to be in the mild degradation range, and the corresponding filtering depth is reduced by one layer and switched to the medium-low priority level.

[0146] When the vehicle navigation decision delay degradation index is between 0.5 and 0.8, it is determined to be in the moderate degradation range, and the corresponding filtering depth is reduced by two layers and switched to the medium-high priority level.

[0147] When the vehicle navigation decision delay degradation index exceeds 0.8, it is determined to be in the severe degradation range, and the corresponding filtering depth is compressed to the minimum and switched to the highest priority level.

[0148] Candidate set compression is driven by a hierarchical screening baseline, forming a combined set of density sampling solutions and boundary-preserving solutions.

[0149] A hierarchical approximate evaluation is performed on the combined set to obtain the coarse ranking result and the preferred refined subset.

[0150] The priority subset is first refined and then weighted to generate a priority queue containing confidence intervals and target weights.

[0151] The priority queue triggers probabilistic sampling and resampling verification to obtain the dimensionality-reduced redundancy suppression solution set.

[0152] The redundancy suppression solution set is protected by edge unloading and recycling to form a hybrid computation solution set.

[0153] Incremental merging of the mixed computation solution set yields a simplified solution set containing confidence reduction and update labels.

[0154] The simplified solution set is subjected to a safety baseline verification to generate a set of candidate paths that meet the safety constraints.

[0155] Perform multi-objective weighted backtracking verification on the candidate path set to produce a final path sequence after deviation correction, and write the final path sequence into the execution queue.

[0156] This invention integrates real-time traffic flow, road events, and environmental perception information from the Internet of Vehicles (IoV), and combines cloud maps with historical evolution data to generate a global road condition situation with time series, enabling detailed characterization and dynamic analysis of multi-objective path factors. In the multi-dimensional path search space, evolutionary inference based on conflict weights can generate Pareto front solution sets covering different optimization objectives. Simultaneously, it comprehensively analyzes the redundancy, complexity, cluster density of candidate paths, target conflict degree, and information entropy of the solution sets, forming a redundancy expansion feature and transmitting it to the vehicle. The vehicle monitors processor usage, task queue accumulation, and latency fluctuations in real time, extracting resource depletion information. Through the interactive mapping between the redundancy expansion feature and resource depletion information, a navigation decision latency degradation model is constructed, generating a quantifiable latency degradation index. Based on the latency degradation index, the path selection depth and priority are dynamically adjusted to effectively suppress redundant solutions while ensuring the priority processing of high-confidence critical paths, thereby outputting executable paths that balance multi-objective optimization and real-time performance.

[0157] This invention can significantly reduce the impact of solution space expansion on vehicle-side computing load, improve path selection and decision response speed, ensure the real-time performance and optimality of the navigation module in multi-target severe conflict scenarios, enhance vehicle-side resource utilization efficiency, realize intelligent, dynamic and traceable path selection, and comprehensively improve the performance and reliability of multi-target vehicle navigation systems.

[0158] Example 2: This example introduces a multi-target in-vehicle navigation system based on vehicle network traffic conditions and cloud map data, such as... Figure 2 As shown, it includes a global road condition status generation module, a multi-objective factor extraction and path space construction module, a multi-objective evolution reasoning module, a redundancy expansion feature analysis module, a vehicle-side resource status monitoring module, a delay degradation modeling and index generation module, and a dynamic filtering and priority adjustment module.

[0159] The global traffic situation generation module is used to integrate vehicle network traffic flow, road events and environmental perception information, and combine cloud maps and historical evolution data to generate a global traffic situation with time series.

[0160] The multi-objective factor extraction and path space construction module is used to analyze the global road condition situation in the spatiotemporal dimension, extract multi-objective factors, and generate a multi-dimensional path search space by combining road topology.

[0161] The multi-objective evolutionary reasoning module is used to perform evolutionary reasoning on the multi-dimensional path search space based on conflict weights, and generate Pareto front solution sets;

[0162] The redundancy expansion feature analysis module is used to analyze the redundancy and complexity of the Pareto front solution set, form redundancy expansion features, and transmit them to the vehicle-side computing engine.

[0163] The vehicle-side resource status monitoring module is used to track the vehicle-side processor usage, task queue backlog, and latency fluctuations in real time, and to extract resource depletion information.

[0164] The delay degradation modeling and index generation module is used to interactively map the redundancy expansion characteristics with resource depletion information, construct the vehicle navigation decision delay degradation model, and generate the vehicle navigation decision delay degradation index.

[0165] The dynamic filtering and priority adjustment module is used to dynamically adjust the filtering depth and priority based on the vehicle-side navigation decision delay degradation index, suppress the computational burden caused by redundant solutions, and output an executable path.

[0166] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0167] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0168] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0169] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the system and method described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0170] In the several embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways.

[0171] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A multi-target in-vehicle navigation method based on vehicle network traffic conditions and cloud map data, characterized in that: Includes the following steps: By integrating vehicle-to-everything (V2X) traffic flow, road events, and environmental perception information, and combining cloud maps with historical evolution data, a global road condition status with time series is generated. Analyze the global road conditions in the spatiotemporal dimension, extract multi-objective factors, and generate a multi-dimensional path search space by combining road topology; Evolutionary reasoning is performed on the multidimensional path search space based on conflict weights to generate a Pareto front solution set; The redundancy and complexity of the Pareto front solution set are analyzed to form a redundancy expansion feature, which is then transmitted to the vehicle-side computing engine. Real-time tracking of vehicle-side processor usage, task queue backlog, and latency fluctuations; extraction of resource depletion information. By interactively mapping the redundancy expansion characteristics with resource depletion information, a vehicle-side navigation decision delay degradation model is constructed, generating a vehicle-side navigation decision delay degradation index. The screening depth and priority are dynamically adjusted based on the vehicle-side navigation decision delay degradation index to suppress the computational burden caused by redundant solutions and output an executable path.

2. The multi-target vehicle navigation method based on vehicle network traffic conditions and cloud map data according to claim 1, characterized in that: By integrating vehicle-to-everything (V2X) traffic flow, road events, and environmental perception information, and combining cloud-based maps with historical evolution data, a global road condition status with time series is generated, as follows: Collect vehicle network traffic flow, road events, roadside and vehicle-mounted environmental perception, as well as cloud maps and historical evolution records, and merge them into a raw multi-source time-series stream with source identification; The original multi-source time-series stream is aligned with a time base and compensated for delay to obtain a time-series corrected stream with a unified timestamp. Spatial semantic mapping is performed on the time series correction stream, and the location information is connected with the cloud map topology to map it into a semantic spatiotemporal segment stream based on road segment / intersection / grid; Spatiotemporal interpolation and short-term reconstruction are performed on missing or anomalous segments of the weighted spatiotemporal fragment stream. Continuity is restored by using adjacent segment backfilling, historical template matching and pattern regression to obtain a coherent time series matrix. The coherent temporal matrix and road event information are fused in parallel within a sliding window to form a labeled temporal situation with event intensity and impact range annotations; The labeled time series situation is subjected to spatiotemporal compression and index extraction at multiple scales. The key sequences of traffic rate, traffic density and event delay are calculated, and an indexable multi-scale time series index set is output. The multi-scale time series index set applies short-term forecasting and uncertainty assessment strategies to generate forecast time series and confidence intervals within future windows; The predicted time series and confidence interval are associated with and versioned with the original map topology to form a global traffic situation package containing timestamps, spatial indexes, confidence levels, event labels, and impact ranges.

3. The multi-target vehicle navigation method based on vehicle network traffic conditions and cloud map data according to claim 2, characterized in that: Within the spatiotemporal dimension, the global road condition situation is analyzed, multi-objective factors are extracted, and a multi-dimensional path search space is generated by combining road topology, as detailed below: Using the global road condition situation package as input, the situation package is spatiotemporally sliced ​​according to a preset time scale and spatial granularity to generate a set of labeled road segment time units, and an event label and original data confidence field are added to each unit; Spectral decomposition and abrupt change detection are performed on the time unit set of road segments to extract the original factors of traffic rate, congestion wave propagation rate, event delay, traffic variance and confidence standard deviation, and output a vector set that constitutes multi-objective factors; Normalize and weight the multi-objective factor vector set to obtain the objective factor matrix; A temporally weighted road network tensor is constructed by mapping the target factor matrix to road topology elements one-to-one and aligning them with time. The temporally weighted road network tensor is subjected to segment combination and attribute accumulation on the time-expanded network according to connectivity rules, generating an ordered sequence of path attribute vectors; By performing diversity-preserving clustering and target dominance analysis on the path attribute vector sequence in the vector space, a candidate multidimensional path search space representation is obtained.

4. The multi-target vehicle navigation method based on vehicle network traffic conditions and cloud map data according to claim 1, characterized in that: The redundancy and complexity of the Pareto front solution set are analyzed to form a redundancy expansion feature, which is then transmitted to the vehicle-side computing engine, as follows: Receive the Pareto front solution set and the target vector and confidence information of each path, and construct a high-dimensional vector space representation of the solution set according to the target dimension; Density estimation is performed on the high-dimensional vector space, and the local clustering density of each solution in the solution set is calculated by the nearest neighbor distance method to generate the spatial distribution feature vector of the solution. Based on the mutual information between target vectors, the target conflict degree of each solution is calculated, and a conflict feature description of the solution is generated. The information entropy of the solution set is calculated on the target dimension, and the information entropy value is used as an indicator of the structural compactness of the solution set; The candidate set size, cluster density, target conflict degree, and information entropy are integrated into a multidimensional redundant expansion feature vector. The multidimensional redundancy expansion features are transmitted to the vehicle-side computing engine through efficient differential compression and a publish-subscribe mechanism.

5. The multi-target vehicle navigation method based on vehicle network traffic conditions and cloud map data according to claim 4, characterized in that: Real-time tracking of vehicle-side processor usage, task queue backlog, and latency fluctuations; extraction of resource exhaustion information, as detailed below: A lightweight acquisition thread is established on the vehicle side to periodically acquire processor utilization, thread utilization, and core frequency, along with operating temperature and power consumption status, forming an initial computing power monitoring stream. Continuous monitoring is performed on the vehicle-side task queue to record the length of tasks to be processed, the priority of task types, the distribution of processing time, and changes in queue backlog, generating a task load time stream. The collected computing power monitoring stream and task load time series stream are synchronized and aligned with the time base to form a resource status matrix under a unified timestamp; Extract resource depletion information from the resource status matrix, including transient fluctuations in processor utilization, rate of change in queue length, bursts in task latency, and combined stress values ​​of temperature and power consumption.

6. The multi-target vehicle navigation method based on vehicle network traffic conditions and cloud map data according to claim 5, characterized in that: The process involves interactively mapping redundancy expansion characteristics with resource depletion information to construct a vehicle-side navigation decision delay degradation model and generate a vehicle-side navigation decision delay degradation index. The formula underlying the vehicle-side navigation decision delay degradation model is as follows: ,in The vehicle-side navigation decision delay degradation index, For the redundancy expansion characteristics and resource depletion information, the first Dimensionless values ​​of each indicator. For the first The weighting coefficients of each indicator, and Greater than 0, .

7. A multi-target in-vehicle navigation system based on vehicle network traffic conditions and cloud map data, used to implement the multi-target in-vehicle navigation method based on vehicle network traffic conditions and cloud map data as described in any one of claims 1-6, characterized in that: It includes a global road condition status generation module, a multi-objective factor extraction and path space construction module, a multi-objective evolution reasoning module, a redundancy expansion feature analysis module, a vehicle-side resource status monitoring module, a delay degradation modeling and index generation module, and a dynamic filtering and priority adjustment module; The global traffic situation generation module is used to integrate vehicle network traffic flow, road events and environmental perception information, and combine cloud maps and historical evolution data to generate a global traffic situation with time series. The multi-objective factor extraction and path space construction module is used to analyze the global road condition situation in the spatiotemporal dimension, extract multi-objective factors, and generate a multi-dimensional path search space by combining road topology. The multi-objective evolutionary reasoning module is used to perform evolutionary reasoning on the multi-dimensional path search space based on conflict weights, and generate Pareto front solution sets; The redundancy expansion feature analysis module is used to analyze the redundancy and complexity of the Pareto front solution set, form redundancy expansion features, and transmit them to the vehicle-side computing engine. The vehicle-side resource status monitoring module is used to track the vehicle-side processor usage, task queue backlog, and latency fluctuations in real time, and to extract resource depletion information. The delay degradation modeling and index generation module is used to interactively map the redundancy expansion characteristics with resource depletion information, construct the vehicle navigation decision delay degradation model, and generate the vehicle navigation decision delay degradation index. The dynamic filtering and priority adjustment module is used to dynamically adjust the filtering depth and priority based on the vehicle-side navigation decision delay degradation index, suppress the computational burden caused by redundant solutions, and output an executable path.

Images