System and method for automatically generating recommendation script based on short-term self-driving travel

Abstract

The invention discloses a system and method for automatically generating a recommendation script based on a short-term self-driving travel journey, and relates to the technical field of intelligent traffic and vehicle-mounted information services, and the system comprises the following modules: a journey planning module which is used for obtaining self-driving travel journey data which is set by a user and comprises at least one journey point and a corresponding planning time window; the data integration engine module is coupled to the travel planning module and used for collecting vehicle state data and environment real-time data in real time and integrating the travel data, the vehicle state data and the environment real-time data to generate a unified structured data packet; and the large model decision center module is coupled to the data integration engine module and is used for receiving the structured data packet and generating a recommendation script comprising at least two ordered actions through situational reasoning, and the types of the actions comprise at least two of navigation, vehicle control and service reservation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent transportation and vehicle information service, and particularly relates to a system and method for automatically generating a recommended script based on a short-term self-driving trip. BACKGROUND

[0002] With the rapid development of intelligent transportation and vehicle information service, a self-driving trip planning system gradually integrates navigation, vehicle control and service booking functions. However, the existing scheme still relies on multiple independent subsystems to process trip data, vehicle state and environmental information, resulting in insufficient functional synergy. This structural problem limits the system's ability to provide coherent services in a dynamic environment, causing users to manually switch between multiple applications and heavy decision-making burdens. SUMMARY

[0003] The purpose of the present application is to provide a system and method for automatically generating a recommended script based on a short-term self-driving trip, which at least solves the problem of users manually switching between multiple applications and heavy decision-making burdens.

[0004] The present application provides the following scheme:

[0005] According to one aspect of the present application, a system for automatically generating a recommended script based on a short-term self-driving trip is provided, comprising the following modules:

[0006] A trip planning module for obtaining a self-driving trip data set by a user, including at least one trip point and a corresponding planned time window;

[0007] A data integration engine module coupled to the trip planning module for real-time collection of vehicle state data and environmental real-time data, and integration of the trip data, vehicle state data and environmental real-time data to generate a unified structured data package;

[0008] A large model decision center module coupled to the data integration engine module for receiving the structured data package and generating a recommended script containing at least two ordered actions through situational reasoning, wherein the types of actions include at least two of navigation, vehicle control and service booking;

[0009] A script execution engine module coupled to the large model decision center module for parsing the recommended script and invoking corresponding downstream systems to perform operations according to the parsed action types, the downstream systems including at least two of a navigation system, a vehicle control system and a third-party service API;

[0010] A system control module coupled to each module for dynamically coordinating the execution timing of each module based on the planned time window in the trip data, forming a closed-loop processing flow from trip planning to automatic script execution.

[0011] Further, the trip planning includes:

[0012] Receiving a plurality of trip points input by a user through a graphical interface, each trip point containing latitude and longitude coordinates or a POI name;

[0013] Configuring a planned arrival time and a planned departure time for each trip point to form a planned time window;

[0014] Organizing the trip points and corresponding planned time windows into an ordered trip sequence to generate machine-readable trip data.

[0015] Further, the data integration engine includes:

[0016] Collecting vehicle state data through the vehicle CAN bus interface, including battery remaining capacity, current latitude and longitude coordinates of the vehicle, and in-vehicle temperature sensor data;

[0017] Calling traffic data service API to obtain real-time traffic congestion index and calling meteorological data service API to obtain weather condition information through wireless network;

[0018] Packaging the trip data, vehicle state data and environmental real-time data according to the predefined JSON format.

[0019] Further, the data integration engine further includes:

[0020] Standardizing the time information in the trip data to uniform 24-hour time format;

[0021] Converting the power data in the vehicle state data to percentage and unifying the position data to coordinate system;

[0022] Normalizing the traffic congestion index in the environmental real-time data and encoding the weather information;

[0023] Packing the processed data according to the predetermined structure to generate complete structured data packets containing timestamps.

[0024] Further, the large model decision center includes:

[0025] Combining the structured data packets with pre-defined prompt word templates to form large model input data;

[0026] Analyzing the correlation between trip progress, vehicle state and environmental conditions through multi-round reasoning of the input data by the large language model;

[0027] Generating an ordered action sequence containing navigation instructions, vehicle control instructions and service reservation instructions based on the reasoning results.

[0028] Further, the large model decision center further comprises:

[0029] Generating navigation actions, including destination coordinates, waypoints, and path preference parameters;

[0030] Generating vehicle control actions, including device control instructions, target parameters, and execution timing;

[0031] Generating service reservation actions, including service type, service provider identification, and reservation time parameters;

[0032] Arranging the actions in chronological order and adding dependencies between actions to form a complete recommended script.

[0033] Further, the script execution engine comprises:

[0034] Parsing the JSON structure of the recommended script and extracting the action chain array;

[0035] Identifying the type field and parameter field of each action;

[0036] Distributing instructions according to action type to corresponding system adapters;

[0037] Converting general instructions to downstream system-specific instruction formats through system adapters.

[0038] Further, the script execution engine further comprises:

[0039] Calling the navigation system API and passing in the destination coordinates and path planning parameters;

[0040] Calling the vehicle control system API and sending device control instructions and execution time parameters;

[0041] Calling the third-party service OpenAPI and submitting service reservation requests and user authentication information;

[0042] Monitoring the execution status of each system and displaying the execution progress on the graphical interface.

[0043] Further, the system control comprises:

[0044] Real-time monitoring of the deviation between system time and trip planning time window;

[0045] Adjusting the data collection frequency of the data integration engine dynamically according to the trip progress;

[0046] Based on the planned departure time and real-time traffic estimation, the best decision-making opportunity is triggered to trigger large model reasoning;

[0047] Coordinating the execution order and time interval of each action in the script execution engine;

[0048] The feedback mechanism ensures that the modules work together to form a complete closed-loop control system.

[0049] According to two aspects of the present application, a method for automatically generating a recommended script based on a short-term self-driving trip is provided, comprising the following steps:

[0050] S1, obtaining self-driving trip data set by a user, containing at least one trip point and a corresponding planned time window;

[0051] S2, coupled to the trip planning module, for real-time collection of vehicle state data and real-time environmental data, and integration of the trip data, vehicle state data and real-time environmental data to generate a unified structured data package;

[0052] S3, coupled to the data integration engine module, for receiving the structured data package and generating a recommended script containing at least two ordered actions through context reasoning, wherein the types of actions include at least two of navigation, vehicle control and service booking;

[0053] S4, coupled to the large model decision center module, for parsing the recommended script and calling corresponding downstream systems to perform operations according to the parsed action types, the downstream systems including at least two of a navigation system, a vehicle control system and a third-party service API;

[0054] S5, coupled to each module respectively, for dynamically coordinating the execution timing of each module based on the planned time window in the trip data, forming a closed-loop processing flow from trip planning to automatic script execution.

[0055] Through the above scheme, the following beneficial technical effects are obtained:

[0056] The present application realizes multi-source data fusion and automatic script generation by constructing a complete intelligent decision-making architecture for self-driving tours, and further integrates traditional scattered trip management, vehicle control and environmental perception functions into a unified processing flow, thereby improving the problem that in traditional schemes, isolated systems are used for separate processing, and due to the lack of a coordination mechanism, the user needs to manually switch between multiple applications and the decision-making burden is heavy.

[0057] The present application collects and fuses trip data, vehicle state data and real-time environmental data through a data integration engine to generate a unified structured data package, and further provides a complete and accurate data basis for subsequent decision-making, thereby improving the problem that in traditional schemes, a single data source is mostly used for processing, and due to incomplete data dimensions, the decision-making basis is insufficient and the recommendation accuracy is insufficient.

[0058] The application performs context reasoning based on a structured data package through a large model decision center and generates a recommended script containing multiple types of actions, thereby realizing intelligent trip decision-making, thereby improving the problem that traditional schemes mostly use rule engines for single function recommendation, and due to the lack of context understanding ability, the recommended content is out of touch with the actual situation. BRIEF DESCRIPTION OF DRAWINGS

[0059] Figure 1 is a module architecture diagram of a system for automatically generating a recommended script based on a short-term self-driving trip according to an embodiment of the present application.

[0060] Figure 2 is a data flow diagram of a system for automatically generating a recommended script based on a short-term self-driving trip according to an embodiment of the present application.

[0061] Figure 3 is a method flow diagram of a method for automatically generating a recommended script based on a short-term self-driving trip according to an embodiment of the present application. DETAILED DESCRIPTION

[0062] The technical solutions of the present application will be described clearly and completely below in conjunction with the drawings. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of protection of the present application.

[0063] Embodiment one:

[0064] Figure 1 is a module architecture diagram of a system for automatically generating a recommended script based on a short-term self-driving trip according to an embodiment of the present application.

[0065] Figure 2 is a data flow diagram of a system for automatically generating a recommended script based on a short-term self-driving trip according to an embodiment of the present application.

[0066] A system for automatically generating a recommended script based on a short-term self-driving trip, as shown in Figures 1-2 includes the following modules:

[0067] A trip planning module for obtaining a self-driving trip data set by a user, including at least one trip point and a corresponding planned time window;

[0068] Further, the trip planning includes:

[0069] Receiving multiple trip points input by a user through a graphical interface, each trip point containing latitude and longitude coordinates or a POI name;

[0070] Configure a planned arrival time and a planned departure time for each trip point to form a planned time window;

[0071] Organize the trip points and corresponding planned time windows into an ordered trip sequence to generate machine-readable trip data.

[0072] Specifically, in the trip planning module, the user inputs the geographic location information of multiple trip points through a graphical interface, including latitude and longitude coordinates or specific POI names. The system configures a planned arrival time and a planned departure time for each trip point to form a complete planned time window. These data are organized into an ordered trip sequence to generate machine-readable trip data.

[0073] The structure of trip data includes trip point sequence, time window information, and metadata. The trip point sequence ensures the order relationship of the points, the time window information defines the planned time range of each point, and the metadata records the trip creation time and the last modification time. This data structure provides a standardized input format for subsequent processing.

[0074] In the data processing process, the system uses a path time estimation algorithm to calculate the trip time between points. The path time calculation formula is: ; where represents the estimated trip time from point to point , represents the distance between the two points, represents the average speed of the road segment, represents the real-time traffic influence coefficient.

[0075] The system verifies the reasonableness of the trip through a time window conflict detection algorithm. The conflict detection formula is:

[0076] ;

[0077] where represents the planned arrival time of the th point, represents the planned departure time of the th point, represents the estimated trip time between the two points. When Conflict is greater than zero, the system detects a time conflict.

[0078] The data output process includes three steps: first, receive the user's input of the original trip information, then perform data standardization processing, and finally output structured trip data. The input data comes from user interface interaction, the processing process includes data verification and time reasonableness check, and the output data is passed to the downstream data integration engine module.

[0079] The technical effect of the module is to convert user input into precise and calculable trip planning data, providing reliable input basis for subsequent intelligent decision-making. Through standardized data structure and verification mechanism, the integrity and consistency of the trip data are ensured.

[0080] The data integration engine module is coupled to the trip planning module for real-time collection of vehicle state data and real-time environmental data, and integration of trip data, vehicle state data and real-time environmental data to generate a unified structured data package;

[0081] Further, the data integration engine includes:

[0082] The vehicle state data is collected through the vehicle CAN bus interface, including the remaining battery capacity, the current latitude and longitude coordinates of the vehicle and the in-vehicle temperature sensor data;

[0083] The real-time traffic congestion index is obtained by calling the traffic data service API through the wireless network, and the weather condition information is obtained by calling the meteorological data service API;

[0084] The trip data, vehicle state data and real-time environmental data are packaged according to the predefined JSON format.

[0085] The data integration engine further includes:

[0086] The time information in the trip data is standardized and unified into 24-hour time format;

[0087] The battery data in the vehicle state data is converted into percentage, and the position data is unified in coordinate system;

[0088] The traffic congestion index in the real-time environmental data is normalized, and the weather information is encoded and converted;

[0089] The processed data is packaged according to the predetermined structure to generate a complete structured data package containing a timestamp.

[0090] Specifically, the data integration engine module collects heterogeneous data through multiple interface protocols. The vehicle CAN bus interface reads the data stream of the battery management system and the vehicle-mounted sensor at a specific frequency to obtain the remaining battery capacity, the latitude and longitude coordinates of the vehicle and the in-vehicle temperature value. The wireless network module establishes a secure connection with the traffic data platform and the meteorological service server, and obtains the real-time traffic congestion index and weather condition information through the standard API interface.

[0091] The data standardization processing adopts unified rules. The time information is converted into 24-hour ISO8601 format to ensure the consistency of the time data. The battery capacity data is converted into standard percentage according to the formula:

[0092] ;

[0093] wherein denotes the normalized percentage of charge, denotes the current charge reading, denotes the minimum range of the charge meter, denotes the maximum range of the charge meter. The position coordinates are converted to longitude and latitude values in the WGS-84 coordinate system.

[0094] The traffic congestion index is processed using a linear normalization method: ; wherein denotes the normalized traffic index, denotes the original traffic index, denotes the lower limit of the traffic index range, denotes the upper limit of the traffic index range. Weather information is converted into a numerical code according to a predefined coding table, for example, a sunny day is coded as 1 and a rainy day is coded as 2.

[0095] The data packaging process is performed according to the predefined JSON schema. The generated structured data packet contains fixed fields: the timestamp field records the data collection time, the trip data field stores the standardized trip information, the vehicle state field contains the processed vehicle parameters, and the environmental data field carries the normalized traffic and weather data. The timestamp uses Unix timestamp format to ensure the accuracy of the time marker.

[0096] This module establishes a data quality verification mechanism. Range checks are performed on the collected raw data to exclude abnormal values that exceed the reasonable range. The nearest valid value completion strategy is used for missing data to ensure the continuity of the data stream. Data packet integrity verification is achieved through the checksum algorithm to ensure that there is no packet loss or damage during data transmission.

[0097] The data output process maintains strict timing. The data integration engine generates new data packets at fixed time intervals, which can be configured according to system requirements. Each data packet carries a unique sequence number to track the generation order of the data packet. The output data is transmitted to the downstream processing module through the message queue to ensure the reliability and orderliness of data transmission.

[0098] The large model decision center module is coupled to the data integration engine module for receiving structured data packets and generating a recommended script containing at least two ordered actions through situational reasoning, wherein the types of actions include at least two of navigation, vehicle control, and service booking;

[0099] Further, the large model decision center includes:

[0100] Combining the structured data packet with the predefined prompt word template to form the large model input data;

[0101] By using a large language model to perform multi-round reasoning on the input data, the correlation between trip progress, vehicle status, and environmental conditions is analyzed.

[0102] Based on the reasoning results, an ordered sequence of actions is generated, including navigation instructions, vehicle control instructions, and service reservation instructions.

[0103] The large model decision center further includes:

[0104] Generate navigation actions, including destination coordinates, waypoints, and path preference parameters;

[0105] Generate vehicle control actions, including equipment control commands, target parameters, and execution timing;

[0106] Generate a service booking action, including service type, service provider identifier, and booking time parameters;

[0107] Arrange the actions in chronological order and add dependencies between the actions to form a complete recommendation script.

[0108] Specifically, the large model decision center module receives structured data packets from the data integration engine. This module combines the data packets with predefined prompt word templates to construct the input sequence for the large language model. The prompt word templates contain system role definitions, task descriptions, and output format specifications, ensuring that the large model reasons in a predetermined direction.

[0109] During the contextual reasoning process, the large model performs multiple rounds of analysis on the input data. The first round of analysis identifies the trip progress status by comparing the time difference between the current time and the planned time window to obtain a progress assessment. The second round of analysis examines the degree of matching between the vehicle status and the trip requirements, particularly the relationship between battery charge and subsequent trip energy consumption. The third round of analysis analyzes the impact of environmental conditions on trip execution, including the delay effect of traffic conditions on trip time and the need for vehicle control due to weather conditions.

[0110] The large model uses a decision weighting algorithm to calculate the priority score for each action: ;in Indicates action The overall priority score, Indicates the first The weighting coefficients of each influencing factor This indicates the corresponding data input. This represents the characteristic function of the influencing factor. The weighting coefficients are dynamically adjusted according to the action type: navigation actions emphasize time factors, vehicle control actions emphasize vehicle status factors, and service booking actions emphasize resource availability factors.

[0111] The navigation action generation process extracts destination coordinates, waypoints along the route, and route preference parameters. Destination coordinates are directly taken from the next trip point location in the trip data. Waypoints are dynamically inserted based on vehicle status and environmental conditions, such as charging station coordinates when the battery level is low. Route preference parameters include options such as avoiding toll roads, prioritizing highways, etc.

[0112] The vehicle control action generation determines device control instructions, target parameters, and execution timing. Device control instructions include air conditioning start-stop, seat heating on-off, etc. Target parameters set specific values, such as an air conditioning target temperature of 22 degrees Celsius. Execution timing is calculated using a look-ahead time algorithm:

[0113]

[0114] where, represents the instruction execution time, represents the planned departure time, represents the time required for device preparation, represents the safety buffer time.

[0115] The service reservation action generation specifies service type, service provider identification, and reservation time parameters. Service types include charging services, parking services, etc. Service provider identification is obtained from user preference data to prioritize selected suppliers. Reservation time parameters are determined based on trip progress to ensure that the reservation time period matches the planned arrival time.

[0116] The action sequence construction process follows the dependency relationship principle. Navigation actions are prioritized over vehicle control actions and service reservation actions. There are resource-dependent action sequence constraints, such as charging station reservations must precede actions that navigate to charging stations. Time-overlapping actions are conflict detected to ensure that mutually exclusive operations are not scheduled at the same time.

[0117] The output data flow process maintains strict format specifications. The recommended script is encapsulated in JSON-LD format, including script metadata, action list, and dependency graph. Each action object contains a type identifier, a parameter set, and an execution condition. The dependency graph uses an adjacency list to represent the sequence constraints between actions, ensuring the logical correctness of script execution.

[0118] This module establishes an output verification mechanism. The generated recommended script is syntax checked to ensure compliance with the predefined JSONschema specification. Semantic verification detects the validity range of action parameters, excluding values outside the reasonable range. Feasibility verification assesses the execution possibility of the action sequence, filtering out actions with unavailable resources or time conflicts.

[0119] ​The script execution engine module is coupled to the large model decision center module, and is used for parsing a recommended script, and calling corresponding downstream system operation according to an action type parsed out, the downstream system including at least two of a navigation system, a vehicle control system and a third-party service API;

[0120] Further, the script execution engine includes:

[0121] The JSON structure of the recommended script is parsed, and an action chain array is extracted;

[0122] The type field and the parameter field of each action are identified;

[0123] The instruction is distributed to the corresponding system adapter according to the action type;

[0124] The general instruction is converted into a special instruction format of the downstream system through the system adapter.

[0125] The script execution engine further includes:

[0126] The navigation system API is called, and the destination coordinates and the path planning parameters are input;

[0127] The vehicle control system API is called, and the device control instruction and the execution time parameter are sent;

[0128] The third-party service OpenAPI is called, and the service booking request and the user authentication information are submitted;

[0129] The execution states of the systems are monitored, and the execution progress is displayed on a graphical interface.

[0130] Specifically, the script execution engine module receives recommended script data from the large model decision center. The module first performs syntax analysis on the script, and extracts the complete structure of the action chain array. The parsing process verifies the compliance of the script format based on a predefined JSONSchema, to ensure that all required fields are complete and the data types are correct.

[0131] In the action recognition stage, each action is classified according to the type field. The navigation class action is identified as NAVIGATION, the vehicle control class action is identified as VEHICLE_CONTROL, and the service booking class action is identified as SERVICE_BOOKING. The parameter field is specifically parsed according to the action type. The navigation action parameters include the destination coordinates, the waypoint list and the path preference setting, the vehicle control action parameters include the device identifier, the control instruction and the target value, and the service booking action parameters include the service type code, the supplier identifier and the booking time window.

[0132] The system adopts an instruction distribution algorithm to determine the target system of each action:

[0133] ;

[0134] in Indicates action The target system to be allocated, Indicates the first One available downstream system A matching function representing action characteristics and system capabilities. Indicates action eigenvectors, system representation The Each capability descriptor. The matching function returns a boolean value, which is 1 if the action requirement is within the system's capabilities, and 0 otherwise.

[0135] The adapter conversion process maps generic commands to system-specific commands. The navigation adapter converts destination coordinates into GeoJSON format accepted by the navigation system, and route preference parameters into navigation system-specific option codes. The vehicle control adapter converts device control commands into message formats conforming to the CAN bus protocol, and execution time parameters into absolute timestamps based on the system clock. The service adapter encapsulates reservation requests into RESTful requests conforming to the OpenAPI specification, and user authentication information is securely transmitted via the OAuth 2.0 protocol.

[0136] Execution status monitoring employs a multi-threaded architecture for parallel processing. Each action is assigned an independent monitoring thread to continuously detect changes in execution status. The status evaluation formula is: ;in Indicates the overall execution progress. Indicates the total number of actions. This indicates the action completion indicator function, which is activated when the action is completed. The value is 1 upon completion and 0 otherwise. Monitoring data is updated to the shared state memory in real time to ensure state consistency among multiple monitoring threads.

[0137] The graphical user interface (GUI) module reads execution progress data from the state storage. The progress display uses a layered visualization scheme: the first layer shows the overall completion rate, the second layer shows the detailed status of each action, and the third layer shows potential anomalies. The user interface provides control options for pausing, resuming, and canceling actions; these operations are immediately synchronized to the execution state manager.

[0138] An exception handling mechanism detects faults during execution. When an action times out or returns an error code, the exception handler selects a response based on a predefined policy table. Possible responses include retrying execution, skipping the current action, or replacing it with a backup action. The exception event is logged to the system log and simultaneously triggers a user notification process.

[0139] The data output process maintains a loosely coupled architecture with the downstream system. The execution engine sends path planning requests to the navigation system through the message queue, sends device instructions to the vehicle control system through the vehicle bus, and sends API calls to third-party services through the HTTPS protocol. Each output channel is configured with a retry mechanism and timeout settings to ensure reliable transmission of instructions.

[0140] A system control module is coupled to each module to dynamically coordinate the execution timing of each module based on the planned time window in the trip data, forming a closed-loop processing flow from trip planning to automatic script execution.

[0141] Further, the system control includes:

[0142] Real-time monitoring of the deviation between system time and trip planning time window;

[0143] Dynamically adjusting the data collection frequency of the data integration engine according to the trip progress;

[0144] Based on the planned departure time and real-time traffic estimation, the optimal decision-making time is triggered to trigger large model inference;

[0145] Coordinating the execution order and time interval of each action in the script execution engine;

[0146] Through the feedback mechanism, the modules work together to form a complete closed-loop control system.

[0147] Specifically, the system control module obtains the current system time through the high-precision clock service and compares it with the trip planning time window in real time. The time deviation calculation uses the following formula: ; Wherein represents the time deviation value, represents the current system time, represents the scheduled time in the planned time window. This deviation value is used to assess the trip execution progress status.

[0148] The data collection frequency adjustment is based on the trip progress status and is dynamically performed. The collection frequency calculation formula is:

[0149] ;

[0150] Wherein, represents the actual collection frequency, represents the basic collection frequency, represents the adjustment coefficient, represents the time deviation value. When the time deviation increases, the system increases the data collection frequency to obtain more timely environmental state information.

[0151] The decision trigger timing calculation comprehensively considers the planned departure time and real-time traffic factors. The trigger time point calculation formula is: ; wherein represents the large model inference trigger time, represents the planned departure time, represents the estimated travel time, represents the real-time congestion coefficient, represents the buffer time. This calculation ensures that the system starts the decision-making process at the right time.

[0152] The action execution coordination mechanism adopts a priority scheduling algorithm. The execution order is determined according to the action type and dependency relationship, and the execution time interval is calculated by the following formula: ; wherein, represents the action execution time interval, represents the available time window, represents the number of actions to be executed. The system dynamically allocates the execution time slice of each action according to the actual available time.

[0153] The feedback mechanism establishes a module state monitoring vector: ; wherein, represents the system state vector, represents the trip planning module state, represents the data integration engine state, represents the large model decision center state, represents the script execution engine state. State monitoring is sampled at a fixed frequency to ensure timely detection of inter-module collaboration abnormalities.

[0154] Closed-loop control is achieved through state feedback regulation. The control algorithm calculates the adjustment amount according to the system state vector:

[0155] ;

[0156] wherein, represents the control output, represents the system state deviation, , , represent the proportional, integral, and derivative coefficients, respectively. This algorithm ensures that each module remains synchronized in terms of execution timing.

[0157] The data input process receives status reports from each module, including the trip progress of the trip planning module, the data readiness status of the data integration engine, the inference completion status of the large model decision center, and the action execution status of the script execution engine. These status information is summarized to the system control center for unified processing.

[0158] The data output process sends control instructions to each module, including adjusting the collection parameters of the data integration engine, triggering the inference timing of the large model decision center, and coordinating the action sequence execution rhythm of the script execution engine. The control instructions are transmitted through the message middleware to ensure reliable delivery and sequential execution of the instructions.

[0159] The system establishes an exception handling mechanism. When a module execution timeout or state exception is detected, an emergency handling process is started. Emergency handling includes rescheduling execution timing, switching to a backup execution path, notifying users to intervene in processing, and other measures to ensure the basic operation capability of the system under abnormal conditions.

[0160] Embodiment two:

[0161] Figure 3 is a method flowchart of a method for generating a recommended script based on a short-term self-driving trip provided by a specific embodiment of the present application.

[0162] In the Beijing one-day self-driving tour scenario, the user plans to visit the Forbidden City in the morning and visit the Summer Palace in the afternoon. When the system detects that the vehicle's power is insufficient, the road is congested, and the temperature in the vehicle is too high, the traditional solution requires the user to operate the navigation system to query the route, find a charging station, remotely start the air conditioner, and manually calculate the departure time to coordinate the charging, driving, and pre-cooling needs. This dispersed operation mode exposes the fundamental problems existing in the prior art: because the trip planning, vehicle status, environmental data, and service resources belong to different systems and lack effective coordination mechanisms, the user must actively discover the needs and manually switch between multiple independent applications, which cannot generate comprehensive decisions based on real-time situations and cannot automatically coordinate the execution order across systems, ultimately causing heavy decision-making burden, low execution efficiency, and suboptimal resource utilization. To solve the above problems, a method for generating a recommended script based on a short-term self-driving trip is provided, as shown in the structure of Figure 3 as shown. The specific implementation process of the method is as follows:

[0163] S1 obtains the user's set trip data, including trip points and planned time windows, and converts the user input into machine-readable trip planning data. S2 collects vehicle status and environmental data in real time, integrates the trip data with real-time data to generate a unified structured data package, and solves the standardization problem of multi-source heterogeneous data.

[0164] S3 performs situational reasoning on the structured data package through a large model to generate a recommended script containing at least two action types of navigation, vehicle control, and service reservation. This step realizes the conversion from data to decision, and analyzes the correlation between trip progress, vehicle status, and environmental conditions through the reasoning capability of the large model. S4 parses the recommended script and calls the corresponding downstream system according to the action type, converts the abstract instructions into specific system executable operation instructions, and completes the transition from decision to execution.

[0165] S5 dynamically coordinates the execution timing of each module based on the planned time window, through real-time monitoring of time deviation, adjusting data acquisition frequency, triggering decision timing and coordinating action execution sequence, forming a closed-loop control from trip planning to automatic execution of scripts. This step ensures that each module is activated at the correct time node, maintaining the coordinated operation of each part of the system.

[0166] The technical scheme of the method realizes multi-dimensional data fusion of trip planning data, real-time vehicle state, environmental conditions and service resources. Through the context reasoning capability of the large model, an executable decision script is generated, which is then converted into specific system instructions by the execution engine, and finally forms a closed-loop process through timing control. The entire process replaces the traditional mode of manual operation by the user between multiple independent systems, solving the technical problems of data fragmentation, passive decision-making and dispersed execution.

[0167] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A system for automatically generating a recommended script based on a short-term self-driving trip, the system comprising: The system comprises the following modules: A trip planning module for obtaining self-driving trip data set by a user, including at least one trip point and corresponding planned time window; A data integration engine module coupled to the trip planning module for real-time collection of vehicle state data and real-time environmental data, and integration of the trip data, vehicle state data and real-time environmental data to generate a unified structured data package; A large model decision center module coupled to the data integration engine module for receiving the structured data package and generating a recommended script containing at least two ordered actions through situational reasoning, wherein the types of actions include at least two of navigation, vehicle control and service reservation; A script execution engine module coupled to the large model decision center module for parsing the recommended script and calling corresponding downstream systems to perform operations according to the parsed action types, the downstream systems including at least two of a navigation system, a vehicle control system and a third-party service API; A system control module coupled to each module for dynamically coordinating the execution timing of each module based on the planned time window in the trip data, forming a closed-loop processing flow from trip planning to automatic script execution. 2.The system of claim 1, wherein, The trip planning includes: Receiving multiple trip points input by a user through a graphical interface, each trip point containing latitude and longitude coordinates or a POI name; Configuring a planned arrival time and a planned departure time for each trip point to form a planned time window; Organizing the trip points and corresponding planned time windows into an ordered trip sequence to generate machine-readable trip data. 3.The system of claim 1, wherein, The data integration engine includes: Collecting vehicle state data through a vehicle CAN bus interface, including battery remaining capacity, current latitude and longitude coordinates of the vehicle and in-vehicle temperature sensor data; Calling a traffic data service API through a wireless network to obtain a real-time traffic congestion index and calling a meteorological data service API to obtain weather condition information; Packaging the trip data, vehicle state data and real-time environmental data in a predefined JSON format.

4. The system for generating a recommended script based on a short-term self-driving trip according to claim 1, wherein, The data integration engine further includes: Standardizing the time information in the trip data to a 24-hour time format; Converting the power data in the vehicle state data to a percentage and unifying the position data to a coordinate system; Normalizing the traffic congestion index in the real-time environmental data and encoding the weather information; Packing the processed data according to the predetermined structure to generate a complete structured data package containing a timestamp.

5. The system for generating a recommended script based on a short-term self-driving trip according to claim 1, wherein, The large model decision center includes: Combining the structured data package with a predefined prompt word template to form large model input data; Analyzing the correlation between trip progress, vehicle state and environmental conditions through multiple rounds of reasoning on the input data by a large language model; Generating an ordered action sequence containing navigation instructions, vehicle control instructions and service reservation instructions based on the reasoning results.

6. The system for generating a recommended script based on a short-term self-driving trip according to claim 1, wherein, The large model decision center further includes: Generating a navigation action including destination coordinates, waypoints and path preference parameters; Generating a vehicle control action including device control instructions, target parameters and execution timing; generating a service reservation action including a service type, a service provider identity, and a reservation time parameter; arranging the actions in a time sequence and adding dependencies between the actions to form a complete recommended script.

7. The system for generating a recommended script based on a short-term self-driving trip according to claim 1, wherein, The script execution engine includes: parsing the JSON structure of the recommended script to extract an action chain array; identifying the type field and the parameter field of each action; distributing the instructions to the corresponding system adapter according to the action type; converting the general instructions into a downstream system-specific instruction format through the system adapter. 8.The system of claim 1, wherein, The script execution engine further includes: calling the navigation system API and inputting the destination coordinates and path planning parameters; calling the vehicle control system API and sending device control instructions and execution time parameters; calling the third-party service OpenAPI and submitting service reservation requests and user authentication information; monitoring the execution status of each system and displaying the execution progress on the graphical interface. 9.The system of claim 1, wherein, The system control includes: real-time monitoring of the deviation between the system time and the travel plan time window; dynamically adjusting the data collection frequency of the data integration engine according to the travel progress; estimating the optimal decision-making opportunity based on the planned departure time and real-time traffic conditions, triggering the large model reasoning; coordinating the execution order and time interval of each action in the script execution engine; ensuring the coordinated work of each module through the feedback mechanism to form a complete closed-loop control system.

10. A method for automatically generating a recommended script based on a short-term self-driving trip, characterized in that, A system for generating a recommended script for a short-term self-driving trip according to any one of claims 1-9, comprising the following steps: S1, obtaining user-set self-driving trip data containing at least one trip point and corresponding planned time window; S2, coupled to the trip planning module, for real-time collection of vehicle state data and environmental real-time data, and integration of the trip data, vehicle state data, and environmental real-time data to generate a unified structured data package; S3, coupled to the data integration engine module, for receiving the structured data package and generating a recommended script containing at least two ordered actions through context reasoning, wherein the action types include at least two of navigation, vehicle control, and service reservation; S4, coupled to the large model decision center module, for parsing the recommended script and calling the corresponding downstream system execution operation according to the parsed action type, the downstream system including at least two of a navigation system, a vehicle control system, and a third-party service API; S5, coupled to each module respectively, for dynamically coordinating the execution timing of each module based on the planned time window in the trip data, forming a closed-loop processing flow from trip planning to automatic script execution.

Images