Unmanned vehicle operation log completeness detection method based on standard process
By constructing a standard process model and aligning it with real logs, the problem of incomplete autonomous vehicle operation logs was solved, enabling efficient log quality assessment and system optimization.
Patent Information
- Application Number
- CN202610024151.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-01-09
- Publication Date
- 2026-03-31
AI Technical Summary
During system upgrades and module iterations, autonomous vehicle operation logs often exhibit issues such as design inconsistencies, lack of recording of key activities, and missing state transitions. Existing methods struggle to quantify the logs' coverage of standard processes, leading to log problems being passively discovered after system failures, which impacts system maintainability and analysis reliability.
By generating simulated operation logs, a standard process model is constructed, and process mining algorithms and Petri net alignment algorithms are used to detect the completeness of the real operation logs, quantify the degree of process coverage, identify missing links, and provide multi-level completeness evaluation indicators.
It enables the discovery of missing log issues from a holistic process perspective, improves the accuracy of log quality identification, reduces maintenance costs, and provides a reliable data foundation for system optimization and anomaly detection.
Smart Images
Figure FT_1 
Figure FT_2 
Figure FT_3
Abstract
Description
Technical Field
[0001] This invention belongs to the field of data processing technology for intelligent transportation systems, and specifically relates to a method for detecting the completeness of unmanned vehicle operation logs based on standard processes, used to detect the completeness of unmanned vehicle operation logs in terms of process coverage. Background Technology
[0002] In applications such as warehousing and logistics, and smart manufacturing, autonomous vehicles generate a large amount of operational logs, recording operations such as order execution, route planning, state transitions, and task scheduling. These logs form the data foundation for operational analysis, anomaly troubleshooting, and system optimization. Autonomous vehicle operational logs are generated separately by multiple software modules, resulting in dispersed sources and complex structures. During system upgrades and module iterations, issues such as inconsistent log design, failure to record key activities, and missing state transitions frequently arise. These defects are difficult to detect in a single log entry but directly impact the results of subsequent data analysis and process mining.
[0003] Currently, log quality inspection methods mainly fall into three categories: format validation, rule-based detection, and process mining. Format validation methods use regular expressions or predefined rules to verify log format, field types, and required fields. However, they only focus on the syntactic correctness of individual log entries and cannot determine whether the overall log covers the complete business process. Furthermore, they lack the ability to detect the semantic rationality of field content. Rule-based anomaly detection methods identify outliers, violations, or timing conflicts by manually defining rule sets. Rule maintenance is costly, and they struggle to cover various missing scenarios in complex processes. When faced with changes in log format or new business processes, rules need to be rewritten, resulting in poor adaptability. Some research attempts to introduce expert systems or decision trees for rule management, but in scenarios involving multi-module collaboration and multi-path concurrency, the number of rules increases dramatically, significantly increasing maintenance difficulty. Process mining methods, such as the Alpha algorithm, Inductive Miner, and Split Miner, extract process models from event logs and perform consistency checks or process consistency detection. They are primarily used to discover deviations between the actual executed process and the expected process. These methods assume that the input logs are complete and reliable. However, when using real runtime logs as the sole input, it is difficult to detect the completeness of the process within the logs themselves, and it is impossible to determine whether the real logs have omitted key activities or paths in the standard process. When the input logs are incomplete, the mined process model may lack key branches or contain pseudo-paths, causing subsequent analysis to fail.
[0004] Compared to traditional information systems, the logs of autonomous vehicle systems have stronger temporal correlation and process dependency. A single log entry often cannot independently reflect the complete business behavior, requiring semantic reconstruction by combining logs from preceding and following entries. Furthermore, autonomous vehicle logs involve the collaboration of multiple modules (such as the scheduling layer, execution layer, and perception layer), with diverse log activity types and frequent state transitions. Standard processes may include dozens of activity nodes and multiple execution paths, making manual inspection inefficient and prone to omissions.
[0005] During the actual development and testing phases, developers lack systematic methods for assessing log completeness. Existing methods cannot quantify the extent to which logs cover standard processes, making it difficult to pinpoint specific missing activities, paths, or state transitions. This results in log issues being passively discovered only after system failures or analysis failures, preventing timely correction during the development phase and impacting the maintainability of the log system and the reliability of subsequent analysis tasks.
[0006] To address the aforementioned issues, this invention proposes a method for detecting the completeness of unmanned vehicle operation logs based on standard processes. Through systematic comparative analysis, the method quantitatively evaluates the process coverage of the logs, identifies missing links, and provides data support for log design optimization. Summary of the Invention
[0007] The purpose of this invention is to provide a method for detecting the completeness of unmanned vehicle operation logs based on standard processes, which solves the problems of incomplete process recording, missing key activities, and lack of quantitative analysis methods for log quality from an overall process perspective in the existing technology.
[0008] The method for detecting the completeness of unmanned vehicle operation logs based on standard procedures according to the present invention includes the construction of a standard procedure based on simulated logs and the detection of the procedure completeness based on real operation logs. The specific steps are as follows:
[0009] Step 1: Based on the preset standard business process of unmanned vehicles, write a script to simulate the operation of unmanned vehicles and generate a simulation operation log; the simulation operation log includes information such as timestamp, order number, vehicle number, operation events and status transitions, organized by order, and its data format is consistent with the real operation log;
[0010] Step 2: Perform structured parsing on the simulated operation log generated in Step 1, extract key field information such as timestamp, order number, log template number and template text, and organize the parsed log in structured data form, maintaining the original log order and order association;
[0011] Step 3: Based on the structured log data obtained in Step 2, perform activity-level abstraction on the log templates, cluster log templates with similar semantic content and business meaning, map log templates of the same category to the same business activity, and assign an activity identifier to each category of activities; use the order number as the process instance identifier, determine the final execution result of the order based on the execution status information in the last log record of the order, and construct event log data containing the order execution result marker;
[0012] Step 4: Based on the event log data obtained in Step 3, a standard process model for unmanned vehicle order execution is constructed using a process mining algorithm, and process structure elements are extracted, including the standard process model structure, the standard activity set, and the standard path set.
[0013] Step 5: Convert the actual operation logs into event log format according to the processing methods in Steps 2 and 3. Use a Petri net-based alignment algorithm to perform process alignment analysis between the actual event logs and the standard process model to identify synchronization steps and model steps. Calculate process completeness indicators based on the alignment results, including trajectory-level process completeness, log-level process completeness, proportion of fully aligned trajectories, activity-level process completeness, and path-level process completeness. Based on the process completeness indicators, output information on missing process elements, including the set of standard activities missing in each order execution trajectory and the set of activity paths that exist in the standard process but do not appear in the actual logs.
[0014] Compared with the prior art, the present invention has the following advantages:
[0015] 1. By performing a completeness check on the operation log from the perspective of the overall process, compared with existing methods that only focus on a single log or local rules, it can discover log missing issues at the process level and improve the accuracy of identifying log quality problems.
[0016] 2. By using quantitative indicators to pinpoint the specific missing activities and execution paths, developers can improve log recording in a targeted manner, avoiding the inefficient method of manually checking logs one by one and reducing the maintenance cost of the log system.
[0017] 3. Provides multi-level completeness evaluation metrics, including trajectory level, activity level and path level, which can comprehensively evaluate log quality from different granularities and provide a reliable data foundation for system optimization and anomaly detection. Attached Figure Description
[0018] Figure 1 A schematic diagram of the overall process for a standard procedure-based method for detecting the completeness of autonomous vehicle operation logs. Figure 2 A schematic diagram of a standard operating procedure model for driverless vehicles built based on simulated logs. Figure 3 A schematic diagram illustrating the process alignment analysis between real operation logs and standard process models. Detailed Implementation
[0019] The technical solution of the present invention will be further described below with reference to the accompanying drawings.
[0020] This invention relates to a method for detecting the completeness of unmanned vehicle operation logs based on a standard process, the overall process of which is as follows: Figure 1 As shown. The method constructs a standard operating procedure model for autonomous vehicles based on simulated operation logs, and performs a procedure completeness check based on the actual operation logs and the standard operating procedure model. The specific implementation steps are as follows:
[0021] Step 1: Based on the preset standard business process for autonomous vehicles, write scripts to simulate the operation of the autonomous vehicle and generate simulation operation logs. The simulation logs include timestamps, order numbers, vehicle numbers, operation events, and state transitions, organized by order, and their data format is consistent with real operation logs. The simulation logs are generated by encoding standard business rules and can accurately reflect the sequence of activities and state changes under the normal operation process of the autonomous vehicle.
[0022] Step 2: Perform structured parsing on the simulated operation logs generated in Step 1. The parsing process can utilize existing log parsing methods to extract key field information from each log record, including the timestamp, order number, log template number, and template text. The parsed logs are organized in a structured data format, preserving the original log order and order relationships. Through this step, the simulated logs are converted into structured data, facilitating subsequent activity clustering and process model construction.
[0023] Step 3: Based on the structured log data obtained in Step 2, perform activity-level abstraction on the log template and construct event log data with order execution result tags to realize the transformation of log data into process mining event logs, specifically including:
[0024] 1) Based on the template information in the structured logs, log templates with similar semantic content and business meaning are clustered, and log templates of the same category are regarded as the same type of business activity, and a unique activity identifier is assigned to each type of activity;
[0025] 2) For each type of activity, combine the text features and business semantics of the log templates contained in the activity to generate the corresponding activity name, and complete the semantic mapping from log templates to process activities;
[0026] 3) Using the order number as the process instance identifier, analyze the log records of the same order, and determine the final execution result of the order based on the execution status information in the last log record of the order;
[0027] 4) Mark the final execution result of the order in the last log record corresponding to the order, indicating whether the order has completed the standard business process or has been abnormally terminated.
[0028] Through the above processing, the structured log data is transformed into event log data with orders as process instances, activities as execution nodes, and order execution result markers.
[0029] Step 4: Based on the event log data obtained in Step 3, a standard process model for unmanned vehicle order execution is constructed using a process mining algorithm. The overall process construction process is as follows: Figure 2 As shown, it specifically includes:
[0030] 1) Convert event log data into a format recognizable by process mining algorithms, using the order number as the process instance identifier (case_id) and the activity name as the event activity attribute. Employ process mining algorithms to discover the process structure of the order execution process, resulting in a process model that reflects the order execution logic under normal autonomous vehicle business scenarios.
[0031] 2) Based on the constructed standard process model, extract process structure elements, including the standard process model structure (such as Petri net structure, initial state and termination state), standard activity set (describing all standardized activities in the order execution process), and standard path set (describing the legal flow paths between activities), as input basis for subsequent process alignment analysis and process completeness detection.
[0032] Step 5: Based on the standard process model constructed in Step 4, convert the actual operation logs into event log format according to the processing methods in Steps 2 and 3, and perform process alignment analysis with the standard process model. The overall alignment and analysis process is as follows: Figure 3 As shown, it specifically includes:
[0033] 1) A Petri net-based process alignment algorithm is employed. By constructing an optimal alignment sequence between the log layer and the model layer, the matching relationship between event behaviors recorded in the actual operation log and activities in the standard process model is identified. During the alignment process, for each order execution trajectory... The corresponding alignment step sequence is obtained, and the alignment steps are divided into the following two categories:
[0034] a) Synchronization step ( ): This represents an activity that should occur in the standard process model and has a corresponding record in the actual operation log;
[0035] b) Model step ( ): This indicates an activity that should occur in the standard process model but is not recorded in the actual runtime log and is only executed on the model side.
[0036] 2) Calculate process completeness metrics based on alignment results to quantify the process completeness of real-world runtime logs from multiple perspectives, specifically including:
[0037] a) Trajectory-level process completeness: This refers to the execution trajectory of a single order. Define its process completeness The calculation formula is as follows:
[0038]
[0039] in, Representing the trajectory The set of synchronization steps generated during the alignment process. Representing the trajectory The set of model steps generated during the alignment process.
[0040] b) Log-level process completeness: Based on the alignment results of all order execution trajectories in the actual operation log with the standard process model, the sufficiency of the actual operation log in recording the standard process model is measured at the overall level. The calculation formula is as follows:
[0041]
[0042] in, This represents the complete set of order execution trajectories contained in the actual operation log.
[0043] c) Fully Aligned Track Ratio: When the order executes the track... If no model step is generated during the alignment process, the trajectory is considered a complete process trajectory, and the proportion of fully aligned trajectories is defined. for:
[0044]
[0045] in, This represents the set of order trajectories for which no model steps were generated. This represents the set of all order traces in the actual operation log.
[0046] d) Activity-level process completeness: For any activity in the standard process model Define its activity-level completeness index The calculation formula is as follows:
[0047]
[0048] in, Indicates activity The total number of times it appears in the synchronization step. Indicates activity The total number of times it appears in the model step.
[0049] e) Path-level process completeness: Based on the standard process model and actual operation logs, direct follow-up diagrams are constructed respectively. The activity flow paths are compared and analyzed, and the path-level process completeness index is defined as follows:
[0050]
[0051] in, This represents the set of activity-direct-following paths obtained from the standard process model. This represents the set of activity-direct-following paths obtained from actual runtime logs.
[0052] 3) Based on the process alignment results, output the missing process element information in the actual operation log, including the set of standard activities missing in each order execution trajectory, and the set of activity paths that exist in the standard process model but do not appear in the actual log, thereby clearly indicating the specific missing locations of the actual operation log at the process node level and process path level.
Claims
1. A standard procedure-based unmanned vehicle operation log completeness detection method, characterized in that, Comprise the following steps: Step 1, based on the preset unmanned vehicle standard business process, write scripts to simulate unmanned vehicle operation, generate simulation running log; The simulation running log contains timestamp, order number, vehicle number, operation event and state switching information, organizes it in order unit, its data format is consistent with the real running log; Step 2, the simulation running log generated in step 1 is structured and parsed, and the timestamp, order number, log template number and template text and other key field information are extracted, the parsed log is organized in structured data form, the original log order and order correlation are maintained; Step 3, based on the structured log data obtained in step 2, the activity level abstraction of log template is carried out, the log templates with similar semantic content and business meaning are clustered, the log templates of the same category are mapped to the same business activity, and the activity identifier is assigned to each category of activity; Take the order number as the process instance identifier, determine the final execution result of the order based on the execution state information in the last log record of the order, and construct the event log data containing the order execution result mark; Step 4, based on the event log data obtained in step 3, the standard process model of unmanned vehicle order execution is constructed by using process mining algorithm, and the process structure elements are extracted, including standard process model structure, standard activity set and standard path set; Step 5, the real running log is converted into event log form according to the processing mode of step 2 and step 3, the process alignment algorithm based on Petri net is used to align the real event log with the standard process model, and the synchronous step and model step are identified; Based on the alignment result, the process completeness index is calculated, including trajectory level process completeness, log level process completeness, complete alignment trajectory proportion, activity level process completeness and path level process completeness, and based on the process completeness index, the missing process element information is output, including the missing standard activity set in each order execution trajectory and the activity path set existing in the standard process but not appearing in the real log.
2. The processing method for activity level abstraction of log templates according to claim 1, wherein, The log templates with similar semantic content and business meaning are classified into the same category, and the log templates of the same category are mapped to the same business activity; For each category of business activity, the corresponding activity name is generated combined with the text features and business semantics of the log template, and the semantic mapping of the log template to the process activity is completed.
3. The process of constructing a standard flow model according to claim 1, wherein, The standard process model adopts Petri net structure representation, including initial state, termination state and control flow relationship between activities; The standard activity set is used to describe all the standard activities in the order execution process, and the standard path set is used to describe the legal flow path between activities.
4. The process alignment analysis method of claim 1, wherein, The real event log is aligned with the standard process model by using the alignment algorithm based on Petri net, the alignment sequence between log layer and model layer is constructed, and the synchronous step and model step are identified; Among them, the synchronous step represents the activity that should occur in the standard process model and has been recorded in the real log, and the model step represents the activity that should occur in the standard process model but has not been recorded in the real log and can only be executed in the model side.
5. The process completeness indicator calculation method of claim 1, wherein, The process completeness indicators include: trajectory-level process completeness, which is used to measure the degree of standard activities in a single order execution trajectory being recorded by the real log; log-level process completeness, which is used to measure the overall recording sufficiency of the standard process model by the real running log at the level of all order execution instances; the proportion of completely aligned trajectories, which is used to measure the proportion of order instances with complete process records; activity-level process completeness, which is used to identify key activity nodes in the process that are prone to log loss; path-level process completeness, which is used to measure the coverage of the standard process control flow structure in the real log.