Intelligent vehicle test driving sales interaction method and device

Abstract

The invention discloses an intelligent vehicle test driving sales interaction method and device, and the method comprises the steps: obtaining the natural language input data of a user in response to the interaction triggering operation of the user for a target vehicle; analyzing a user demand tag set representing the car purchase intention and preference dimension of the user; screening a to-be-displayed target vehicle function set from a displayable function set of the target vehicle based on the user demand tag set, and determining a target experience scene set matched with the target vehicle function set according to a preset function scene mapping relationship; and acquiring the current position of the target vehicle as a route starting point, determining road network data based on the map data, and positioning a scene section corresponding to the target experience scene set in the road network data so as to plan and generate a personalized test driving route capable of covering the target experience scene set on the premise of meeting a preset route constraint condition. Therefore, the active test driving experience organization driven by the vehicle purchase intention and preference of the user is realized.

Description

Technical Field

[0001] This application relates to the field of vehicle marketing system technology, and in particular to an intelligent vehicle test drive sales interaction method and device. Background Technology

[0002] Currently, the automotive sales industry primarily relies on the traditional "person-to-person" interaction model, where users obtain information in a relatively passive and singular way. In showrooms or test drive sessions, vehicle information is mainly conveyed through verbal introductions by sales staff, brochures, or static video displays.

[0003] Although some systems have introduced touchscreens or basic voice assistants for assistance, most of these systems are in a passive waiting state for instructions and lack the ability to actively recognize user intentions and guide interactions. Such one-way and rigid information output methods are difficult to arouse users' deep interest and cannot reflect the high-tech attributes of smart cars.

[0004] Furthermore, the test drive is a crucial step in sales conversion, yet current systems suffer from severe gaps in the experience and a lack of personalization. Traditional test drives typically follow fixed routes and standardized feature introductions, failing to adapt in real time to the user's specific usage scenarios (such as urban commuting or family trips) or key concerns (such as intelligent driving features and power performance). Summary of the Invention

[0005] This application provides an intelligent vehicle test drive sales interaction method, device, storage medium, and program product to at least solve one of the above-mentioned technical problems.

[0006] In a first aspect, embodiments of this application provide an intelligent vehicle test drive sales interaction method, comprising: responding to a user's interactive trigger operation on a target vehicle, acquiring the user's natural language input data, wherein the natural language input data includes voice input data and / or text input data; performing semantic parsing on the natural language input data to generate a user demand tag set representing the user's car purchase intention and preference dimensions; filtering a set of target vehicle functions to be displayed from the set of displayable functions of the target vehicle based on the user demand tag set, and determining a set of target experience scenarios matching the set of target vehicle functions according to a preset function scenario mapping relationship; wherein the function scenario mapping relationship defines the association between vehicle function identifiers and experience scenario types or experience scenario road segment identifiers; acquiring the current location of the target vehicle as the route starting point, determining road network data based on map data, and locating the scene road segment corresponding to the set of target experience scenarios in the road network data, so as to plan and generate a personalized test drive route that can cover the set of target experience scenarios under the premise of satisfying preset route constraints.

[0007] Secondly, embodiments of this application provide a storage medium storing one or more programs including execution instructions, which can be read and executed by electronic devices (including but not limited to computers, servers, or network devices) to perform the steps of the method described above in this application.

[0008] Thirdly, a computer device is provided, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the steps of the method described above in this application.

[0009] Fourthly, embodiments of this application also provide a computer program product, the computer program product including a computer program stored on a storage medium, the computer program including program instructions, which, when executed by a computer, cause the computer to perform the steps of the method described above.

[0010] Fifthly, embodiments of this application also provide a mobile platform, including: a platform body; and a control system disposed on the platform body; wherein the control system is configured to execute the steps of the above method to manage intelligent driving scenarios during the operation of the mobile platform.

[0011] The beneficial effects of the embodiments of this application are as follows: By performing semantic parsing on user natural language input data to generate a set of user demand tags, and based on these tags, the target vehicle function set and target experience scenario set are selected. Further, a correspondence is established between the target vehicle function set and experience scenario types or scenario road segment identifiers according to the function scenario mapping relationship. Based on the vehicle's current location and road network data, the scenario road segments corresponding to the target experience scenario set are located. Under the premise of meeting preset route constraints, a personalized test drive route covering the scenario road segments is planned and generated. This achieves proactive test drive experience organization driven by user purchase intentions and preferences, ensuring consistent matching between function displays and experiential scenarios, improving the relevance, coherence, and feasibility of the test drive process, and alleviating the lack of personalization and experience gaps caused by fixed test drive routes and standardized content in traditional test drive systems. Attached Figure Description

[0012] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 A flowchart illustrating an example of an intelligent vehicle test drive sales interaction method according to an embodiment of this application is shown; Figure 2 A flowchart illustrating an example of generating a user requirement tag set in a method according to an embodiment of this application is shown. Figure 3 A flowchart illustrating an example of generating a personalized test drive route according to an embodiment of this application is shown. Figure 4 This document shows an example of an operation flowchart of a method based on the confidence-weighted fusion of multimodal psychological features based on semantic conflict in an embodiment of this application. Figure 5 A flowchart illustrating an example of adaptive interactive guidance and vehicle parameter fine-tuning control based on psychological state categories in a method according to an embodiment of this application is shown. Figure 6 This paper presents an example of an operation flowchart for functional adaptation evaluation and recommendation output based on test drive behavior logs in a method according to an embodiment of this application; Figure 7 This is a schematic diagram of the structure of an embodiment of the electronic device of this application. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application. It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other.

[0015] It should also be noted that, in this document, the terms "comprising" or "including" include not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0016] It should be noted that, in current technologies, some brands have introduced touchscreen queries or basic voice assistants into vehicles to provide searchable displays of vehicle parameters, configurations, and function descriptions. However, most of these interactive components still operate on a "command-triggered - passive response" model, lacking the ability to proactively greet and guide users when they enter the vehicle or begin a test drive. Consequently, it is difficult to continuously clarify user concerns and form a continuous profile of their needs across multiple interactions.

[0017] Furthermore, regarding the recommendation and display process, current technologies often use fixed rules or preset scripts to output recommended content. This makes it difficult to structure the car purchase intentions and preferences reflected by users' natural language input into a calculable tag system. It is also difficult to further map this tag system to the correspondence between the vehicle's displayable functions and experience scenarios. As a result, the display content and display order are fixed, making it difficult to achieve personalized test drive function explanations and experience organization.

[0018] Furthermore, while the test drive phase is a crucial conversion stage, current technologies generally lack effective utilization of users' nonverbal feedback. A user's facial expressions, gaze, and gestures can reflect their state of confusion, tension, or increased interest, but current technologies often lack real-time perception and adaptive adjustment mechanisms for these states. This makes it difficult to dynamically adjust the depth of explanation, the pace of interaction, or parameter suggestions accordingly, thus affecting the efficiency of comprehension and the establishment of trust during the test drive.

[0019] It should be understood that the above description of the relevant technologies is intended only to help the public better understand the inventive spirit and motivation of this application, and is not intended to limit this application. Furthermore, the technical solutions described in the above-mentioned relevant technologies are not prior art, and may also be undisclosed technical solutions, such as those under research or in the laboratory stage.

[0020] The technical solutions in this application, including the collection, storage, use, processing, transmission, provision, and disclosure of users' personal information, comply with relevant laws and regulations and do not violate public order and good morals.

[0021] Figure 1 A flowchart illustrating an example of an intelligent vehicle test drive sales interaction method according to an embodiment of this application is shown.

[0022] Regarding the execution subject of the method in the embodiments of this application, it can be any controller or processor with computing or processing capabilities. In some examples, the method in the embodiments of this application can be integrated and configured in an electronic device or terminal through software, hardware or a combination of software and hardware, and the type of terminal or electronic device can be diverse, such as mobile phone, tablet computer, desktop computer or vehicle terminal, etc.

[0023] For example, the execution entity of the method in this application embodiment may be a test drive interaction control module integrated in the vehicle intelligent cockpit domain controller. This test drive interaction control module can be executed by the processor (CPU and / or AI computing unit) calling the program instructions stored in the memory to realize semantic parsing of the user's natural language input, generation of user demand tag set, determination of target vehicle function set and target experience scenario set, and planning and generation of personalized test drive route. The test drive interaction control module can communicate with the vehicle's human-machine interaction unit (such as microphone array, speaker, touch screen) and vehicle positioning and map service unit (such as GNSS positioning module, map component or cloud map service interface) to obtain voice / text input data, vehicle current location and road network data, and output human-machine interaction information or route navigation information for test drive guidance.

[0024] like Figure 1 As shown, in step S110, in response to the user's interactive trigger operation on the target vehicle, the user's natural language input data is acquired, which includes voice input data and / or text input data.

[0025] In some implementations, the vehicle terminal responds to interactive triggering operations by listening to preset trigger signals (such as a signal that a user enters the cabin or an active wake-up operation based on the screen / button), thereby activating the front-end input interface of the in-vehicle intelligent assistant.

[0026] For example, the system collects the user's voice signal in real time through a microphone array deployed in the cockpit, and uses an automatic speech recognition (ASR) module to perform noise reduction, echo cancellation and transcription processing to convert the voice stream into text data; at the same time, the system can also support receiving the user's text input through the central control screen or associated mobile terminal.

[0027] In step S120, the natural language input data is semantically parsed to generate a set of user demand tags that represent the user's car purchase intention and preference dimensions.

[0028] Here, after acquiring the natural language text, the system utilizes a Natural Language Processing (NLP) engine to perform deep semantic analysis on the unstructured input data. In some implementations, the semantic analysis process goes beyond simple keyword matching and may also include intent recognition and entity extraction. On one hand, it identifies the user's implicit macro-level car-buying intent in the dialogue (e.g., whether they prefer "understanding intelligent driving performance" or "focusing on family comfort"); on the other hand, it precisely extracts specific preference-related entities from the text (such as "budget range," "acceleration performance requirements," "comparison with specific competitors," etc.). Furthermore, the system maps this parsed information to a predefined standardized tagging system, generating a structured set of user demand tags.

[0029] In step S130, the set of target vehicle functions to be displayed is filtered from the set of displayable functions of the target vehicle based on the user demand tag set, and the set of target experience scenarios that match the set of target vehicle functions is determined according to the preset function scenario mapping relationship.

[0030] Here, the functional scenario mapping relationship defines the association between vehicle function identifiers and experience scenario types or experience scenario road segment identifiers.

[0031] In some implementations, a weighted search is performed in the cloud or local vehicle function database based on the generated user demand tag set, and the target vehicle function set with the highest matching degree with the current user tag (i.e. the selling points that are the focus of this test drive) is selected from the full range of functions that the vehicle has.

[0032] Specifically, when filtering the set of target vehicle features, the system can employ a two-stage matching algorithm. The first stage is hard filtering, which performs an initial screening of the feature library based on hard conditions such as budget range and power preference in user tags; the second stage is soft scoring, which calculates a personalized matching score. An exemplary scoring formula can be expressed as: Equation (1) In the formula, For user number The weight of each demand tag, The system assigns a score to the matching degree of the function on the corresponding tag. The system then sorts the functions in descending order of total score and selects the Top N functions as the target vehicle function set.

[0033] Furthermore, the system supports dynamic linkage of demonstration content. For the same selected function, the system calls up different explanatory materials based on the different focuses of the user profile. For example, for users tagged "Safety First," when introducing the Automatic Emergency Braking (AEB) function, the system prioritizes displaying a "test video of a ghost lurking in a sudden incident"; while for "Technology Adventurer" users, the system prioritizes displaying a "cool rendering animation of the memory parking function," thus achieving a personalized content display strategy.

[0034] Subsequently, the system invokes a pre-built database of functional scenario mapping relationships. This database logically defines the corresponding links between technical functions and physical environments (for example, "City Navigation Assist" corresponds to "complex urban intersections" or "congested road sections," and "Air Suspension Comfort" corresponds to "speed bumps" or "potholes"). Based on this mapping relationship, the system can automatically deduce the "target experience scenario set" that needs to be covered to fully demonstrate these functions, thereby ensuring that the subsequent test drive experience is highly targeted, avoiding showing users functions that they are not interested in, and effectively improving the efficiency of information delivery.

[0035] In step S140, the current location of the target vehicle is obtained as the starting point of the route. Road network data is determined based on map data, and the scene road segment corresponding to the target experience scene set is located in the road network data, so as to plan and generate a personalized test drive route that can cover the target experience scene set under the premise of meeting the preset route constraints.

[0036] Here, the system uses geographic information technology to transform abstract experience scenarios into executable physical paths. In some implementations, starting from the target vehicle's current Global Positioning System (GPS) coordinates, the system loads map data (such as high-precision map data) to obtain the surrounding road network topology and attribute information.

[0037] Specifically, the system retrieves and locates specific geographical road segments in the road network that correspond to the set of target experience scenarios (for example, the system will look for nearby streets that meet the "congestion" attribute or ramps that meet the "high curvature" attribute on the map), and marks these road segments as planned must-pass areas or high-priority areas. Under the premise of strictly adhering to the preset route constraints such as the total test drive time, total mileage limit, and traffic regulations, the path planning algorithm calculates a personalized test drive route that can maximize the connection or coverage of these target scenario road segments.

[0038] This application breaks away from the traditional fixed route pattern of test drives, which is "one-size-fits-all," and achieves the effect of dynamic planning of personalized test drive routes. This allows users to experience vehicle performance in the most suitable real-world scenarios, significantly enhancing the persuasiveness and immersion of the test drive.

[0039] Regarding the implementation details of obtaining natural language input data in step S110, in some examples of embodiments of this application, if the natural language input data includes voice input data, then vehicle status data collected by a multimodal sensor group deployed in the cockpit of the target vehicle is obtained.

[0040] Here, to achieve intelligent proactive interaction, the system does not rely on a single wake word, but rather on a multimodal sensor array deployed within the target vehicle's cabin for comprehensive state perception. In some implementations, this sensor array enters a low-power monitoring mode when the vehicle is in standby or unlocked state, collecting real-time data on the in-vehicle environment and occupant status. For example, this may include collecting pressure distribution data of the seat via a pressure sensor or gravity sensor array embedded within the driver's seat, capturing visual image streams within the cabin via a driver monitoring system (DMS) camera mounted on the A-pillar or rearview mirror, and reading real-time door opening / closing status, door lock status, and vehicle power status signals via the vehicle's body control module (BCM).

[0041] Then, when the vehicle status data meets the preset welcome interaction conditions, it is determined that a valid interaction trigger operation has been detected; the welcome interaction conditions include one or more of the following: the pressure value detected by the gravity sensor of the driver's seat is within the preset adult weight range, the in-vehicle camera recognizes a valid facial feature of a non-staff member who is looking forward, and the door closing signal is received from the body controller.

[0042] Here, after obtaining the above vehicle status data, the system will execute a logical verification algorithm to determine whether to trigger the welcoming process, so as to avoid accidental triggering (such as placing heavy objects or cleaning staff operations).

[0043] In some implementations, the logic verification algorithm can perform multi-dimensional filtering based on preset welcoming interaction conditions. First, the system determines whether the seat pressure value falls within a preset adult weight range (e.g., 45kg-120kg) to exclude interference from children or backpacks. Second, it uses computer vision algorithms to process camera data, performing face detection and feature comparison. This not only confirms that the occupant is in an "interaction-ready" state, looking directly at the screen or ahead, but also compares the extracted facial features with a preset "staff / salesperson facial feature database," generating a positive signal only when the occupant is identified as a non-staff member (i.e., a potential customer). Finally, combined with the door closing signal from the vehicle body controller, it confirms that the cabin has formed a closed acoustic environment. Only when one or more of the above key conditions (preferably a combination of conditions) are met does the system recognize that a valid interactive trigger operation has been detected, thereby accurately targeting the target user. In addition, to prevent repeated false triggers caused by users frequently adjusting their sitting posture or opening and closing the car door, the system is also equipped with session anti-shake logic, which means that after recognizing the facial features of the same user, the active greeting process will not be triggered repeatedly within a preset time window (such as 30 seconds).

[0044] Subsequently, in response to effective interactive triggering operations, the in-vehicle intelligent assistant's proactive greeting process is activated, and the microphone array is started to collect the user's voice input data during the greeting dialogue.

[0045] Here, once the interactive trigger is deemed valid, the system immediately switches from a silent background state to an active service state, activating the in-vehicle intelligent assistant's proactive greeting process. In some implementations, the system can invoke the text-to-speech (TTS) module to play a pre-set welcome message or guiding dialogue (such as "Hello, welcome to the intelligent cockpit. What functions would you like to learn about?"), breaking the interaction deadlock by proactively asking questions. Simultaneously, the system can also activate the microphone array within the cockpit and utilize beamforming technology to directionally focus the pickup beam towards the driver's position to suppress other noise interference within the vehicle. Subsequently, the system enters a listening state, capturing the user's natural language feedback to the greeting in real time, completing the transition from a "passive waiting to be awakened" to an "actively guiding interaction" mode, ensuring clear and effective voice input data is obtained the moment the user speaks.

[0046] Figure 2 A flowchart illustrating an example of generating a set of user requirement tags in a method according to an embodiment of this application is shown.

[0047] like Figure 2 As shown, in step S210, intent recognition is performed on the natural language input data to determine the target intent category to which the user's car purchase intent belongs.

[0048] In some implementations, a pre-trained natural language processing model (such as a text classification model based on an improved BERT or RoBERTa architecture) can be used to vectorize the user's natural language input data and then input into an intent classifier. This classifier can be fine-tuned and trained on a massive amount of dialogue data from the automotive sales field, enabling it to accurately calculate the confidence probability of the user's input belonging to different preset intent categories. Furthermore, the system can select the category with the highest confidence as the target intent category. For example, "How fast does this car accelerate from 0 to 100 km / h?" can be identified as a "performance parameter inquiry" intent, and "Can I take it for a test drive?" can be identified as a "test drive request" intent. This effectively filters out casual conversation or invalid information and establishes the dominant direction of the current interaction.

[0049] In step S220, key information is extracted from the natural language input data, and key information slot values ​​are extracted from at least one information dimension.

[0050] Here, the information dimensions include any one of the following: budget range, usage scenarios, family structure, power preferences, and focus on intelligent driving.

[0051] In some implementations, based on the determined intent, the system invokes a sequence labeling model (such as BiLSTM-CRF or a Transformer-based entity extraction model) to extract key information from the text at a fine-grained level. Specifically, it can identify and extract entity fragments with specific business meanings from the text based on predefined slot definitions. For example, the system can accurately extract key slot values ​​such as "family structure: two-child family" and "car usage scenario: suburbs / camping" from the sentence "We have two children and want a car that is convenient for camping in the suburbs." Simultaneously, the system also performs corresponding entity recognition for dimensions such as "budget," "power preference," and "intelligent driving focus."

[0052] In step S230, the target intent category is mapped to the intent dimension label, and the key information slot value is mapped to the preference dimension label. The intent dimension label and the preference dimension label are aggregated to generate a user demand label set.

[0053] Here, to eliminate the diversity and ambiguity of natural language expressions (for example, a user might say "strong push-back feeling" or "fast acceleration," both referring to the same power demand), a standardized mapping operation can be performed. In some implementations, the system queries a pre-built thesaurus or knowledge graph, mapping the identified target intent category to standardized intent dimension labels (such as Intent_TestDrive), and normalizing the extracted key information slot values ​​to preference dimension labels (such as Preference_Power_High). Subsequently, the system logically aggregates these two types of labels to construct a user demand label set in JSON or XML format containing full-dimensional user characteristics, realizing the translation from raw language to machine labels.

[0054] In step S240, a temporary identity is created for the current user, and the user demand tag set is bound to the temporary identity to establish user profile data that supports synchronization across vehicle terminals.

[0055] Here, in order to support continuous interaction across scenarios, the system generates a unique temporary identity identifier for the current visitor in the background user database while generating a tag set (such as a SessionID generated based on UUID or associated with a temporary FaceID).

[0056] In some implementations, the system uses key-value pair storage to strongly bind the generated set of user demand tags to the temporary identity and stores it in a high-speed cache on the cloud or edge server, ensuring the immediate availability and fluidity of user data.

[0057] In terms of business scenarios, when a user moves from a static display vehicle in the showroom to a test drive vehicle outdoors, the test drive vehicle terminal only needs to recognize the user's identity to instantly retrieve the user's tag stored in the cloud. This eliminates the need for the user to repeat their needs verbally, thus achieving a seamless cross-terminal experience from static vehicle experience in the showroom to dynamic test drive of vehicle outdoors.

[0058] In some cases, to ensure the security and convenience of the identity association process from static display vehicles in the showroom to outdoor test drive vehicles, the system supports multiple association handshake methods. For example, the first method is sales consultant-assisted association, where the sales consultant manually pairs the session ID generated by the showroom vehicle with the prepared test drive vehicle using a handheld tablet in the background; the second method is QR code authorization association, where the user scans a QR code inside the test drive vehicle to authorize the system to retrieve the profile data previously generated in the showroom vehicle; the third method is seamless association, where, with the user's authorization to enable biometric recognition, the test drive vehicle automatically retrieves a matching temporary identity identifier from the cloud using the recognized facial feature token, achieving a seamless "synchronization upon boarding" connection.

[0059] Figure 3 A flowchart illustrating an example of generating a personalized test drive route according to an embodiment of this application is shown.

[0060] like Figure 3 As shown, in step S310, a weighted road network search map is constructed based on the road network data, and the scene road segments that have been located in the road network data are marked as reward road segments in the weighted road network search map.

[0061] In some implementations, the system first abstracts the acquired road network data into a computer-processable directed weighted graph data structure; in this graph, road intersections are mapped as nodes of the graph, the actual roads connecting the intersections are mapped as edges of the graph, and the length or travel time of the road is set as the basic travel cost of the edge.

[0062] Building upon this foundation, the system performs a crucial reward labeling operation. Specifically, it traverses the road segments corresponding to the target experience scenario set located in previous steps, finds specific edges corresponding to these road segments in the search graph, and assigns these edges a positive "scenario reward value." For example, a bumpy road segment that showcases the vehicle's suspension vibration damping performance is given a high reward score, while an ordinary road segment has a reward score of zero. In this way, the experiential value of the physical world is digitized into mathematical weights in a graph theory model, constructing a composite road network map that includes both travel costs and experiential benefits.

[0063] In step S320, preset route constraints, including a total test drive duration threshold and / or a total test drive mileage threshold, are obtained.

[0064] Here, to ensure that the generated test drive route conforms to the actual specifications of sales operations, the system reads and loads preset hard constraint parameters, mainly including a total test drive duration threshold (e.g., 30 minutes) and a total test drive mileage threshold (e.g., 15 kilometers). In some implementations, the constraints may originate from the dealer's standardized test drive management regulations, the time window set by the user during the reservation, or be dynamically calculated based on the vehicle's current remaining range. These constraints act as cutoff conditions for the search boundaries in subsequent algorithm execution, used to prune in a timely manner during path exploration, filtering out infeasible solutions that, while offering a rich experience, are too time-consuming or too far.

[0065] In step S330, the path search algorithm is run to find the optimal path in the weighted road network search graph.

[0066] Here, the path search algorithm is configured with a scene reward function. When the search path passes through a reward segment, the cumulative reward value of that path is increased through the scene reward function, so as to output the path that maximizes the cumulative reward value while satisfying the preset route constraints, as the personalized test drive route. It should be understood that the type of path search algorithm can be diverse, such as an improved A* algorithm based on multi-objective optimization or an algorithm for solving orienteering motion problems, and no restrictions are imposed here.

[0067] In some implementations, the system's path search algorithm seeks the optimal solution in a weighted road network search graph. This algorithm is equipped with a unique scene reward function, whose core logic differs from traditional navigation's pursuit of "shortest time," instead aiming for "maximum experience benefits." Specifically, during path exploration and expansion, whenever the planned path passes an edge marked as a "reward segment," the scene reward function adds the reward value corresponding to that segment to the current path's total score. By continuously iterating and searching for possible path combinations, and verifying in real time whether the total time and total mileage of each candidate path exceed the thresholds in the route constraints, the system outputs the path with the highest accumulated reward value (i.e., the sum of the number or quality of covered target experience scenarios) as the personalized test drive route, while satisfying all preset route constraints. Thus, within a limited test drive time, it can intelligently connect the functional experience points that users care about most, generating a test drive trajectory with the highest "cost-effectiveness."

[0068] In some examples of embodiments of this application, when the target vehicle is driving along a personalized test drive route, and it is determined that the target experience scenario set contains a specific demonstration scenario that is not suitable for triggering under real road conditions, a virtual simulation mode is activated on the in-vehicle display terminal.

[0069] It is worth mentioning that, in addition to virtual simulation for extreme scenarios, this application embodiment also provides augmented reality (AR) interaction in real-world scenarios. When a vehicle triggers an assisted driving function (such as adaptive cruise control (ACC) to stop the vehicle in front, or lane keeping assist (LCC)) during real-world driving, the system provides visual feedback using the central control screen, digital instrument panel, or augmented reality head-up display (AR-HUD). Specifically, the system overlays and renders the structured data output by the perception module (such as identified lane lines, 3D bounding boxes of vehicles or pedestrians ahead, and the green driving trajectory line planned by the system) as visual graphics onto the driver's real field of vision or the screen display. By adopting a WYSIWYG feedback mechanism, users can intuitively understand the current perception state and decision-making intent of the intelligent driving system, effectively establishing user trust in the system.

[0070] Specifically, when the system detects that it is about to enter a specific demonstration scenario marked as "high-risk" or "not necessarily triggered" from the target experience scenario set (such as an emergency braking scenario where a pedestrian or non-motorized vehicle suddenly crosses the road (commonly known as a "ghost peek"), a scenario where the vehicle in front brakes suddenly to avoid a collision, or an extreme weather perception scenario), the system will automatically determine that actively reproducing these scenarios on real public roads poses a significant safety hazard or legal compliance risk, and belongs to the type "unsuitable for triggering under real road conditions." At this time, the system triggers a mode switching command, activating the virtual simulation mode on the central control screen or the passenger entertainment screen without interfering with the vehicle's underlying power control and the driver's normal driving. Thus, through logical isolation at the software level, the risks of physical testing are cleverly avoided, ensuring the absolute safety and compliance of the test drive process.

[0071] Then, a virtual road environment is constructed based on the graphics rendering engine, virtual target objects are injected into the virtual road environment, and the virtual vehicle model is driven according to the real-time motion state data of the target vehicle to dynamically demonstrate the perception and decision-making process of the intelligent driving system for the virtual target object.

[0072] Specifically, the system utilizes the vehicle's high-performance graphics rendering engine (such as Unreal Engine or Unity) to load and render a high-fidelity 3D virtual road environment in real time, based on the topology of real roads or a pre-set standard test field model. Within this environment, the system dynamically generates and injects virtual target objects that do not exist in reality (such as a virtual pedestrian suddenly crossing the road or a virtual vehicle cutting in front of you) according to a pre-set test script. These virtual objects possess physical properties, including motion trajectories and collision volumes, but exist only in the digital world. This allows the vehicle to construct thrilling and complex virtual traffic scenarios on open or ordinary real roads, greatly expanding the boundaries of the test drive experience.

[0073] Furthermore, the system can read the target vehicle's motion status data in the real world (including vehicle speed, steering angle, yaw rate, and acceleration / deceleration signals) in real time via the vehicle bus (CAN / Ethernet), and map this data to the virtual world. This allows the system to drive the virtual vehicle model in real time, ensuring its posture is synchronized with the real vehicle. Simultaneously, the intelligent driving algorithm runs perception and decision-making logic for virtual target objects in the background (e.g., identifying a virtual pedestrian in the virtual world and triggering AEB braking). For example, the system dynamically presents this perception recognition frame, decision path planning line, and the virtual vehicle's avoidance actions on the screen in picture-in-picture or full-screen animation. Thus, while driving the real vehicle smoothly, the driver can intuitively see how the vehicle's intelligent driving system thinks and responds to sudden dangers, effectively increasing the user's trust in the safety of the intelligent driving system.

[0074] As a preferred embodiment of this application, during the process of the target vehicle driving according to the personalized test drive route, the user's facial images, eye movement trajectory and body posture data can also be collected by sensors.

[0075] Here, to acquire high-precision behavioral data without interfering with the user's test drive experience, the system utilizes a Driver Monitoring System (DMS) acquisition unit deployed in key locations within the vehicle (such as the steering wheel pillar, rearview mirror, and A-pillars). Specifically, this unit integrates infrared (IR) supplemental lighting and a high-frame-rate RGB-IR binocular camera to ensure clear capture of the user's facial texture and pupil features in bright daylight or low-light environments such as tunnels. Simultaneously, the system combines a pressure sensor matrix inside the seat or in-cabin millimeter-wave radar to monitor the user's spinal morphology, head tilt angle, and torso position changes in real time, thereby synchronously acquiring the user's limb posture data and establishing a full-time perception capability for both overt facial expressions and covert body postures.

[0076] Then, discrete emotion categories in facial images are identified, and the user's gaze hotspots in the in-vehicle display interface and the area outside the vehicle are calculated based on eye movement trajectories. Finally, the user's posture characteristics are analyzed based on body posture data.

[0077] In some implementations, after acquiring raw data, the system can run multiple dedicated computer vision and signal processing algorithm models in parallel. For facial images, the system uses a convolutional neural network (CNN) to identify facial key points and extract facial action units, mapping them to discrete emotion categories such as happiness, confusion, and tension, along with their confidence levels. For eye movement trajectories, the system constructs a gaze mapping model, projecting pupil vectors onto a 3D cockpit space model to accurately determine whether the user's gaze is focused on specific UI controls on the central control screen (a hotspot on the in-vehicle display interface) or across the windshield focusing on road conditions (a hotspot in the external area). For limb data, the system analyzes the distribution and frequency of changes in the body's center of gravity, resolving postural features such as leaning forward (high attention), leaning back (relaxation), or frequent adjustments to posture (anxiety). Thus, pixel-level and signal-level raw data are transformed into behavioral descriptors with semantic information.

[0078] Furthermore, discrete emotion categories, gaze hotspots, and posture features are weighted and fused to quantify and generate user psychological state features, which include instantaneous emotion vectors and / or comprehensive immersion indices.

[0079] In some implementations, the aforementioned unimodal features can be input into a multimodal fusion engine (such as an attention-based Transformer fusion network or a Bayesian network) to perform normalization and weighting processing. The system assigns weights based on the contribution of different features to the representation of mental states (for example, facial expressions have a higher weight for emotion determination, while gaze duration has a higher weight for immersion determination), and finally quantifies and outputs two core indicators: an instantaneous emotion vector and a comprehensive immersion index.

[0080] For example, the instantaneous emotion vector is a multi-dimensional numerical vector that accurately characterizes the valence (positive / negative) and arousal (calm / excited) of the user's current emotion; the comprehensive immersion index is a scalar score that reflects the user's focus and engagement in the current test drive. Thus, the system successfully quantifies the elusive user psychology into digital state parameters that can be processed by a computer.

[0081] Figure 4 A flowchart illustrating an example of a method for weighted fusion of multimodal psychological features based on semantic conflict in accordance with embodiments of this application is shown.

[0082] like Figure 4As shown, in step S410, during the process of quantifying and generating user psychological state features, if a semantic conflict is detected between the explicit semantics expressed by the user's voice input data and the implicit emotional tendency expressed by discrete emotion categories or posture features, then a first priority confidence weight is assigned to the voice input data, a second priority confidence weight is assigned to discrete emotion categories or posture features whose duration exceeds a preset time threshold, and a third priority confidence weight is assigned to discrete emotion categories or posture features that are presented instantaneously.

[0083] Here, after acquiring multimodal data, cross-modal sentiment polarity alignment detection is first performed. In some implementations, the system can compare the textual semantic sentiment labels of the voice input (such as "positive / affirmative") with the implicit sentiment tendencies of facial expressions or body language (such as "negative / fear" or "high arousal / nervousness") on the same timeline. When the system detects a significant semantic conflict between the two (i.e., "inconsistency between words and actions," for example, the user verbally says "no problem" but their face shows fear or their body stiffens), in order to prevent the system from making misjudgments that interfere with the user due to oversensitivity, the system initiates a multi-level confidence decision logic.

[0084] Specifically, the weights are dynamically redistributed based on the principle of "explicit intent as the primary driver and implicit state as the reference." The system first identifies the user's explicit voice commands, which, because they represent the user's rational will, are given a high confidence weight (e.g., 0.6) as the first priority. Second, for non-verbal features that last for more than a preset time threshold (e.g., 3 seconds) (such as a persistent tense posture or stiff expression), which reflect the user's true physiological or psychological state, they are given a medium confidence weight (e.g., 0.3) as the second priority. Finally, for micro-expressions or brief movements that appear instantaneously (such as a momentary frown), which may only be a subconscious reaction or noise, they are given a low confidence weight (e.g., 0.1) as the third priority.

[0085] In step S420, the user's psychological state characteristics are recalculated based on the weighted multimodal data to prioritize responding to the user's explicit intentions.

[0086] Based on the dynamically assigned confidence weights, the system invokes a weighted fusion algorithm to recalculate the final user psychological state features, constructing a revised emotion vector space. In some implementations, the feature vector of the speech modality has a higher priority confidence weight, which dominates the direction of the final state vector (i.e., intent classification), while discrete emotion and posture features serve as correction factors, primarily affecting the magnitude (i.e., emotion intensity) or additional dimensions (such as immersion) of the state vector. Thus, the system can ensure that it prioritizes responding to the user's explicit intent at the interaction decision level (i.e., not interrupting the test drive or arbitrarily taking over), reflecting respect for the user's subjective will, while retaining a digital record of potential risks (nervousness). For example, the psychological state ultimately generated by the system might be judged as "appearing calm but harboring underlying tension," thereby achieving an optimal balance between system robustness and sensitivity.

[0087] Figure 5 A flowchart illustrating an example of adaptive interactive guidance and vehicle parameter fine-tuning control based on psychological state categories in a method according to an embodiment of this application is shown.

[0088] like Figure 5 As shown, in step S510, after quantifying and generating the user's psychological state characteristics, the user's psychological state characteristics are comprehensively analyzed to determine the user's current psychological state category.

[0089] In some implementations, a comprehensive analysis can be performed by combining a pre-defined state classifier or decision tree model. This analysis process not only examines the numerical level of a single dimension but also incorporates the combined features of dimensions. For example, the system maps "high arousal + negative emotion" to "tension state," "low arousal + high cognitive load" to "confusion state," and "positive emotion + high immersion" to "pleasure / flow state." Thus, through the division of a multi-dimensional feature space, the system can discretize continuously fluctuating numerical signals into specific psychological state labels that can be executed by a computer.

[0090] In step S521, when the current psychological state category is confused, the vehicle display interface is controlled to present a detailed schematic diagram of the current operating function.

[0091] In some implementations, when the system determines that the user is confused, such as when the vehicle is performing complex automatic lane changing or avoidance maneuvers, the system automatically triggers an explanatory enhancement display strategy to eliminate the user's distrust due to not knowing what the vehicle is doing.

[0092] Specifically, the rendering layer on the central control screen or instrument panel visualizes the thought process of the current intelligent driving system. For example, it highlights key obstacles identified by the system, draws the predicted trajectory line planned by the system, or briefly displays the decision logic of the current function in the form of a pop-up window (such as "A vehicle has been detected cutting in from the left, and we are slowing down to avoid it"), effectively reducing the cognitive load and doubts of the user.

[0093] In step S523, when the current psychological state category is a state of tension and the vehicle speed meets the preset high-speed conditions, the vehicle control parameters are fine-tuned by voice output. The fine-tuning suggestions include reducing the cruising speed or increasing the following distance.

[0094] Here, when the system determines that the user is in a state of tension, it will further introduce the vehicle speed as a secondary verification condition. The system will only intervene when the vehicle is traveling at high speed (such as more than 80 km / h, at which point the user’s risk sensitivity is the highest) and the tension is indeed caused by driving behavior.

[0095] During this process, the state of tension can be triggered based on a tension intensity threshold. Specifically, the tension dimension score in the instantaneous emotion vector can be obtained in real time and compared with a preset tension intensity threshold; only when the score exceeds the threshold does the system determine that the user's tension level has reached a level requiring intervention. Subsequently, the system outputs a fine-tuning plan in a gentle and suggestive tone through the voice assistant, such as "The current speed is relatively fast, should we slightly reduce the cruising speed or increase the following distance?"

[0096] In step S530, in response to a user confirmation command for the fine-tuning suggestion, fine-tuning of the vehicle control parameters is performed within a preset vehicle safety boundary range.

[0097] Here, when the system receives a confirmation command from the user regarding the fine-tuning suggestion (such as a voice reply "okay" or a click of the screen confirmation button), it does not directly overwrite the underlying control parameters. Instead, it first performs a safety check. Specifically, it queries the current vehicle dynamics model and the safety boundary range of the ADAS system (such as minimum safe following distance and minimum highway cruise speed limit). Under the premise of ensuring compliance with regulations and without affecting driving safety, it generates smooth control commands. The system then gradually adjusts the set speed or following distance parameters of the adaptive cruise control (ACC), allowing the vehicle to complete the adjustment through linear deceleration or reversing actions, ensuring safe vehicle operation.

[0098] In addition to control parameters related to power and safety, the system also supports adaptive fine-tuning for cabin comfort features. In some implementations, when the system detects that the user's posture characteristics indicate a "relaxed" or "enjoyed" state (such as leaning back and a stable heart rate), it can automatically fine-tune environmental parameters based on a preset "comfort mode" baseline. For example, it might automatically reduce the air conditioning fan speed by 10% to lower noise, or fine-tune the seat massage intensity to a gentler level. All such adjustments are limited to the vehicle's factory-set safety and comfort range (e.g., ±10%), aiming to further enhance the user's immersive driving experience through detailed optimization.

[0099] In step S540, within the preset monitoring window period after fine-tuning, the changing trend of the user's psychological state characteristics is continuously monitored. If the negative emotion score in the user's psychological state characteristics shows a downward trend and the comprehensive immersion index shows an upward trend, the current interaction strategy is determined to be effective, and a positive feedback signal is generated.

[0100] In some implementations, after parameter fine-tuning is completed, the system can automatically open a preset monitoring window (e.g., 30 to 60 seconds after execution). During this period, the system continuously samples and calculates the user's psychological state characteristics, focusing on the derivatives or trends of their changes. If the system detects a significant downward trend in the negative emotion score, representing a negative experience, and a rebound in the comprehensive immersion index, representing experience engagement, the system logically determines that the fine-tuning strategy (such as reducing vehicle speed) has successfully alleviated the user's anxiety, and is therefore an effective interaction method. Based on this determination, the system internally generates a logically positive feedback signal, which can be used as a reward marker for system self-optimization.

[0101] In step S550, based on the positive feedback signal, the stress intensity threshold is reduced within the current session period.

[0102] Here, in response to the aforementioned positive feedback signal, the system dynamically adjusts the algorithm parameters within the current test drive session, and appropriately lowers the threshold of stress required to trigger similar fine-tuning suggestions (e.g., lowering the trigger threshold from 0.8 to 0.7). For example, the system learns that the current user finds the strategy of "actively reducing difficulty / speed" very effective; therefore, in the following journey, the system should become more sensitive, intervening earlier to provide assistance as soon as the user shows slight signs of stress. Thus, through this closed-loop optimization within a single session, the vehicle can quickly adapt to the different psychological tolerance and driving style preferences of various users.

[0103] Figure 6 A flowchart illustrating an example of the operation of the functional adaptability assessment and recommendation output based on test drive behavior logs in the method according to an embodiment of this application is shown.

[0104] like Figure 6 As shown, in step S610, after the target vehicle completes the personalized test drive route, the behavior log of the entire test drive process is obtained.

[0105] Here, the behavior log records the scene coverage depth, user attention duration, and frequency of active interaction in each scenario within the target experience scenario set.

[0106] In some implementations, when the test drive ends (the system detects that the vehicle has shifted into Park and navigation has ended), the system immediately freezes and retrieves the full-dimensional behavior log stored in the onboard computing unit. This log is not a simple video recording, but a structured data stream aligned to the timeline. The system correlates vehicle perception data (such as the number of lane changes triggered and the duration of traffic jams) with user interaction data.

[0107] For example, the log records in detail three core dimensions of data for each target experience scenario: the first is the scenario coverage depth, that is, whether the user has fully experienced the entire process of the function in a specific scenario (such as automatic parking space) or only made a superficial attempt; the second is the user attention dwell time, which is the cumulative time that the user gazes at the smart driving visualization interface on the instrument panel or the functional area of ​​the central control screen based on eye tracker data; the third is the frequency of active interaction, which records the number of times the user asks about specific functions by voice (such as "How do I turn on this ACC?"), touches related buttons by physical touch, and actively intervenes in driving during the smart driving period (such as lightly pressing the brake pedal, taking over the steering wheel, or adjusting the following distance lever).

[0108] In step S620, a multi-factor evaluation is performed based on the behavior log to calculate the compatibility score between the target vehicle's various intelligent driving functions and the current user.

[0109] Here, the fit score is obtained by weighted summation of the scene relevance sub-score, the user engagement sub-score, and the explicit interest sub-score. The scene relevance sub-score is calculated based on the scene coverage depth, the user engagement sub-score is calculated based on the user attention dwell time, and the explicit interest sub-score is calculated based on the frequency of active interaction.

[0110] In some implementations, the system establishes a multi-factor weighted evaluation model to quantify the degree of matching between the vehicle's functions and the user's needs, and can calculate the adaptation score for each intelligent driving function (such as NOA, APA, AEB, etc.).

[0111] For example, firstly, a sub-score for scene relevance is calculated based on scene coverage depth. If a user spends a long time on highways and frequently activates cruise control during a test drive, this sub-score will be high. Secondly, a sub-score for user engagement is calculated based on user attention duration, reflecting the user's level of attention to the function's operation. Finally, a sub-score for explicit interest is calculated based on the frequency of active interaction, reflecting the user's willingness to actively explore. The system then performs a weighted sum of these three normalized sub-scores (e.g., giving explicit interest a higher weight) to arrive at a precise value that directly represents the user's true dependence on and liking of the function.

[0112] In step S630, each intelligent driving function is sorted according to its adaptability score, and the target recommended intelligent driving function set is selected from the sorted results in combination with the hardware configuration constraints of the target vehicle.

[0113] After obtaining the compatibility scores of all functions, the system first sorts them in descending order to identify the top 5 functions that the user is most interested in. Then, because vehicle functions are often strongly tied to hardware (sensors, computing platforms), the system introduces crucial hardware configuration constraint verification logic. Specifically, it queries the vehicle's Bill of Materials (BOM) to check for hardware dependency conflicts or mutual exclusions among the top-scoring functions. For example, if the user scores "Advanced Urban Driving Assistance" extremely highly, the system will confirm that this function requires "LiDAR" and "dual Orin chips." Thus, the system eliminates functions that cannot be implemented in a single configuration model through a legal combination, ultimately selecting a technically feasible set of recommended intelligent driving functions that best matches the user's preferences.

[0114] In step S640, based on the preset function configuration package mapping strategy, the target recommended intelligent driving function set is matched with the corresponding vehicle optional configuration scheme, and a personalized test drive analysis report is generated accordingly.

[0115] For example, based on the compatibility scores of each function and the feasibility of the vehicle configuration, the system packages and recommends personalized intelligent driving option packages. For users identified as "urban commuters," if their test drive data shows high coverage of "congested road sections" and positive feedback on "automatic start-stop smoothness" (e.g., frequent inquiries), the system's recommended solution may include a combination of "Noise Control (NOA)," "Traffic Jam Assist (TJA)," and "Automatic Parking Assist (APA)." For users of the "highway intercity travel" type, if their data shows a positive focus on "large vehicle avoidance" and "curving speed control" scenarios, the recommended solution may emphasize the option combination of "highway navigation-assisted driving," "intelligent headlights," and "side and rear cross-traffic alert." This data-driven recommendation logic significantly enhances the persuasiveness of sales recommendations.

[0116] Specifically, if the recommended package includes "seat massage" and "premium audio system," the system automatically matches the "Comfort Package" that includes both of these features; if it includes "City NOA," it matches the "Intelligent Driving High-Spec Model" or the "Premium Intelligent Driving Subscription Package." Based on this, the system automatically generates a personalized test drive analysis report in H5 or PDF format. This report not only lists the recommended models and optional packages but also displays data highlights from the test drive process in chart form (such as "During this test drive, the intelligent driving system took over 80% of the journey for you, saving 30% of your energy"). This data-driven approach supports the purchase recommendation, thereby improving the success rate of sales conversion.

[0117] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of combined actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Secondly, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application. In the above embodiments, the descriptions of each embodiment have their own emphasis; for parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0118] In some embodiments, this application also provides a computer program product, the computer program product including a computer program stored on a non-volatile computer-readable storage medium, the computer program including program instructions, which, when executed by a computer, cause the computer to execute any of the above-described intelligent vehicle test drive sales interaction methods.

[0119] In some embodiments, this application also provides an electronic device, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform an intelligent vehicle test drive sales interaction method.

[0120] The apparatus described in the embodiments of this application can be used to execute the intelligent vehicle test drive sales interaction method of the embodiments of this application, and accordingly achieve the technical effects achieved by the intelligent vehicle test drive sales interaction method of the embodiments of this application, which will not be elaborated here. In the embodiments of this application, the relevant functional modules can be implemented by a hardware processor.

[0121] Figure 7 This is a schematic diagram of the hardware structure of an electronic device for executing a smart vehicle test drive sales interaction method according to another embodiment of this application, as shown below. Figure 7 As shown, the device includes: One or more processors 710 and memory 720, Figure 7 Take the 710 processor as an example.

[0122] The device for implementing the intelligent vehicle test drive sales interaction method may also include: an input device 730 and an output device 740.

[0123] The processor 710, memory 720, input device 730, and output device 740 can be connected via a bus or other means. Figure 7 Taking the example of a connection between China and Israel via a bus.

[0124] The memory 720, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules, such as the program instructions / modules corresponding to the intelligent vehicle test drive sales interaction method in the embodiments of this application. The processor 710 executes various functional applications and data processing of the server by running the non-volatile software programs, instructions, and modules stored in the memory 720, thereby implementing the intelligent vehicle test drive sales interaction method in the above-described method embodiments.

[0125] The memory 720 may include a program storage area and a data storage area, wherein the program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the device, etc. Furthermore, the memory 720 may include high-speed random access memory and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory 720 may optionally include memory remotely located relative to the processor 710, and these remote memories may be connected to the device via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0126] Input device 730 can receive input digital or character information and generate signals related to user settings and function control of the device. Output device 740 may include display devices such as a display screen.

[0127] The one or more modules are stored in the memory 720, and when executed by the one or more processors 710, they execute the intelligent vehicle test drive sales interaction method in any of the above method embodiments.

[0128] The above-described product can perform the methods provided in the embodiments of this application, and has the corresponding functional modules and beneficial effects for performing the methods. Technical details not described in detail in this embodiment can be found in the methods provided in the embodiments of this application.

[0129] The electronic devices in this application embodiments exist in various forms, including but not limited to: (1) Mobile communication devices: These devices are characterized by their mobile communication capabilities and primarily aim to provide voice and data communication. These terminals include: smartphones (e.g., iPhones), multimedia phones, feature phones, and low-end phones, etc.

[0130] (2) Ultra-mobile personal computer devices: These devices fall under the category of personal computers, possessing computing and processing capabilities, and generally also have mobile internet access features. These terminals include PDAs, MIDs, and UMPCs, such as the iPad.

[0131] (3) Portable entertainment devices: These devices can display and play multimedia content. This category includes audio and video players (such as iPods), handheld game consoles, e-book readers, as well as smart toys and portable car navigation devices.

[0132] (4) Server: A device that provides computing services. The components of a server include a processor, hard disk, memory, system bus, etc. Servers are similar to general computer architectures, but because they need to provide highly reliable services, they have higher requirements in terms of processing power, stability, reliability, security, scalability, and manageability.

[0133] (5) Other electronic devices with data interaction functions.

[0134] In some embodiments, this application also provides a mobile platform on which the computer device described in any embodiment of this application is installed. The mobile platform includes, but is not limited to, vehicles, tracked robots, bipedal robots, quadrupedal robots, etc., wherein the vehicle can be a passenger car, pickup truck, truck, etc. It should be noted that the above are merely examples, and this application does not limit the specific form of the mobile platform.

[0135] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0136] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented using software plus a general-purpose hardware platform, or of course, using hardware. Based on this understanding, the above technical solutions, in essence or the parts that contribute to the related technology, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0137] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A smart vehicle test drive sales interaction method, comprising: In response to a user's interactive trigger operation on a target vehicle, the user's natural language input data is acquired, wherein the natural language input data includes voice input data and / or text input data; Semantic parsing is performed on the natural language input data to generate a set of user demand tags that represent the user's car purchase intention and preference dimensions; Based on the user demand tag set, the set of target vehicle functions to be displayed is filtered from the set of displayable functions of the target vehicle, and the set of target experience scenarios that match the set of target vehicle functions is determined according to the preset function scenario mapping relationship; wherein, the function scenario mapping relationship defines the association between vehicle function identifiers and experience scenario types or experience scenario road segment identifiers. The current location of the target vehicle is obtained as the starting point of the route. Road network data is determined based on map data, and the scene road segment corresponding to the target experience scene set is located in the road network data. Under the premise of meeting the preset route constraints, a personalized test drive route that can cover the target experience scene set is planned and generated.

2. The method according to claim 1, wherein, When the natural language input data includes voice input data, the step of acquiring the user's natural language input data in response to the user's interactive trigger operation on the target vehicle includes: Acquire vehicle status data collected by a multimodal sensor array deployed in the cockpit of the target vehicle; When the vehicle status data meets the preset welcoming interaction conditions, it is determined that a valid interactive trigger operation has been detected; the welcoming interaction conditions include one or more of the following: the pressure value detected by the gravity sensor of the driver's seat is within the preset adult weight range, the in-vehicle camera recognizes a valid face feature of a non-staff member who is looking forward, and receives a door closing signal sent by the body controller. In response to the effective interactive trigger operation, the in-vehicle intelligent assistant's proactive greeting process is activated, and the microphone array is started to collect the user's voice input data during the greeting dialogue.

3. The method according to claim 1, wherein, The step of semantically parsing the natural language input data to generate a set of user demand tags representing the user's car purchase intention and preference dimensions includes: Perform intent recognition on the natural language input data to determine the target intent category to which the user's car purchase intent belongs; Key information is extracted from the natural language input data, and key information slot values ​​are extracted from at least one information dimension; the information dimension includes any one of the following: budget range, car usage scenario, family structure, power preference and intelligent driving focus; The target intent category is mapped to a label of intent dimension, and the key information slot value is mapped to a label of preference dimension. The labels of intent dimension and preference dimension are aggregated to generate the user demand label set. A temporary identity is created for the current user, and the user requirement tag set is bound to the temporary identity to establish user profile data that supports synchronization across vehicle terminals.

4. The method according to claim 1, wherein, The process of planning and generating a personalized test drive route that covers the target experience scenario set, under the premise of meeting preset route constraints, includes: A weighted road network search map is constructed based on the road network data, and the scene road segments that have been located in the road network data are marked as reward road segments in the weighted road network search map; Obtain preset route constraints, including a total test drive duration threshold and / or a total test drive mileage threshold. The path search algorithm searches for the optimal path in the weighted road network search graph. The path search algorithm is configured with a scene reward function. When the search path passes through the reward segment, the cumulative reward value of the path is increased by the scene reward function. The path that maximizes the cumulative reward value while satisfying the preset route constraints is output as the personalized test drive route.

5. The method according to claim 1, wherein, During the process of the target vehicle driving according to the personalized test drive route, the method further includes: When it is determined that the target experience scenario set contains a specific demonstration scenario that is not suitable for triggering under real road conditions, the virtual simulation mode is started on the vehicle display terminal; A virtual road environment is constructed based on a graphics rendering engine. Virtual target objects are injected into the virtual road environment, and the virtual vehicle model is driven according to the real-time motion state data of the target vehicle to dynamically demonstrate the perception and decision-making process of the intelligent driving system for the virtual target object.

6. The method according to claim 1, wherein, During the process of the target vehicle driving according to the personalized test drive route, the method further includes: The system collects facial images, eye movement patterns, and body posture data from the user through sensors. The system identifies discrete emotion categories in the facial images, calculates the user's gaze hotspots on the in-vehicle display interface and outside the vehicle based on the eye movement trajectory, and analyzes the user's posture features based on the body posture data. The discrete emotion categories, the gaze hotspot regions, and the posture features are weighted and fused to quantify and generate user psychological state features, which include instantaneous emotion vectors and / or comprehensive immersion indexes.

7. The method according to claim 6, wherein, After weightedly fusing the discrete emotion categories, the gaze hotspot regions, and the posture features to quantify and generate user psychological state features, the method further includes: A comprehensive analysis of the user's psychological state characteristics is performed to determine the user's current psychological state category; When the current psychological state category is a confused state, the vehicle display interface is controlled to present a detailed schematic diagram of the current operating function. When the current psychological state category is a state of tension and the vehicle speed meets the preset high-speed conditions, the system outputs fine-tuning suggestions for vehicle control parameters via voice. These fine-tuning suggestions include reducing the cruising speed or increasing the following distance.

8. The method according to claim 6, wherein, In the process of quantifying and generating the user's psychological state characteristics, the method further includes: If a semantic conflict is detected between the explicit semantics expressed by the user's voice input data and the implicit emotional tendency expressed by the discrete emotion category or the posture feature, then a first priority confidence weight is assigned to the voice input data, a second priority confidence weight is assigned to the discrete emotion category or the posture feature whose duration exceeds a preset time threshold, and a third priority confidence weight is assigned to the discrete emotion category or the posture feature that is presented instantaneously. The user's psychological state characteristics are recalculated based on the weighted multimodal data to prioritize responses to the user's explicit intentions.

9. The method according to claim 7, wherein, The state of tension is triggered based on a threshold value for the intensity of the state of tension. After providing fine-tuning suggestions for vehicle control parameters via voice output, the method further includes: In response to a user confirmation command regarding the fine-tuning suggestion, fine-tuning of the vehicle control parameters is performed within a preset vehicle safety boundary range; During the preset monitoring window period after the fine-tuning, the changing trend of the user's psychological state characteristics is continuously monitored; If the negative emotion score in the user's psychological state characteristics shows a downward trend and the overall immersion index shows an upward trend, then the current interaction strategy is deemed effective and a positive feedback signal is generated. Based on the positive feedback signal, the stress intensity threshold is reduced within the current session period.

10. The method according to claim 1, wherein, After the target vehicle completes the personalized test drive route, the method further includes: Obtain behavior logs for the entire test drive process. The behavior logs record the scene coverage depth, user attention duration, and frequency of active interactions for each scenario in the target experience scenario set. A multi-factor evaluation is performed based on the behavior logs to calculate the compatibility score between the target vehicle's various intelligent driving functions and the current user. The compatibility score is obtained by weighted summation of scene relevance sub-score, user engagement sub-score, and explicit interest sub-score. The scene relevance sub-score is calculated based on the scene coverage depth, the user engagement sub-score is calculated based on the user's attention dwell time, and the explicit interest sub-score is calculated based on the frequency of active interactions. The intelligent driving functions are sorted according to the adaptation score, and the target recommended intelligent driving function set is selected from the sorted results in combination with the hardware configuration constraints of the target vehicle. Based on a preset function configuration package mapping strategy, the target recommended intelligent driving function set is matched with the corresponding vehicle optional configuration scheme, and a personalized test drive analysis report is generated accordingly.

11. A computer device, comprising a memory, a processor, and a computer program stored in the memory, wherein, The processor executes the computer program to implement the steps of the method according to any one of claims 1-10.

Images