Cabin conference scene active pushing method, electronic device and vehicle

Abstract

The application relates to the technical field of intelligent automobiles, in particular to a cockpit conference scene active pushing method, an electronic device and a vehicle. The method part comprises the following steps: acquiring multi-modal sensing data corresponding to a user in a cockpit of a vehicle; generating scene scheme information corresponding to a conference transition scene on the basis of the multi-modal sensing data, wherein the conference transition scene comprises at least one of a pre-conference scene before a preset conference scene is started and a post-conference scene after the preset conference scene is ended; when being in the conference transition scene, pushing the scene scheme information corresponding to the conference transition scene, comprising: if currently being in the pre-conference scene, pushing reminding information containing cockpit environment pre-adjustment options on the basis of schedule information, the cockpit environment pre-adjustment options comprising at least one of air conditioner setting and office light and shadow atmosphere adjustment; if currently being in the post-conference scene and the continuous time length of the preset conference scene satisfies a preset time length threshold, pushing a rest scheme containing a rest guide option. The method is used for solving the problem that a user cannot be provided with smooth and efficient intelligent experience.

Description

Technical Field

[0001] This application relates to the field of intelligent vehicle technology, and in particular to a method for proactively pushing information in a cockpit conference scenario, as well as related electronics and vehicles. Background Technology

[0002] With the rapid development of intelligent connected vehicles, smart cockpits have gradually evolved and taken on the attributes of a third living space for users. For business travelers, conducting online voice or video conferences in the car has become a frequent daily application scenario.

[0003] Currently, while some smart cockpit systems are equipped with relevant conferencing functions, they typically employ a mechanical command-trigger-response passive execution mode. Specifically, the system heavily relies on users manually clicking on the vehicle's screen or issuing voice commands at the start of the meeting to adjust the cockpit environment individually or collectively (such as closing windows, turning down the volume, etc.).

[0004] However, this traditional control method has significant limitations. Existing systems often focus only on the single time slice of a meeting in progress. When faced with high-intensity mobile work, users need to manually adjust various hardware resources to create the environment before starting work, and then manually intervene again after the meeting to obtain space for rest and relaxation. It is evident that traditional environmental control solutions rely excessively on immediate human intervention and cumbersome operations, resulting in low overall scenario efficiency and failing to provide users with a smooth, efficient, and intelligent experience. Summary of the Invention

[0005] This application provides a method for proactively pushing information in a cockpit conference scenario, an electronic device, and a vehicle to solve the technical problem that traditional smart cockpits cannot provide smooth and continuous intelligent services.

[0006] To solve the above-mentioned technical problems, this application provides the following technical solution: A method for proactively pushing content in a cockpit conference scenario includes: Acquire multimodal sensor data corresponding to users in the vehicle cabin; Based on the multimodal sensing data, scenario scheme information corresponding to the meeting transition scenario is generated, wherein the meeting transition scenario includes at least one of the pre-meeting scenario before the preset meeting scenario starts and the post-meeting scenario after the meeting ends. When in the meeting transition scenario, push the corresponding scenario plan information, including: if currently in the pre-meeting scenario, push the reminder information containing cabin environment pre-adjustment options based on the schedule information, the cabin environment pre-adjustment options including at least one of air conditioning settings and office lighting atmosphere adjustment; if currently in the post-meeting scenario, and the duration of the preset meeting scenario is detected to meet the preset duration threshold, push the rest plan containing rest guidance options. An electronic device, comprising: processor; And memory, used to store computer programs; The computer program, when executed by the processor, implements the method described therein.

[0007] A vehicle including the aforementioned electronic equipment.

[0008] One of the solutions provided in this application acquires multimodal sensor data and focuses service intervention nodes on two specific meeting transition scenarios: the pre-meeting scenario before the meeting starts and the post-meeting scenario after the meeting ends. Upon identifying a transition scenario, the solution proactively pushes the corresponding scenario plan, breaking the limitation of existing technologies that can only passively respond to a single, fixed state during a meeting. This fills the gap in system perception and service before and after meetings. Relying on multimodal sensor data, it achieves advance decoding and personalized solution generation for pre-meeting preparation and post-meeting relaxation needs. It effectively compensates for the operational gap caused by the lack of predictability in the system, which necessitates cumbersome manual adjustments to various cabin hardware. It realizes an intelligent leap from passively waiting for instructions after the meeting to proactively anticipating and pre-processing specific transition periods, providing business professionals with a seamless experience throughout the entire meeting cycle and achieving smooth, continuous intelligent services in smart cabin meeting scenarios. Attached Figure Description

[0009] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 This is a schematic diagram of the processing framework of an active push method for a cockpit meeting scenario according to an embodiment of this application; Figure 2 This is a flowchart illustrating an active push method for a cockpit meeting scenario according to one embodiment of this application; Figure 3This is a schematic diagram of the meeting mode triggering interface of a preset meeting scene in a cockpit meeting scene active push method according to an embodiment of this application; Figure 4 This is a schematic diagram of the meeting mode triggering and opening interface of a preset meeting scenario in a cockpit meeting scenario active push method according to an embodiment of this application; Figure 5 This is a schematic diagram of a system structure for an active push system for a cockpit meeting scenario according to one embodiment of this application; Figure 6 This is a schematic diagram of the structure of an electronic device according to one embodiment of this application. Detailed Implementation

[0011] To make the technical problems, technical solutions, and beneficial effects solved by this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0012] In existing technical solutions, there are some preset meeting modes for business office scenarios. The core logic of these modes usually relies on manual clicks or voice commands from the user inside the meeting room to be passively triggered. While this trigger-response mode can provide basic environmental adjustment functions, it often only adjusts the system during the meeting, resulting in a lack of service predictability during the meeting preparation period and the transition phase after the meeting ends.

[0013] In actual business travel, users often need to manually adjust meeting settings such as seats, air conditioning, and privacy before the meeting officially begins, or find ways to relax after a long, high-pressure meeting. This fragmented interaction not only increases the user's operational burden, but also makes it difficult to meet the needs of business people for an efficient and immersive experience.

[0014] Based on this, this application provides a proactive push solution for cockpit meeting scenarios, which aims to achieve a service transformation from passive response to proactive anticipation by proactively perceiving and predicting meeting transition scenarios, thereby realizing smooth and continuous intelligent services for intelligent cockpit meeting scenarios.

[0015] To more clearly illustrate the technical solution of this application, before describing the specific method embodiments in detail, the core control logic of the active push method for the intelligent cockpit conference scenario provided by this application will be given an overall overview in conjunction with the overall architecture diagram.

[0016] like Figure 1 As shown, the intelligent cockpit proactive push and learning architecture provided in this application mainly includes two core functional loops: personalized learning iteration for preset meeting scenarios (corresponding to...). Figure 1 (showing the upper part) and proactive push notifications for personalized scenarios (corresponding to) Figure 1 (See the lower half).

[0017] The system has pre-built standardized preset meeting modes. When a user triggers and opens this mode via voice, touch, or other means, the system accurately records the user's personalized operations on various underlying hardware components of the cabin during the meeting (such as fine-tuning the air conditioning fan speed and turning ambient lighting on and off). This implicit interaction data is aggregated in real-time to the vehicle-human-environment data center, analyzed, refined, and precipitated as unique user behavioral habits, which are then stored in a dynamic memory system. The underlying personalized model uses this habit data for reinforcement feedback learning, adjusting and updating the default function configurations of the preset meeting modes accordingly. This mechanism allows the vehicle to continuously self-correct as the user interacts with it, becoming increasingly adept at understanding the basic meeting scenarios with each use.

[0018] The system proactively pushes personalized scenarios, particularly addressing service anticipation issues during critical transition phases such as before, during, and after meetings. It doesn't limit itself to fixed preset modes but comprehensively records users' actual actions and states in pre-, during, and post-meeting scenarios. This data, combined with a human-vehicle-environment data center and memory system, is used to accumulate and learn user habits. Based on this multimodal data, the personalized model can deeply characterize the user's potential intentions and needs at specific meeting junctures. When the system detects a user in a similar pre- or post-meeting transition scenario, the model proactively generates a personalized scenario plan that fits that state (such as a pre-meeting focused preparation mode or a post-meeting rest mode) and sends it directly to the user.

[0019] In summary, this application, through the aforementioned dual-layer closed-loop architecture, not only solidifies the self-iteration capability of scenario parameters during the meeting, but also extends the service reach to the transition zone before and after the meeting, completely changing the traditional passive mechanical interaction logic of the cockpit and realizing proactive anticipation and seamless connection of scenario services.

[0020] The following describes the specific implementation examples.

[0021] In one embodiment, such as Figure 2 As shown, a method for proactively pushing information in intelligent cockpit meeting scenarios is provided for use in an intelligent cockpit meeting scenario proactive push system. The method includes the following steps: S101. Acquire multimodal sensor data corresponding to the user in the vehicle cabin; S102. Generate scenario scheme information corresponding to the conference transition scenario based on multimodal sensing data, wherein the conference transition scenario includes at least one of the pre-conference scenario before the preset conference scenario starts and the post-conference scenario after the conference ends. S103. When in a meeting transition scenario, push the corresponding scenario plan information for the meeting transition scenario, including: if currently in the pre-meeting scenario, push reminder information containing cabin environment pre-adjustment options based on schedule information, the cabin environment pre-adjustment options including at least one of air conditioning settings and office lighting and atmosphere adjustment; if currently in the post-meeting scenario, and the duration of the preset meeting scenario is detected to meet the preset duration threshold, push a rest plan containing rest guidance options.

[0022] Specifically, this embodiment is applied to the intelligent cockpit environment control and scenario-based service recommendation scenarios in the field of intelligent connected vehicle technology, especially for business travel scenarios with high requirements for in-cabin privacy, work efficiency, and rest quality. In traditional intelligent cockpit services, only fixed meeting scenarios are usually preset, lacking proactive perception and response to the potential state of users before and after the meeting.

[0023] To overcome the limitations of this passive response, step S101 is first executed to acquire the user's corresponding multimodal sensor data. Multimodal sensor data, at the physical and data logic levels, refers to a collection of multidimensional information reflecting the cockpit environment. The system primarily relies on the vehicle-human-environment data center to acquire multimodal sensor data in real time.

[0024] As a concrete example, this multimodal sensing data can include visual data, voice data, environmental and travel data, vehicle signal data, and user data. Visual data is the user's facial expressions (such as fatigue or focus) and body postures (such as sitting upright or leaning back in relaxation) accurately captured and identified by the in-cabin camera; voice data consists of voice features collected and analyzed through the cockpit microphone array, which includes not only semantic keywords of the call content (such as identifying business terms like meetings and contracts), but also emotional features and stress level indicators extracted based on speech rate and tone; environmental and travel data covers the vehicle's current geographical location, navigation destination, current time, and, more importantly, calendar event information synchronized in real time from the user's linked electronic devices (e.g., 15:00-16:00, project review meeting with Company A); vehicle signal data can include actuator feedback parameters such as the current seat angle, air conditioning temperature, driving mode, and multimedia volume; and user data integrates the user's basic identity attributes, travel frequency, and historical interaction data within the vehicle. By collecting all of the aforementioned multi-dimensional data, the Human-Vehicle-Environment Data Center has achieved a high-precision digital characterization of the user's current physical environment and subjective intentions, providing a very solid and rich data foundation for subsequent proactive scene recommendation processing.

[0025] After acquiring multimodal sensor data, the system proceeds to step S102, generating scenario plan information corresponding to the meeting transition scenarios based on the multimodal sensor data. Here, the meeting transition scenarios are buffer and preparation intervals specifically defined by the system on the timeline for standard meeting events. Specifically, the system defines a preset duration before the formal start of the meeting (e.g., 5 minutes before the meeting mode is turned on) as the pre-meeting scenario, aiming to provide environmental pre-adjustment and emotional preparation for the upcoming high-intensity work; simultaneously, it defines another preset duration after the end of the meeting (e.g., 5 minutes after the meeting mode is turned off) as the post-meeting scenario, aiming to provide users with an opportunity for relaxation and recovery after intense mental work. By locking in these two meeting transition scenarios, the system is no longer limited to a single state during a meeting. When generating scenario plan information, the system analyzes the user's scenario plan information within these two specific time windows, ensuring it aligns with the current context. On one hand, at the human-computer interaction level, the scenario plan information can be displayed content and prompts pushed to the user for confirmation. For example, the system generates visual scene cards based on the Scene Master protocol and displays them as pop-ups on the screen (such as the second-row ceiling-mounted screen), or proactively asks the user in conjunction with TTS voice broadcasts (such as broadcasting "Do you want to switch to meeting mode in advance?" or popping up a message asking "Do you need to rest for 15 minutes?"). On the other hand, at the underlying control execution level, it can represent a set of multi-dimensional cockpit hardware collaborative control instructions generated by a large model decision. Once the user clicks to confirm (or the system executes automatically when specific conditions are met), this scheme is immediately transformed into specific control instructions issued to the cockpit domain controller to directly drive and coordinate the adjustment of multiple hardware actuators such as air conditioning, seats, windows, and lights, thereby physically reconstructing the cockpit environment. In short, the pushed scene scheme can be either a scene recommendation card / voice for the user to select and confirm, or a package of underlying hardware control instructions ready to be executed at any time.

[0026] As scenario plan information is generated, the system executes step S103, pushing the corresponding scenario plan information when in a meeting transition scenario. Specifically, for the pre-meeting scenario, since switching from a normal car ride to a high-intensity focused work state requires an environmental pre-adjustment and mental preparation process, the system monitors the time difference between the current time and the schedule information synchronized from the external terminal (such as the meeting start time) in real time. When this time difference shortens to a preset range (e.g., 5 minutes before the meeting starts), the system determines that it has officially entered the pre-meeting scenario. At this time, the system does not directly and forcibly change the cabin state, but pushes a reminder message to the user containing cabin environment pre-adjustment options. At the physical output level, this reminder message can be gently read aloud through the cabin text-to-speech module (e.g., "A meeting is in 5 minutes, you can adjust to meeting mode in advance to prepare for the meeting"), and simultaneously pops up a visual card on the vehicle's desktop or the second-row ceiling screen. The cabin environment pre-adjustment options represent a series of underlying hardware control commands to pre-warm up the work environment. As a specific example, this includes not only air conditioning settings (such as adjusting the second-row zone air conditioning to avoid direct airflow onto people and setting the fan speed to level 1 natural wind mode to reduce wind noise), but also adjusting the office lighting atmosphere (such as closing the front and rear sunshades, adjusting the light transmittance of each window to the lowest level to block external visual interference, and adjusting the second-row ambient lighting to a light blue tone to promote concentration). Furthermore, when a meeting is about to begin, the system can also work with the noise reduction module to block irrelevant notifications and adjust the seat to a semi-formal support posture. Specific combinations of these functions are not strictly limited here.

[0027] On the other hand, regarding the post-meeting scenario, considering that high-intensity online or offline meetings often lead to mental exhaustion, the system has set up a conditional trigger mechanism based on meeting intensity. When the meeting ends and the system enters the post-meeting scenario, it first extracts and calculates the duration of the preceding preset meeting scenario. Subsequently, the system compares this duration with a preset duration threshold set within the system (e.g., 60 minutes). Short, brief communications usually do not cause deep fatigue, and frequent push notifications for rest can be disruptive; however, when the duration exceeds the preset threshold, it indicates that the user has just experienced a long, energy-intensive, in-depth meeting, and is highly likely to be in a state of high-pressure fatigue. Under this condition, the system proactively pushes a short rest plan with rest guidance options to the user. As a specific implementation method, after a long meeting ends, the rear ceiling-mounted screen will automatically pop up an inquiry window, proactively asking the user if they need to rest for a certain period of time (e.g., a 10-15 minute rest countdown). If the user confirms, the rest program will directly drive the underlying actuators to activate the seat's soothing massage function, dim or turn off the second-row reading lights, and play natural white noise through the multimedia system, thus creating a dimly lit, quiet, and comfortable space for the user to recharge during limited travel time.

[0028] In this embodiment, by acquiring multimodal sensor data and focusing service intervention nodes on two specific meeting transition scenarios—pre-meeting scenarios before the meeting begins and post-meeting scenarios after the meeting ends—the system proactively pushes corresponding scenario solutions when these transition scenarios are identified. This fills the gap in system perception and services before and after the meeting. Relying on multimodal sensor data, it enables the early decoding and personalized solution generation of pre-meeting office preparation and post-meeting relaxation needs. This effectively compensates for the operational gap caused by the lack of predictability in the system, which necessitates cumbersome manual adjustment of various cabin hardware. It achieves an intelligent leap from passively waiting for instructions after the meeting to proactively anticipating and pre-processing specific transition periods, creating a seamless experience for business professionals throughout the entire meeting cycle and realizing smooth and continuous intelligent services in smart cabin meeting scenarios.

[0029] Furthermore, by sending pre-meeting reminders based on the schedule, including those for adjusting air conditioning and lighting, and by conditionally pushing short break options based on meeting duration thresholds after the meeting, the system leverages the mental workload curve during business work to pre-meeting the physical environment to aid focus and provides targeted physical and mental recovery after meetings by accurately identifying high-energy-consuming sessions. This significantly improves user efficiency and fatigue relief during complex business trips.

[0030] It should be noted that in this embodiment, the preset cockpit meeting mode is used in a preset meeting scenario. Based on the vehicle's SOA service capabilities, a meeting mode is pre-set in the cockpit to provide second-row users with basic meeting privacy space and to provide a basis for iterative iteration of user personalized habits. The cockpit meeting mode is as follows: The trigger conditions for this meeting mode can be click-triggered, voice-triggered, or condition-triggered.

[0031] The second row of ceiling-mounted screens can be displayed as follows: Figure 3 As shown. The meeting scene mode is prominently displayed on the vehicle's infotainment system desktop. Second-row passengers can manually access the meeting mode by clicking on the desktop (e.g., ...). Figure 3 As shown), the preset vehicle control settings in the cockpit can be opened. Users can also trigger the "meeting mode" scenario automatically, opening the preset vehicle control settings in the cockpit, when the large model on the device recognizes similar semantics such as "Help me open meeting mode," "Open meeting mode," or "Start meeting mode." For example... Figure 4As shown, the condition trigger can be a schedule trigger. When the system detects that a user has opened a meeting-related app, it automatically reminds the user whether to activate the meeting scenario mode. The user can activate / deactivate the mode manually or via voice. As an example, mobile meeting schedule information could be synchronized to the vehicle's schedule app, automatically recognizing the online meeting time based on the schedule. When the vehicle's infotainment system starts (driving / parked status), it automatically reminds the user of their schedule 10 minutes before the online meeting begins, and automatically joins the online meeting 10 minutes later, while the cabin switches to meeting mode.

[0032] In one embodiment, the multimodal sensing data includes historical interaction data of users in historical meeting transition scenarios. Generating scenario solution information corresponding to the meeting transition scenario based on the multimodal sensing data includes the following steps: S201. Analyze historical interaction data based on a preset habit matrix to extract user habit features that reflect user operating habits. The habit matrix includes the user's habitual operating parameters for cabin environment, cabin privacy and cabin audio in the historical meeting transition scenario. S202. Based on the user habit characteristics, generate the scenario solution information corresponding to the meeting transition scenario.

[0033] Specifically, this embodiment further defines the logic for generating scenario solution information based on the above embodiments. During the long-term use of the smart cockpit, different business professionals often have highly personalized and preferred methods for alleviating the high pressure or fatigue before and after meetings.

[0034] To this end, this embodiment first introduces a historical dimension assessment. By executing step S201, historical interaction data is analyzed based on a preset habit matrix to extract user habit features reflecting user operating habits. Here, historical interaction data is essentially the full amount of embedded data collected by the system in past pre-meeting or post-meeting scenarios, such as user records of manually adjusting seat angles, turning on specific music, or adjusting specific temperatures. The preset habit matrix is ​​a multi-dimensional mathematical statistics and mapping model used to quantitatively analyze the correlation between the frequency of operation of specific scenarios (i.e., meeting transition scenarios) and specific cockpit function items. It should be noted that, specifically, the system or cockpit big model transforms long-term historical interaction data into a structured user habit matrix and performs feature extraction. The system does not blindly process data from all day, but uses the start and end times of each formal meeting in the historical itinerary as clear time anchors, strictly capturing all underlying vehicle control signals within specific time windows before and after these anchors. These originally discrete physical interactions, such as pressing the air conditioning button or issuing a voice command to lower the window, are uniformly extracted and encoded into multi-dimensional feature vectors containing time deviation, initial environmental baseline, and specific hardware execution parameters. These feature vectors are then batch-input into a habit matrix constructed within the large-scale cockpit model. This habit matrix is ​​a dynamically updated high-dimensional probability space. The model uses an unsupervised density clustering algorithm to calculate distances between massive amounts of data points within a set meeting transition time window, thereby identifying clusters of operations with highly concentrated parameters. For example, it can keenly detect a large number of operation data points pointing to closing the rear sunshade in the dimensional intersection area at a specific time before the meeting when external illumination is strong. Based on this, a time decay weighting mechanism is introduced into the habit matrix. This means that operation samples closer to the current date have a higher confidence weight, allowing the system to calculate the conditional probability expectation of a specific hardware control action in a specific meeting transition scenario. When the probability expectation of a specific operation exceeds a preset threshold, it is determined that the operation has changed from an occasional behavior to a high-frequency habitual operation strongly associated with the user. The system extracts the parameter set of this operation from the habit matrix and writes it into the user's exclusive scene linkage rule library. This allows the system to directly call these rules to automatically combine and proactively push the scene solution that best suits the user's current preferences when the vehicle enters a similar meeting time window again in the future.

[0035] As a concrete example, if the system analyzes the data matrix and finds that the user frequently manually turns on the seat massage function and dims the reading light after multiple meetings, the system can accurately extract the user's habitual characteristics of wanting massage relaxation and dim lighting after the meeting, and then build a multi-dimensional dynamic user memory system that is continuously updated with the interaction.

[0036] After extracting the user's behavioral habits, the system executes step S202, generating corresponding scenario solution information based on user habit characteristics. This process transforms abstract user habit characteristics into a set of specific control instructions that can be recognized by the actuators at the bottom layer of the cockpit. The system does not pre-define or limit the combination range of cockpit functions, but rather, within the specific scenario boundary of meeting transition, it accumulates the user's personalized behavioral patterns. After the human-vehicle environment data center refines and accumulates the collected user data into the memory system, the originally independent high-frequency operations (such as simply lowering the air conditioning fan speed, closing the windows, adjusting the seat back, or switching multimedia audio sources) are not directly issued as loose single-point instructions. Instead, based on the user's condition before or after the meeting, these scattered actions are logically categorized and time-sequentially aligned. For example, actions such as lowering the fan speed, closing the windows, and brightening the reading lights are extracted and clustered into parallel control instructions under the theme of focused preparation; while actions such as reclining the seat, turning on the back massage, and playing soothing white noise are clustered into serial or parallel instructions under the theme of energy recovery. After completing logical clustering, the system performs physical interference checks and execution sequence orchestration on these combinations (e.g., ensuring that the seat back tilt command is executed before the leg rest lift command to prevent spatial interference). Finally, the system packages this set of time-sequentially orchestrated cross-domain hardware control messages along with the UI materials required for front-end interaction (such as automatically generated scene card text and icons) into a highly cohesive, one-click-triggered personalized scene solution. This completely transforms the fragmented control actions that previously relied on users clicking through different menus one by one into an automated scene service with a specific business theme and high hardware-software linkage. The next time a user is in a pre- or post-meeting scenario, the system can directly invoke and proactively push this scene based on the user's unique memory.

[0037] In this embodiment, user habit characteristics are extracted by analyzing historical interaction data of users in meeting transition scenarios based on a habit matrix, and personalized scenario solution information is generated accordingly. This endows the intelligent cockpit with dynamic memory and continuous learning capabilities, transforming the implicit feedback of each manual adjustment by the user into data nutrients for optimizing recommendation strategies and refining unique behavioral patterns. This breaks through the limitations of traditional intelligent cockpits that are fixed at the factory and have rigid rules, constructing a continuously evolving intelligence that becomes more intelligent with use and constantly evolves. This ensures that the generated scenario solution information matches the deep personalized preferences of different business professionals, effectively avoiding ineffective recommendations and greatly enhancing the system's service accuracy and long-term user stickiness.

[0038] In one embodiment, the multimodal sensing data includes the user's profile information and real-time environmental parameters in the current meeting transition scenario. The real-time environmental parameters include vehicle status information and cabin environment information. Generating corresponding scenario solution information based on the multimodal sensing data includes: S301. Utilize a preset intent prediction model to perform the following steps to generate the scenario solution information: S302. Extract the portrait information, the vehicle status information, and the cabin environment information, and perform feature encoding on each to obtain the corresponding multimodal heterogeneous feature vectors; S303. The profile feature vector representing user preference features in the multimodal heterogeneous feature vector is used as a weighted constraint matrix, and a matrix dot product operation is performed with the vehicle state feature vector and the cabin environment feature vector representing instantaneous environmental state to obtain the aligned cross-modal joint feature vector. S304. The cross-modal joint feature vector is synchronously input into two parallel branch networks to calculate the current state index characterizing the current cabin environment adaptability through the first branch network mapping, and to calculate the potential intention probability value characterizing the user's conference preparation tendency through the second branch network mapping. S305. Perform joint feature matching between the current state indicator and the potential intent probability value and the preset scene in the preset scene library to determine the preset scene information that corresponds to both the current state indicator and the potential intent probability value, and use it as the scene solution information.

[0039] Specifically, those skilled in the art will understand that in complex mobile office scenarios, single-dimensional sensor data is highly susceptible to interference from background noise or transient, sporadic behaviors, leading to data failure. To ensure the fitting accuracy of proactive push notifications, the system introduces multi-dimensional input sources. The profile information is a relatively long-term, multimodal composite data stream. In actual acquisition and aggregation, it combines user facial expressions and body postures captured and recognized by the in-cabin camera, voice emotion features containing specific business keywords (such as "meeting" and "contract") picked up by the microphone array and semantically interpreted by a large model, and user basic preference data such as real-time calendar events, historical travel frequency, and high-frequency destinations synchronized from the user's electronic devices. In contrast, the real-time environmental parameters focus on characterizing the instantaneous physical and spatiotemporal state of the vehicle and its passengers. Vehicle status information includes, but is not limited to, the vehicle's current absolute geographical location, navigation target destination, seat frame angle, and current driving or parking status. Cabin environment information is further refined to include the real-time temperature and airflow of the second-row zone air conditioning, the audio source and volume of the in-vehicle multimedia system, the current color rhythm of the ambient lighting, and the opening and closing status of the windows and sunshades, among other vehicle control signals.

[0040] To eliminate the inherent heterogeneity of the various modal data in terms of physical dimensions, temporal span, and data structure, the system invokes a pre-defined intent prediction model for deep neural network computation. First, the model drives multiple parallel feature extractors (such as convolutional neural networks (CNNs), network encoders, or Transformer layers) to perform feature dimensionality reduction and nonlinear mapping encoding on the extracted static profile information, dynamic vehicle state information, and cabin environment information, thereby transforming fragmented physical signals into standardized multimodal heterogeneous feature vectors under the same mathematical dimension. Since the historical preferences (profile features) of business professionals have a strong guiding and constraining effect on the adjustment of the current instantaneous environment, the system uses the profile feature vector representing user preference features as a weighted constraint matrix, and performs a matrix dot product operation with the vehicle state feature vector and cabin environment feature vector representing the instantaneous environmental state. This matrix operation utilizes the user's long-term personalized behavioral background as an attention filter to perform weighted correction and projection alignment on the dynamic spatiotemporal features of people and vehicles, thereby fusing them into a cross-modal joint feature vector that can deeply represent the strong correlation between people, vehicles, and the environment.

[0041] To improve inference efficiency and enable concurrent multitasking, the model synchronously inputs the cross-modal joint feature vector into two parallel deep branch networks. The first branch network (such as a feature mapping classifier) ​​focuses on analyzing the objective environment. It calculates a current state index representing the adaptability of the current cabin environment through nonlinear mapping, which is used to quantitatively assess the gap between the current physical space inside the vehicle (such as temperature, light, and privacy) and the standard conditions for business work or rest. Meanwhile, the second branch network focuses on predicting subjective psychological tendencies. It calculates the probability value of the potential intention representing the user's tendency to prepare for meetings through probability density mapping, which is used to predict how likely the passenger is to urgently switch to a working state or a relaxing state in the current time slice. Ultimately, these two dynamic assessment indicators, which take into account both objective environment and subjective psychology, are used as multi-dimensional retrieval keys. These keys are then combined with digital features in a local or cloud-based preset scenario library for joint similarity matching. This allows for the precise filtering and identification of specific preset scenario information with the highest confidence level in relation to the current spatiotemporal environment and intent (e.g., matching one of "deep focus mode," "meeting reminder mode," or "travel break energy recovery mode"). This information serves as the final proactively pushed scenario solution information. As mentioned earlier, this scenario solution information can be macroscopically presented as HMI visual scenario cards or TTS voice queries proactively shown to second-row users for interactive selection and confirmation, or as a package of underlying multi-hardware collaborative control instructions encapsulated and ready to be sent to all SOA control nodes throughout the vehicle.

[0042] In this embodiment, a pre-defined intent prediction model, incorporating multimodal feature encoding, matrix dot product constraint alignment, and a dual-track parallel branch network, is introduced into the smart cockpit. This model transforms heterogeneous long-term profiles and short-term instantaneous environmental parameters into joint features, which are then matched and retrieved against a pre-defined scene library. This breaks away from the limitations of traditional in-vehicle systems that rely on linear control triggered by a single mechanical threshold or simple voice commands. By utilizing the attention constraint mechanism of neural networks, it achieves cross-modal deep alignment between static preferences and dynamic instantaneous operating conditions. Furthermore, it constructs a bidirectional complementary intent parsing mechanism from two dimensions: objective cockpit state adaptability and subjective user engagement tendencies. It significantly improves the accuracy and robustness of the intelligent cockpit system in proactively predicting the potential work, privacy, and rest needs of business travelers under changing and complex travel conditions. It effectively avoids the problems of false or missed push notifications caused by missing single-modal data or occasional noise (such as a single unconscious sigh or misclicking a webpage). This ensures that the final generated scenario solution information can be perfectly aligned with the user's real and immediate desires at the automotive grade level, further solidifying the intelligent technology foundation of the cockpit as a highly empathetic and professional AI business assistant.

[0043] In one embodiment, pushing scenario solution information corresponding to a conference transition scenario includes: pushing a scenario card corresponding to the scenario solution information, and executing the scenario solution corresponding to the scenario solution information in response to a user instruction or a preset trigger condition.

[0044] Specifically, this embodiment details how to deliver the generated abstract scenario solution information to the user and complete the final vehicle control loop. After generating personalized scenario solution information at the system's underlying level (e.g., the large cockpit model), an intuitive medium is needed to interact with the user, avoiding any abruptness caused by direct actions of the underlying hardware. To this end, the system, based on a preset scenario encapsulation protocol (such as the Scenario Master Protocol), renders the complex underlying hardware control instruction set into a front-end visual human-machine interface element, namely a scenario card. As a graphical user interface (GUI) asset, this scenario card typically pops up proactively in prominent locations such as the vehicle's central control desktop or the rear ceiling screen in the form of a pop-up window, floating window, or quick panel. The scenario card not only includes the theme name of the scenario solution information (e.g., "Energy Recovery During Travel Breaks"), but also intuitively displays the status of the cockpit functions that are about to be adjusted to the user.

[0045] After the scene card is pushed, the system enters a state of waiting for authorization and execution. The system supports two dimensions of execution triggering mechanisms: the first is responding to user commands. User commands refer to explicit operation signals issued by the user, aiming to return the final control decision to the user. For example, the user can click the open button on the ceiling-mounted screen scene card via physical touch, or issue a voice confirmation command (such as "Okay, open it") via the cockpit microphone array. After the system receives and parses the command, the cockpit domain controller will immediately send the instruction set behind the scene card to each actuator, synchronously adjusting the seats, air conditioning, ambient lighting, and other devices.

[0046] The second type is execution in response to preset trigger conditions. Considering that business professionals may be busy with their hands (e.g., working on a computer to organize documents) or extremely tired and unwilling to speak before or after meetings, the system is configured with automatic execution logic as a fallback. Preset trigger conditions refer to a set of rules that the system's underlying logic determines meet the trigger threshold without requiring explicit user input. As a concrete example, this preset trigger condition could be a do-not-disturb countdown mechanism. For instance, when the system pushes a scene card, it includes a countdown (e.g., 10 seconds). If the user does not cancel before the countdown ends, the system's underlying logic considers this tacit approval and automatically executes the scene plan information. Alternatively, the preset trigger condition could be a multimodal perception composite condition. For example, when the system pushes a scene card for a post-meeting nap, if the visual sensor detects in real time that the user has closed their eyes or tilted their head back, the system directly triggers and executes the scene plan information based on this feedback, without forcibly waking the user for confirmation.

[0047] In this embodiment, by pushing scene cards corresponding to scene solution information and combining user instructions or preset trigger conditions to execute the scene solution information, the system's intent is concretely expressed through graphical cards. The combination of explicit authorization and implicit conditions ensures the flexibility of control flow. This makes the intelligent cockpit's proactive recommendation process neither abruptly depriving users of their control, nor does it provide seamless automated services when users are unable to intervene. Thus, while fully respecting the privacy and autonomy of business people, a smooth, natural, and highly boundary-conscious intelligent interactive experience is achieved.

[0048] In one embodiment, the process of executing a preset meeting scenario further includes: S601. Record the user's proactive adjustments to the target cockpit functions during the execution of a preset meeting scenario; S602. Within multiple consecutive driving cycles, the frequency of operation for the target cockpit function items is statistically analyzed. S603. If the operation frequency meets the preset ratio threshold, the initial parameters of the corresponding function item in the preset meeting scenario are updated using the operating parameters corresponding to the active adjustment operation.

[0049] In practical applications, preset meeting scenarios are typically used as standardized cold starts, with strictly defined triggering conditions and the scope of underlying functions. For example, in the initial factory state, the default settings of a preset meeting scenario might include fixed initial parameters such as turning off the second-row fragrance and setting the second-row zone air conditioning fan speed to level 1 (natural wind). However, different users have vastly different business preferences and comfort perceptions. Therefore, the system executes step S601 to record the user's active adjustments to the target cabin functions during the execution of the preset meeting scenario. These active adjustments refer to secondary interventions performed manually by the user after the system automatically executes the preset scenario, due to discomfort or differences in preference. The target cabin functions encompass the underlying sub-functions involved in the preset meeting scenario, without specific limitations. Examples include actuator functions such as window switches and light transmission, sunshade switches, audio zone modes, reading light switches, ceiling screen switches, driving modes, ambient light switches and modes, fragrance switches and types, and air conditioning switches and modes. The system will capture all of these implicit interactive feedback data.

[0050] Subsequently, to filter out accidental or spontaneous user actions, the system proceeds to steps S602 and S603. Over multiple consecutive driving cycles, the system calculates the frequency of operations on the target cabin function and updates parameters intelligently based on this frequency. Assume the total number of consecutive driving cycles monitored by the system is M1, and the number of times the user actively adjusts a target cabin function within these M1 cycles is N1. The system calculates the operation frequency by the ratio N1 / M1 and compares it to a preset threshold (e.g., set to 90%). If the operation frequency meets the preset threshold, it is determined to be a fixed user habit, not an accidental action. As a specific logical example, when the meeting mode is activated, if the user actively activates the fragrance 5 times in 5 driving cycles (M1=5), the operation frequency related to the fragrance meets the 90% threshold, and the system identifies this adjustment as a user habit. If the air conditioning fan speed is only adjusted 2 times, the operation frequency does not meet the threshold and is considered an accidental action, not a habit. For actions recognized as habitual, the system will use the corresponding operating parameters of the proactive adjustment action (e.g., changing the fragrance status from off to on) to overwrite and update the initial parameters of the corresponding function items in the preset meeting scenario. It is worth noting that this update mechanism only modifies the specific function option parameters within the defined standard scenario function scope, without expanding or shrinking the overall function control boundary of the original scenario.

[0051] For example, specific adjustments can be shown in the table below: In this embodiment, by recording the user's proactive adjustment operations during the execution of a preset meeting scenario, and based on the statistics of operation frequency over multiple driving cycles and verification with a preset proportion threshold, the initial parameters of the scenario are automatically updated. A closed-loop update mechanism using implicit user feedback signals as reinforcement learning samples is constructed. Noise data (i.e., sporadic behavior) is effectively stripped away through mathematical frequency thresholds, extracting the true preference parameters. This enables the meeting mode to continuously learn and evolve with user use, achieving a high-level intelligence that becomes increasingly sophisticated and constantly evolving, significantly reducing the redundancy of repeated similar adjustments by business users and establishing strong long-term user stickiness.

[0052] In one embodiment, for a meeting transition scenario, the method further includes: acquiring schedule information synchronized from an external terminal, the schedule information including meeting start time and meeting end time; determining the trigger point of the pre-meeting scenario based on the meeting start time and a first preset advance time; and determining the trigger point of the post-meeting scenario based on the meeting end time or a monitored meeting process termination signal.

[0053] In traditional smart cockpits, scene switching often relies on users' immediate manual button presses or voice commands, lacking predictability in terms of time. To address this, the system acquires schedule information synchronized from external terminals, including meeting start and end times. These external terminals typically refer to mobile devices such as smartphones and tablets carried by the user; the schedule information refers to the meeting event stream data created by the user in collaborative office software (such as Lark or WeChat Work) on their mobile devices. The system seamlessly synchronizes this schedule information to the vehicle's built-in schedule app via vehicle-cloud collaboration or in-vehicle communication modules, silently resolving the precise planned start and end times for each online or offline meeting in the background. This data synchronization process breaks down the data barriers between the user's personal mobile office ecosystem and the cockpit control ecosystem.

[0054] After successfully acquiring the schedule timeline, the system determines the trigger point for the pre-meeting scenario based on the meeting start time and a first preset advance time. Since the physical adjustments to the cabin environment (such as temperature drop, window closure, ambient lighting change) and the user's psychological preparation both require a buffer period, the system will not abruptly intervene at the absolute zero of the meeting start time. The system sets a first preset advance time, which, as a specific example, can be flexibly set to 5 minutes or 10 minutes. The system subtracts this first preset advance time from the parsed meeting start time to calculate an absolute timestamp, which is the precise trigger point for the pre-meeting scenario. When the current cabin time reaches this trigger point, the system automatically recognizes that the user is about to enter a meeting state, thereby waking up the relevant service modules in advance and executing pre-meeting reminder actions.

[0055] Similarly, to capture the moment a meeting ends and push relevant fatigue relief services, the trigger point for post-meeting scenarios is determined based on the meeting end time or the detected meeting process termination signal. Considering the highly uncertain nature of actual business meetings (they may end on time, or they may end early or significantly exceed the time limit), on the one hand, the system can statically use the originally scheduled meeting end time as a benchmark to determine the trigger point; on the other hand, and more preferably, the system monitors the meeting process termination signal in the cabin or on the synchronous device in real time. This termination signal can be a system-level event of the meeting app being closed, or a network protocol callback instruction indicating that the audio / video call data stream has been disconnected. Once a genuine termination signal is captured, the system accurately defines this moment as the absolute trigger point for post-meeting scenarios and immediately enters the logic for evaluating and pushing post-meeting specific scenario solution information.

[0056] In this embodiment, by acquiring the schedule information synchronized from the external terminal and combining it with a preset advance time and a dynamic meeting process termination signal, the triggering time of the scenarios before and after the meeting is accurately calculated. The user's static digital calendar and dynamic application running status are transformed into an absolute time coordinate axis for cabin environment control. This allows the underlying algorithm model to be precisely activated at the correct business flow node, giving the cabin scenario recommendation system a precognitive capability in the time dimension. This ensures that the cabin environment is ready on time before the meeting and responds instantly afterward, greatly enhancing the timeliness and accuracy of proactive scenario pushes and reducing the risk of users missing meetings or forgetting to adjust the environment.

[0057] In one embodiment, when in a meeting transition scenario, push the corresponding scenario solution information for the meeting transition scenario, including: S701. Obtain the meeting start time of the preset meeting scenario based on the schedule information, and calculate the time deviation between the current time and the meeting start time; S702. If the time deviation meets the preset time threshold condition, then based on the current vehicle power mode or parking status, determine that the current scenario is a pre-meeting scenario and trigger the push of scenario solution information corresponding to the pre-meeting scenario.

[0058] In practical smart cockpit applications, adjustments to the cockpit environment (such as seat movement or dimming of light) can easily lead to serious safety hazards if these adjustments occur under inappropriate driving conditions. Therefore, the system first executes step S701 to obtain the start time of a preset meeting scenario based on schedule information. The system's background process service accurately parses the absolute time node of the planned online or offline meeting, i.e., the meeting start time. Simultaneously, the system outputs the current absolute system time in real time. Then, it calculates the time deviation between the current time and the meeting start time. This time deviation represents the absolute time difference or countdown time remaining until the official start of the meeting.

[0059] After acquiring the dynamically updated time deviation, the system determines whether the time deviation meets the preset time threshold condition. This preset time threshold condition refers to the time window boundary defined by the system for pre-meeting emotional preparation and environmental pre-adjustment. As a specific example, this threshold can be set to 5 minutes. However, meeting the time dimension standard alone is not enough to trigger action. The system then reads the current vehicle physical signals through the vehicle bus (such as a CAN / LIN network) and performs a secondary safety dimension judgment based on the current vehicle power mode or parking status. The power mode refers to the current electrification power supply level of the vehicle, such as OFF (power off), ACC (accessory power on), ON (main circuit power on), or READY (high voltage ready for driving); the parking status refers to the physical static lock feedback of the vehicle chassis and braking system, such as whether the transmission is indeed in P (parking gear) or whether the electronic parking brake (EPB) system is engaged. Only when the time conditions are met and the system detects that the vehicle is in an absolutely safe stationary state (e.g., the system detects that the gear is in P (Park) or the power mode is in a specific parking power supply state that is not in motion), will the system logically lock and determine that the current scenario is the pre-meeting scenario. Once this dual safety barrier based on time and physical state is breached, the system immediately releases the control signal, triggering the push of scenario solution information corresponding to the pre-meeting scenario, such as actively displaying a pre-meeting preparation card containing seat pre-adjustment and window closing options on the vehicle's central control screen. The specific combination of power mode and parking state used as the safety verification threshold can be flexibly mapped according to different automotive-grade safety levels, and is not specifically limited.

[0060] In this embodiment, by calculating the time deviation between the current time and the meeting start time, and based on meeting the threshold conditions in the time dimension, the vehicle's power mode or parking status is introduced as a joint judgment criterion to ultimately determine the triggering timing of the pre-meeting scenario. This ensures that before issuing any predictive recommendation instructions that may change the cabin's physical environment, the system must verify whether the vehicle currently possesses the physical prerequisites for safe execution. This eliminates the fatal hidden danger of accidental triggering of scenario adjustments (such as sudden seat movement affecting driving operations) during high-speed driving or complex traffic conditions. While adhering to the absolute bottom line of vehicle driving safety, it achieves high-precision, zero-disruption proactive identification and intelligent push of pre-meeting preparation windows for business people.

[0061] In one embodiment, after user confirmation or automatic system execution, personalized scene execution and effect recording are also included.

[0062] After user confirmation or automatic system execution, commands are sent to the cockpit domain controller to precisely coordinate multiple actuators, including the seat (angle, massage), air conditioning (temperature, airflow), multimedia (audio source, volume), ambient lighting, windows, and fragrance system. Corresponding service applications (such as conferencing software and white noise applications) are also invoked to collaboratively realize the meeting scenario. Simultaneously, the system records the user's implicit interaction data throughout the scenario execution (such as which parameters were manually adjusted), serving as crucial feedback data for the closed-loop learning mechanism and supporting continuous model iteration.

[0063] For example, the data recording scope includes sub-functions involved in preset meeting scenarios: window switch, window light transmittance, sunshade switch, audio zone mode, reading light switch, ceiling screen switch, driving mode, etc. Ambient lighting switch, ambient lighting mode, fragrance switch, fragrance type, air conditioning switch, air conditioning mode, etc. The specific functions can be added or removed according to the specific configuration of the vehicle model, and there are no specific limitations.

[0064] In one embodiment, triggering the push of the scenario plan information corresponding to the pre-meeting scenario includes: if the time deviation meets a preset time threshold condition, then broadcasting a voice reminder before the execution of the meeting schedule.

[0065] In complex in-vehicle mobile office environments, business users facing intensive business trips often have their attention focused on driving, in-vehicle conversations, or other miscellaneous tasks. In such situations, relying solely on visual notifications like pop-up cards on the in-vehicle screen is not only easily ignored by users, leading to missed important appointments, but can also forcibly distract the driver's visual attention under certain driving conditions, posing automotive-grade safety hazards. Therefore, this embodiment introduces a non-visual channel voice broadcast mechanism. Specifically, when the time deviation (i.e., the pre-meeting countdown duration) between the current system time and the start time of the online or offline preset meeting scenario parsed from the schedule information meets a preset time threshold, it indicates that the vehicle has effectively entered the pre-meeting buffer period. As a specific application example, the preset time threshold can be flexibly set to 5 minutes or 10 minutes. Once the countdown enters this time limit, the system automatically calls the text-to-speech module inside the cabin and broadcasts a pre-meeting voice reminder to the target passengers through the in-vehicle digital audio amplifier system. For example, when the time deviation reaches 5 minutes (i.e., the system detects that the user needs to start a meeting in 5 minutes), the system can proactively remind the second-row business core area via AI voice: "A meeting is in 5 minutes. You can adjust to meeting mode in advance and prepare for the meeting." Alternatively, 10 minutes before the meeting starts, the system can automatically remind the user of their schedule via voice, and at the same time, a prominent meeting mode entry or prompt card will be displayed on the top of the vehicle's infotainment system desktop, so that the user can easily pre-adjust the air conditioning, seats, lighting, windows, and privacy environment before the meeting through extremely simple manual or voice commands.

[0066] In this embodiment, proactive and predictive information is delivered to the user within a specific time and space buffer before the meeting officially begins, seamlessly transforming static schedule information into dynamic in-vehicle proactive care. This effectively solves the pain points mentioned in the background technology, where the system can only focus on the meeting in isolation, resulting in a lack of continuous intelligent experience for users before the meeting and reliance on cumbersome manual operations. It ensures that business people can obtain high-value meeting reminders in a timely manner without being distracted by checking the in-vehicle screen during business trips, avoiding the risk of being late for the meeting due to focusing on other matters.

[0067] In one embodiment, pushing the corresponding scenario solution information for the meeting transition scenario when the meeting is in the transition scenario further includes: When the system detects that a user has opened a meeting application on the vehicle's infotainment system, it generates a scene card to ask whether to start the preset meeting scene and pushes it to the system via a pop-up window. The system then executes the scene plan in response to the user's manual click command or voice command to start the meeting.

[0068] In real-world business travel scenarios, not all meeting arrangements are meticulously entered into the calendar system in advance. Often, business professionals face unexpected, unscheduled meetings. If the system relies solely on a countdown timer based on time discrepancies, it will have blind spots in its ability to detect these unexpected meetings.

[0069] To eliminate this blind spot, the system monitors the switching of foreground processes or the broadcast signals for application launches on the vehicle's infotainment system in real time. When it detects that a user is clicking and opening a specific meeting-type application on the vehicle's main screen or rear entertainment screen, the user is about to engage in online work within the physical cabin and is currently in a critical pre-meeting preparation period. At this time, the system will not allow the user to frantically search for the control menus for the windows, air conditioning, and seats after opening the app. Instead, it will immediately generate a floating scene card on the currently launched meeting application interface (or in a prominent position on the edge of the screen) and push it as a pop-up. This scene card usually contains clear guiding text, such as: "We have detected that you are about to start a meeting. Do you want to start your private meeting environment with one click?"

[0070] After the card is pushed to the system, users can manually tap the scene card on the screen (e.g., tap the confirm button) or issue a voice command (e.g., say "confirm" or "start meeting mode") depending on their convenience. Once the system receives and responds to either confirmation command, it immediately converts the preset scene into multi-hardware collaborative execution signals, reconstructing a series of physical environmental changes such as closing windows, lowering air conditioning, and repositioning seats, and smoothly handing over screen control to the meeting application.

[0071] In this embodiment, by recognizing the user's behavior of opening a meeting application on the in-vehicle infotainment system, a scenario card is proactively generated and pushed via pop-up. This constructs a scenario trigger defense line that combines schedule prediction and application behavior monitoring, connecting the previously independent software application ecosystem with the underlying cockpit hardware control logic. As a supplement to the schedule trigger mechanism, it gives the system the ability to respond to unplanned and unexpected meetings. At the same time, through lightweight card pop-ups and extremely simple confirmation instructions, it minimizes the user's operation path when dealing with emergency meetings, avoiding the cumbersome switching between the software interface and the vehicle control interface, further solidifying the AI ​​assistant experience of full-scenario coverage and seamless integration in the intelligent cockpit.

[0072] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0073] In one embodiment, such as Figure 5As shown, a cockpit meeting scenario proactive push system 50 is provided, including: Data acquisition module 510 is used to acquire multimodal sensor data corresponding to users in the vehicle cabin; The scheme generation module 520 is used to generate scenario scheme information corresponding to the meeting transition scenario based on multimodal sensing data. The meeting transition scenario includes at least one of the pre-meeting scenario before the preset meeting scenario starts and the post-meeting scenario after the meeting ends. The push module 530 is used to push scenario solution information corresponding to the meeting transition scenario when the meeting is in a transition scenario, including: if the meeting is in a pre-meeting scenario, a reminder message containing cabin environment pre-adjustment options is pushed based on the schedule information, the cabin environment pre-adjustment options including at least one of air conditioning settings and office lighting and atmosphere adjustment; if the meeting is in a post-meeting scenario and the duration of the preset meeting scenario is detected to meet a preset duration threshold, a rest solution containing rest guidance options is pushed.

[0074] Specifically, this embodiment provides a virtual device or hardware system architecture corresponding to the aforementioned method embodiments, which can be embedded in the cockpit domain controller of an intelligent connected vehicle in a software, hardware, or a combination of both. The various modules in this system work collaboratively to break down information silos in traditional cockpit systems and achieve proactive recommendations for business and office scenarios.

[0075] Specific limitations regarding the proactive push system for cockpit conference scenarios can be found in the limitations of the proactive push method for cockpit conference scenarios described above, and will not be repeated here. Each module in the aforementioned proactive push system for cockpit conference scenarios can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in the computer device in hardware form, or stored in the memory of the computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0076] In one embodiment, such as Figure 6 As shown, an electronic device is provided, including a memory 610, a processor 620, and a computer program stored in the memory 610 and executable on the processor. When the processor 620 executes the computer program, it implements the cockpit conference scene active push method in the above embodiment. To avoid repetition, it will not be described again here.

[0077] In one embodiment, a vehicle is provided, including electronic devices as described in the foregoing embodiments.

[0078] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, it implements the cockpit conference scene active push method described in the above embodiment. To avoid repetition, it will not be described again here.

[0079] 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, and should all be included within the protection scope of this application.

Claims

1. A method for proactively pushing content in a cockpit conference scenario, characterized in that, include: Acquire multimodal sensor data corresponding to users in the vehicle cabin; Based on the multimodal sensing data, scenario scheme information corresponding to the meeting transition scenario is generated, wherein the meeting transition scenario includes at least one of the pre-meeting scenario before the preset meeting scenario starts and the post-meeting scenario after the meeting ends. When in the meeting transition scenario, push the corresponding scenario plan information for the meeting transition scenario, including: if currently in the pre-meeting scenario, push the reminder information containing cabin environment pre-adjustment options based on the schedule information, the cabin environment pre-adjustment options including at least one of air conditioning settings and office lighting and atmosphere adjustment; if currently in the post-meeting scenario, and the duration of the preset meeting scenario is detected to meet the preset duration threshold, push the rest plan containing rest guidance options.

2. The method according to claim 1, characterized in that, The multimodal sensing data includes the user's historical interaction data in historical meeting transition scenarios. The step of generating scenario solution information corresponding to the meeting transition scenario based on the multimodal sensing data includes: The historical interaction data is analyzed based on a preset habit matrix to extract user habit features that reflect the user's operating habits. The habit matrix includes the user's habitual operating parameters for cabin environment, cabin privacy and cabin audio in the historical meeting transition scenarios. Based on the user habit characteristics, the scenario solution information corresponding to the meeting transition scenario is generated.

3. The method according to claim 1, characterized in that, The multimodal sensing data includes the user's profile information and real-time environmental parameters in the current meeting transition scenario. The real-time environmental parameters include vehicle status information and cabin environment information. Generating corresponding scenario solution information based on the multimodal sensing data includes: The following steps are performed using a pre-defined intent prediction model to generate the scenario solution information: The image information, vehicle status information, and cabin environment information are extracted and feature encoded to obtain corresponding multimodal heterogeneous feature vectors. The profile feature vector representing user preference features in the multimodal heterogeneous feature vector is used as a weighted constraint matrix, and a matrix dot product operation is performed with the vehicle state feature vector and the cabin environment feature vector representing instantaneous environmental state to obtain the aligned cross-modal joint feature vector. The cross-modal joint feature vector is synchronously input into two parallel branch networks, so as to calculate the current state index characterizing the current cabin environment adaptability through the first branch network mapping, and to calculate the potential intention probability value characterizing the user's conference preparation tendency through the second branch network mapping. The current state indicator and the potential intent probability value are combined with preset scenarios in the preset scenario library to perform joint feature matching, and preset scenario information corresponding to both the current state indicator and the potential intent probability value is determined as the scenario solution information.

4. The method according to claim 1, characterized in that, The push of scenario solution information corresponding to the meeting transition scenario includes: Push scene cards corresponding to the scene solution information, and execute the scene solution corresponding to the scene solution information in response to user commands or preset trigger conditions.

5. The method according to any one of claims 1-4, characterized in that, The process of executing the preset meeting scenario also includes: Record the user's active adjustment operations on the target cockpit function items during the execution of the preset meeting scenario; Within multiple consecutive driving cycles, the frequency of operation for the target cockpit function items is statistically analyzed; If the operation frequency meets the preset ratio threshold, the initial parameters of the corresponding cockpit function item in the preset meeting scenario are updated using the operating parameters corresponding to the active adjustment operation.

6. The method according to any one of claims 1-4, characterized in that, When in the meeting transition scenario, pushing the corresponding scenario solution information for the meeting transition scenario includes: The meeting start time for a preset meeting scenario is obtained based on the schedule information, and the time deviation between the current time and the meeting start time is calculated. If the time deviation meets the preset time threshold condition, then based on the current vehicle power mode or parking status, it is determined that the current situation is the pre-meeting scenario, and the push of the scenario plan information corresponding to the pre-meeting scenario is triggered.

7. The method according to claim 6, characterized in that, The triggering of the push of the scenario plan information corresponding to the pre-meeting scenario includes: If the time deviation meets the preset time threshold condition, a voice reminder before the execution of the meeting schedule will be played.

8. The method according to claim 1, characterized in that, The step of pushing the corresponding scenario solution information for the meeting transition scenario when the meeting is in the transition scenario further includes: When the system detects that a user has opened a meeting application on the vehicle's infotainment system, it generates a scene card to ask whether to start the preset meeting scene and pushes it to the system via a pop-up window. The system then executes the scene plan in response to the user's manual click command or voice command to start the meeting.

9. An electronic device, characterized in that, include: processor; And memory, used to store computer programs; The computer program, when executed by the processor, implements the method as described in any one of claims 1 to 6.

10. A vehicle, characterized in that, Including the electronic device as described in claim 9.

Images