Adaptive cabin environment-based personalized configuration recommendation method and system

Abstract

The application provides a kind of personalized configuration recommendation method and system based on adaptive cabin environment, first, obtain the real-time perception data set of cabin environment and user behavior data set, cover physical environment parameters, equipment operating state and user historical operation, physiological signal, voice interaction and other information, the real-time perception data set of cabin environment and user behavior data set are carried out feature fusion, generate target association feature set to describe the corresponding relationship between environment and user behavior, call knowledge association graph to carry out semantic reasoning, generate configuration demand intention feature set to mine user potential demand, generate candidate configuration scheme based on the configuration demand intention feature set and sort, generate final personalized configuration recommendation result, at the same time, according to the semantic association rule and feature association strength parameter of knowledge association graph updated by user actual interaction feedback, realize the dynamic optimization of recommendation strategy, provide accurate, personalized cabin configuration recommendation for user.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent cockpit, in particular to a personalized configuration recommendation method and system based on adaptive cockpit environment. BACKGROUND

[0002] In modern cockpit application scenarios, such as automotive cockpit, aviation cockpit, etc., it is crucial to provide users with personalized and comfortable configuration experience. Currently, existing cockpit configuration recommendation methods have many limitations. On the one hand, some methods only make configuration recommendations based on simple preset rules. These rules are usually fixed and cannot be flexibly adjusted according to the dynamic changes of the cockpit environment and the diverse behavior characteristics of users. For example, in an automotive cockpit, when the external environment temperature suddenly changes, the preset rules may not be able to timely perceive and adjust the configurations of the air conditioner, seat heating, etc. of the cockpit to meet the user's needs. On the other hand, although some methods take into account the user's historical configuration operation records, they ignore real-time environmental factors of the cockpit and multi-dimensional data such as user physiological signals and voice interactions, resulting in a deviation between the recommended results and the user's actual needs. In addition, existing methods lack effective use of user feedback, cannot continuously optimize the recommendation strategy according to the user's actual interaction, and are difficult to continuously improve the accuracy and personalization of configuration recommendations, making it difficult to meet the growing personalized needs of users for cockpit configurations. SUMMARY

[0003] In view of the above-mentioned problems, in combination with the first aspect of the present application, the embodiments of the present application provide a personalized configuration recommendation method based on adaptive cockpit environment, which comprises:

[0004] Obtaining a cockpit environment real-time perception data set and a user behavior data set, the cockpit environment real-time perception data set containing physical environment parameter information and cockpit device running state information collected by environmental monitoring devices, and the user behavior data set containing user historical configuration operation records, physiological signal data and voice interaction content;

[0005] Performing feature fusion processing on the cockpit environment real-time perception data set and the user behavior data set to generate a target associated feature set, the target associated feature set containing a description of the correspondence between environmental state features and user behavior features;

[0006] Calling a pre-constructed knowledge association graph to perform semantic reasoning processing on the target associated feature set to generate a configuration demand intent feature set, the configuration demand intent feature set containing a description of the user's potential cockpit function configuration demand;

[0007] generate a candidate configuration scheme set based on the configuration demand intention feature set, and prioritize the candidate configuration scheme set according to the feature correlation strength in the target correlation feature set to generate a final personalized configuration recommendation result;

[0008] According to actual interaction feedback information of the user on the final personalized configuration recommendation result, update the semantic correlation rule and the feature correlation strength parameter of the knowledge correlation graph.

[0009] In still another aspect, the embodiment of the present application also provides a personalized configuration recommendation system based on an adaptive cockpit environment, comprising a processor, a machine readable storage medium, the machine readable storage medium is connected with the processor, the machine readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine readable storage medium to realize the above method.

[0010] Based on the above aspects, the embodiment of the present application comprehensively covers cockpit physical environment parameters, device running states, and multi-dimensional information such as user historical configuration operations, physiological signals and voice interactions by acquiring a cockpit environment real-time perception data set and a user behavior data set, performs feature fusion processing on the cockpit environment real-time perception data set and the user behavior data set, generates a target correlation feature set containing a description of the correspondence relationship between environment state features and user behavior features, realizes deep correlation analysis between the environment and the user behavior, calls a pre-constructed knowledge correlation graph for semantic reasoning processing, generates a configuration demand intention feature set describing the user's potential cockpit function configuration demand description, further mines the user's potential demand, generates a candidate configuration scheme set based on the configuration demand intention feature set, and prioritizes the candidate configuration scheme set according to the feature correlation strength to generate a final personalized configuration recommendation result, ensures the accuracy and individuality of the recommendation result, finally, updates the semantic correlation rule and the feature correlation strength parameter of the knowledge correlation graph according to the actual interaction feedback information of the user on the recommendation result, realizes dynamic optimization of the recommendation strategy, can continuously adapt to changes in the cockpit environment and changes in the user's demand, and provides more accurate and personalized cockpit configuration recommendation services for the user. BRIEF DESCRIPTION OF DRAWINGS

[0011] Figure 1 is an execution flow diagram of the personalized configuration recommendation method based on an adaptive cockpit environment provided by the embodiment of the present application.

[0012] Figure 2 is a schematic diagram of exemplary hardware and software components of the personalized configuration recommendation system based on an adaptive cockpit environment provided by the embodiment of the present application. DETAILED DESCRIPTION

[0013] The present application will be specifically described below in conjunction with the drawings of the specification.Figure 1 is a flowchart of a personalized configuration recommendation method based on an adaptive cockpit environment provided by an embodiment of the present application. The personalized configuration recommendation method based on an adaptive cockpit environment will be described in detail below.

[0014] Step S110: Obtain a cockpit environment real-time perception data set and a user behavior data set. The cockpit environment real-time perception data set includes physical environment parameter information and cockpit device operating state information collected by an environment monitoring device. The user behavior data set includes user historical configuration operation records, physiological signal data, and voice interaction content.

[0015] This embodiment takes a smart car cockpit as a unified application scenario for illustration. In order to provide personalized cockpit configuration recommendations for users, first, the relevant cockpit environment real-time perception data set and user behavior data set are obtained.

[0016] Step S111: Collect physical environment parameter information through temperature sensors, humidity sensors, light sensors, and noise sensors arranged in the cockpit. The physical environment parameter information includes continuous time series of temperature measurement values, humidity measurement values, light intensity measurement values, and noise intensity measurement values.

[0017] In this embodiment, the temperature sensor in the cockpit continuously perceives the cockpit temperature, records the temperature measurement values at different times, forms continuous time series of temperature measurement values, and reflects the change of the cockpit temperature over time. The humidity sensor collects the cockpit humidity in real time to obtain continuous time series of humidity measurement values, which reflects the dynamic change of the cockpit humidity. The light sensor monitors the light intensity in the cockpit to generate continuous time series of light intensity measurement values, which helps to understand the change rule of the cockpit light. The noise sensor collects the noise in the cockpit to generate continuous time series of noise intensity measurement values, which reflects the change of the cockpit noise level.

[0018] Step S112: Collect cockpit device operating state information through a state monitoring module of a cockpit device control system. The cockpit device operating state information includes the current working mode of the air conditioning system, the current position parameter of the seat adjustment system, the current volume level of the car audio system, and the current opening degree of the window control system.

[0019] In this embodiment, the state monitoring module of the cockpit device control system obtains the current working mode of the air conditioning system in real time, which has discrete modes such as refrigeration, heating, and ventilation, and can be switched according to the cockpit temperature and user demand. At the same time, the current position parameter of the seat adjustment system is collected, including continuous coordinate values in terms of front and back, up and down, and angle, which accurately represent the specific position state of the seat. The current volume level of the car audio system and the current opening degree of the window control system are also recorded to fully understand the cockpit device operating state.

[0020] Step S113: Extract the user's historical configuration operation records from the cabin interaction log database, which contains the user's adjustment timestamp information and adjustment amplitude information for air conditioning temperature, seat angle, sound effect, and window opening degree.

[0021] In this embodiment, the user's historical configuration operation records are extracted from the cabin interaction log database, which records in detail the user's past adjustment timestamp and adjustment amplitude information for the configuration of devices such as air conditioning temperature, seat angle, sound effect, and window opening degree in the cabin. By analyzing the above records, the user's operation habits and preferences can be understood.

[0022] Step S114: Collect the user's physiological signal data through the biological sensors installed in the cabin, which includes heart rate fluctuation sequence, skin electric response intensity sequence, and respiratory frequency change sequence.

[0023] In this embodiment, the biological sensors installed in the cabin monitor the user's physiological state in real time, collecting physiological signal data such as heart rate fluctuation sequence, skin electric response intensity sequence, and respiratory frequency change sequence. These physiological signal data can reflect the user's physiological response under different environments and operations.

[0024] Step S115: Extract the user's voice interaction content through the dialogue record storage module of the voice interaction system, which includes the user's explicit demand expression sentences and implicit preference evaluation sentences for the cabin environment.

[0025] In this embodiment, the user's voice interaction content is extracted from the dialogue record storage module of the voice interaction system, which includes the user's explicit demand expression sentences for the cabin environment, such as explicitly proposing to adjust the air conditioning temperature or sound effect, and implicit preference evaluation sentences, such as expressing satisfaction or dissatisfaction with certain aspects of the current cabin environment.

[0026] Step S116: Align the physical environment parameter information, cabin device running state information, historical configuration operation records, physiological signal data, and voice interaction content by timestamp to generate a set of cabin environment real-time perception data and a set of user behavior data with time synchronization relationship.

[0027] In this embodiment, the collected and extracted physical environment parameter information, cabin device running state information, historical configuration operation records, physiological signal data, and voice interaction content are aligned by timestamp, so that these data have a synchronous relationship in time, thereby generating a set of cabin environment real-time perception data and a set of user behavior data, facilitating subsequent feature fusion and analysis.

[0028] Step S120: Feature fusion processing is performed on the cockpit environment real-time perception data set and the user behavior data set to generate a target association feature set, which contains a corresponding relationship description of environment state features and user behavior features.

[0029] In this embodiment, after obtaining the cockpit environment real-time perception data set and the user behavior data set, feature fusion processing is needed to mine the corresponding relationship between environment state features and user behavior features. Different types of data are processed respectively.

[0030] Step S121: Time window statistical processing is performed on the physical environment parameter information in the cockpit environment real-time perception data set to calculate the temperature average value, humidity variance value, light intensity median value, and noise intensity maximum value in each time window, and to generate environment state statistical features.

[0031] In this embodiment, time window statistical processing is adopted for the physical environment parameter information in the cockpit environment real-time perception data set. The continuous time sequence is divided into several time windows, and different physical environment parameters are calculated in each time window. For temperature measurement values, the average value of all temperature measurement values in the time window is calculated, which can reflect the overall level of the cockpit temperature in the time window. For humidity measurement values, the variance value is calculated, which can reflect the dispersion degree of humidity in the time window and reflect the fluctuation of humidity. For light intensity measurement values, the median value is calculated, which can represent the intermediate level of light intensity in the time window. For noise intensity measurement values, the maximum value in the time window is found, which can reflect the maximum level of the noise in the cockpit in the time window. Through these statistical calculations, environment state statistical features are generated, which comprehensively reflect the physical environment state of the cockpit in each time window.

[0032] Step S122: Mode encoding processing is performed on the cockpit device running state information in the cockpit environment real-time perception data set to convert the air conditioning system working mode into discrete encoding values of refrigeration / heating / ventilation, and to convert the seat adjustment system position parameters into continuous coordinate values of front / back / up / down / angle, to generate device state mode features.

[0033] For the cockpit equipment running state information, mode coding processing is performed. For the working mode of the air conditioning system, since it has different discrete modes such as refrigeration, heating, ventilation, etc., these modes are respectively converted into corresponding discrete code values, so that subsequent processing and analysis can be facilitated. For the position parameters of the seat adjustment system, they are converted into continuous coordinate values in terms of front and back, up and down, and angle, which can accurately represent the specific position state of the seat. Through the above mode coding processing, the equipment state mode features are generated, which reflect the running mode and position state of the cockpit equipment.

[0034] Step S123: Frequency analysis processing is performed on the historical configuration operation records in the user behavior data set, and the adjustment frequency and single adjustment duration of the user to each cockpit equipment configuration are counted, to generate user operation habit features.

[0035] The historical configuration operation records in the user behavior data set are subjected to detailed frequency analysis processing to generate features reflecting the user's operation habits.

[0036] Step S1231: The historical configuration operation records are arranged in chronological order to generate an operation time sequence.

[0037] The historical configuration operation records extracted from the cockpit interaction log database are arranged in chronological order to form an operation time sequence. The operation time sequence effectively shows the sequence and time points of the user's adjustment of the cockpit equipment configuration at different times.

[0038] Step S1232: According to the operation time sequence, the time interval between two adjacent adjustments of the same equipment configuration is calculated, and the number of adjustments per unit time is counted as the adjustment frequency.

[0039] According to the generated operation time sequence, the time interval between two adjacent adjustments of the same cockpit equipment configuration is calculated. For example, for the air conditioning system, the time interval between two adjacent adjustments of the air conditioning temperature is calculated. Then the number of adjustments of the equipment configuration per unit time is counted, which is the adjustment frequency. By analyzing the adjustment frequency, the frequency of the user's adjustment of different cockpit equipment configurations can be understood.

[0040] Step S1233: According to the operation time sequence, the start time stamp and end time stamp of each adjustment operation are extracted, and the time difference is calculated as the single adjustment duration.

[0041] The start time stamp and the end time stamp of each adjustment operation of the cabin equipment configuration are extracted from the operation time sequence, and the time difference obtained by subtracting the start time stamp from the end time stamp is the single adjustment duration. For example, when the user adjusts the seat angle, the start time and the end time of the adjustment are recorded, and the time difference between the two is the single adjustment duration of adjusting the seat angle.

[0042] Step S1234: The adjustment frequency and the single adjustment duration of different equipment types are counted respectively to generate the air conditioning system adjustment frequency, the seat adjustment system adjustment frequency, the vehicle audio system adjustment frequency, the window control system adjustment frequency, and the corresponding single adjustment duration distribution.

[0043] The adjustment frequency and the single adjustment duration of different types of cabin equipment, such as the air conditioning system, the seat adjustment system, the vehicle audio system, and the window control system, are counted. For each equipment type, the adjustment frequency and the single adjustment duration distribution are obtained. For example, the adjustment frequency of the air conditioning system in a period of time and the single adjustment duration distribution of each adjustment of the air conditioning temperature are counted, so that the user's operation habit characteristics of different equipment can be understood in detail.

[0044] Step S1235: The adjustment frequency and the single adjustment duration are standardized to generate the standardized user operation habit characteristics.

[0045] In order to make the adjustment frequency and the single adjustment duration of different equipment types comparable, they are standardized. Standardization can eliminate the dimensional difference and the numerical range difference between different equipment data. After standardization, the generated standardized user operation habit characteristics can more accurately reflect the user's operation habit of each cabin equipment.

[0046] Step S124: The physiological signal data in the user behavior data set is processed for feature extraction, the peak-valley difference of heart rate fluctuation, the rising slope of skin electric response, and the periodic interval of respiratory frequency are calculated, and the user physiological feedback characteristics are generated.

[0047] For the physiological signal data in the user behavior data set, feature extraction processing is performed. For the heart rate fluctuation sequence, the peak-valley difference thereof is calculated, which can reflect the fluctuation amplitude of the heart rate within a set time and embody the change degree of the user's physiological state. For the galvanic skin response intensity sequence, the rising edge slope thereof is calculated, which can reflect the change speed of the galvanic skin response and reflect the change of the user's emotion and the like. For the respiratory frequency change sequence, the periodic interval thereof is calculated, which can embody the regularity and rhythm of respiration. Through these feature extraction processes, user physiological feedback features are generated, which reflect the physiological reactions of the user in the current cabin environment.

[0048] Step S125: performing semantic analysis processing on the voice interaction content in the user behavior data set to extract specific demand keywords in the explicit demand sentence and extract emotional tendency words in the implicit preference evaluation sentence to generate user semantic demand features.

[0049] The semantic analysis processing is performed on the voice interaction content in the user behavior data set. For the explicit demand sentence, the specific demand keywords therein are extracted through the semantic analysis method. For example, if the user says "turn up the air conditioner temperature a bit", then "air conditioner temperature" and "turn up" are the specific demand keywords. For the implicit preference evaluation sentence, the emotional tendency words therein are extracted, such as "comfortable", "uncomfortable", "satisfied", "dissatisfied", and the like, which can reflect the user's preference and attitude towards the cabin environment. Through the above semantic analysis processing, the user semantic demand features are generated, which embody the user's semantic demand and preference.

[0050] Step S126: performing feature association calculation on the environment state statistical features, device state mode features, user operation habit features, user physiological feedback features and user semantic demand features to determine the co-occurrence frequency and conditional probability between each pair of environment features and user features to generate a target association feature set containing feature association degree values.

[0051] After obtaining the environment state statistical features, device state mode features, user operation habit features, user physiological feedback features and user semantic demand features, feature association calculation is performed on these features to determine the relationship between the environment features and the user features.

[0052] For example, step S1261: the environment state statistical features and the device state mode features are classified into an environment dimension feature group, and the user operation habit features, the user physiological feedback features and the user semantic demand features are classified into a user dimension feature group.

[0053] The environmental state statistical features and the equipment state mode features are classified into one category to form an environmental dimension feature group. The environmental state statistical features are obtained by time window statistical processing of cabin physical environment parameter information, and the equipment state mode features are obtained by mode encoding processing of cabin equipment running state information. They are all related to the environment and equipment state of the cabin. Meanwhile, the user operation habit features, the user physiological feedback features and the user semantic demand features are classified into another category to form a user dimension feature group. The user operation habit features are obtained by frequency analysis processing of user historical configuration operation records, the user physiological feedback features are obtained by feature extraction processing of user physiological signal data, and the user semantic demand features are obtained by semantic analysis processing of user voice interaction content. The above features all reflect the behavior, physiology and demand of the user.

[0054] Step S1262: Extract a time synchronization relationship identifier from the set of cabin environment real-time perception data and the set of user behavior data, and establish a time alignment mapping table of the environmental dimension feature group and the user dimension feature group based on the time synchronization relationship identifier.

[0055] The time synchronization relationship identifier is extracted from the set of cabin environment real-time perception data and the set of user behavior data generated previously, which have the time synchronization relationship. The time synchronization relationship identifier can be a timestamp or other information that can reflect the time correlation of the data. Based on the time synchronization relationship identifier, a time alignment mapping table of the environmental dimension feature group and the user dimension feature group is established. The mapping table corresponds the features in the environmental dimension feature group and the user dimension feature group in time, so that the environmental features and the user features can be correlated and analyzed at the same time point.

[0056] Step S1263: Traverse each time synchronization node in the time alignment mapping table, and extract a first feature instance in the environmental dimension feature group and a second feature instance in the user dimension feature group corresponding to the time synchronization node.

[0057] Each time synchronization node in the established time alignment mapping table is traversed. At each time synchronization node, a corresponding first feature instance is extracted from the environmental dimension feature group, and a corresponding second feature instance is extracted from the user dimension feature group. For example, at a certain time point, the specific instances of the environmental state statistical features and the equipment state mode features at the time point are extracted from the environmental dimension feature group, and the specific instances of the user operation habit features, the user physiological feedback features and the user semantic demand features at the time point are extracted from the user dimension feature group.

[0058] Step S1264: Perform feature pair matching processing on the first feature instance and the second feature instance to generate a feature pair sequence composed of a single environmental dimension feature instance and a single user dimension feature instance.

[0059] The extracted first feature instances and second feature instances are subjected to feature pair matching processing. Single environmental dimension feature instances are combined with single user dimension feature instances to form feature pairs. For example, an environmental temperature statistical feature at a certain time point is combined with a user operation habit feature on air conditioner temperature at the same time point to form a feature pair. Through the above matching processing on all extracted feature instances, a feature pair sequence composed of multiple feature pairs is generated.

[0060] Step S1265: The number of times that a same feature pair repeatedly appears at different time synchronization nodes in the feature pair sequence is counted to generate a feature pair co-occurrence number statistical result.

[0061] The generated feature pair sequence is subjected to statistical analysis, and the number of times that a same feature pair repeatedly appears at different time synchronization nodes is counted. For example, for a feature pair composed of an environmental temperature statistical feature and a user operation habit feature on air conditioner temperature, the number of times that it repeatedly appears at different time points is counted. Through the above statistical analysis, a co-occurrence number statistical result of each feature pair is obtained, which reflects the co-occurrence frequency of environmental features and user features in time.

[0062] Step S1266: The ratio of the feature pair co-occurrence number statistical result to the total number of time synchronization nodes is calculated to generate a feature pair co-occurrence frequency parameter.

[0063] The feature pair co-occurrence number statistical result is divided by the total number of time synchronization nodes to obtain a feature pair co-occurrence frequency parameter. The feature pair co-occurrence frequency parameter represents the frequency of the feature pair appearing in all time synchronization nodes, and reflects the co-occurrence degree between environmental features and user features. For example, the co-occurrence number of a certain feature pair is A, and the total number of time synchronization nodes is B, then the co-occurrence frequency parameter of the feature pair is A divided by B.

[0064] Step S1267: The number of times that a single user dimension feature instance in the user dimension feature group independently appears in all time synchronization nodes is counted to generate a user feature independent appearance number statistical result.

[0065] The single user dimension feature instance in the user dimension feature group is counted, and the number of times that it independently appears in all time synchronization nodes is counted, without considering the association with other environmental features. For example, for a user operation habit feature on air conditioner temperature, the number of times that it independently appears at all time points is counted. Through the above statistical analysis, an independent appearance number statistical result of each user dimension feature instance is generated, which can reflect the appearance frequency of the user feature in time.

[0066] Step S1268: Count the number of times that each instance of the environment dimension feature in the environment dimension feature group appears independently in all time synchronization nodes, and generate a number-of-time-independent-appearances statistical result of the environment feature.

[0067] Similarly, count the number of times that each instance of the environment dimension feature in the environment dimension feature group appears independently in all time synchronization nodes. For example, count the number of times that the environment temperature statistical feature appears independently at all time points. Through the above counting, generate a number-of-time-independent-appearances statistical result of each instance of the environment dimension feature, which can reflect the frequency of appearance of the environment feature in time.

[0068] Step S1269: Calculate the ratio of the number-of-time-independent-appearances statistical result of the user feature to the number-of-time-independent-appearances statistical result of the corresponding environment feature, and generate an environment feature appearance probability under the condition of the user feature.

[0069] Divide the number-of-time-independent-appearances statistical result of the user feature by the number-of-time-independent-appearances statistical result of the corresponding environment feature to obtain the environment feature appearance probability under the condition of the user feature. The environment feature appearance probability represents the possibility of the appearance of the corresponding environment feature under the condition that the user appears a certain feature. For example, if the number-of-time-independent-appearances statistical result of the corresponding environment feature is C, and the number-of-time-independent-appearances statistical result of the corresponding user feature is A, then the environment feature appearance probability under the condition of the user feature is A divided by C.

[0070] Step S12610: Calculate the ratio of the number-of-time-independent-appearances statistical result of the user feature to the number-of-time-independent-appearances statistical result of the corresponding environment feature, and generate a user feature appearance probability under the condition of the environment feature.

[0071] Divide the number-of-time-independent-appearances statistical result of the user feature by the number-of-time-independent-appearances statistical result of the corresponding environment feature to obtain the environment feature appearance probability under the condition of the user feature. The environment feature appearance probability represents the possibility of the appearance of the corresponding environment feature under the condition that the user appears a certain feature. For example, if the number-of-time-independent-appearances statistical result of the corresponding environment feature is C, and the number-of-time-independent-appearances statistical result of the corresponding user feature is A, then the environment feature appearance probability under the condition of the user feature is A divided by C.

[0072] Step S12611: Perform weighted fusion processing on the number-of-time-independent-appearances statistical result of the feature pair, the environment feature appearance probability under the condition of the user feature, and the user feature appearance probability under the condition of the environment feature, and generate a feature pair correlation degree value.

[0073] The feature pair co-occurrence frequency parameter, the environment feature appearance probability under the user feature condition, and the user feature appearance probability under the environment feature condition are weighted and fused. Different weights can be assigned to the three parameters according to different importance, and then they are fused to obtain a feature pair correlation degree value. The feature pair correlation degree value comprehensively reflects the correlation closeness between the environment feature and the user feature.

[0074] Step S12612: The above-mentioned co-occurrence number statistics, co-occurrence frequency calculation, conditional probability calculation, and correlation degree value generation processing are performed on all possible feature pairs in the environment dimension feature group and the user dimension feature group, to generate a feature correlation degree matrix containing the correlation degree values of the feature pairs.

[0075] The co-occurrence number statistics, co-occurrence frequency calculation, conditional probability calculation, and correlation degree value generation processing are performed on all possible feature pairs in the environment dimension feature group and the user dimension feature group. The correlation degree values of each feature pair are arranged into a matrix form to obtain a feature correlation degree matrix. The feature correlation degree matrix comprehensively shows the correlation degrees between all feature pairs in the environment dimension feature group and the user dimension feature group.

[0076] Step S12613: The feature correlation degree matrix is subjected to structured conversion processing to generate a target correlation feature set containing environment feature identifiers, user feature identifiers, and corresponding correlation degree values.

[0077] The feature correlation degree matrix is subjected to structured conversion processing. The environment feature identifier and the user feature identifier are added to each element in the matrix to clearly indicate the environment feature and the user feature corresponding to each correlation degree value. Through the above-mentioned conversion, a target correlation feature set is generated, which effectively records the correlation degree information between the environment feature and the user feature, facilitating subsequent semantic reasoning and configuration recommendation.

[0078] Step S130: The pre-constructed knowledge correlation graph is called to perform semantic reasoning processing on the target correlation feature set to generate a configuration requirement intention feature set containing the user's potential cabin function configuration requirement description.

[0079] After obtaining the target correlation feature set, the pre-constructed knowledge correlation graph is called to perform semantic reasoning processing. The knowledge correlation graph is a graph containing a large amount of environment information, user behavior information, and their correlation relationships. Different features in the target correlation feature set are input into different matching units of the knowledge correlation graph.

[0080] Step S131: The environment state statistical feature in the target correlation feature set is input into the environment entity matching unit of the knowledge correlation graph to match the environment parameter entity nodes stored in the knowledge correlation graph.

[0081] The environmental state statistical features in the target association feature set are input to an environmental entity matching unit of the knowledge association graph. The environmental entity matching unit searches for environmental parameter entity nodes matching the input environmental state statistical features in the knowledge association graph. For example, if the environmental state statistical features include temperature average value, humidity variance value and other information in a certain time period, the environmental entity matching unit finds corresponding temperature, humidity and other environmental parameter entity nodes in the knowledge association graph, which store related environmental information and association relationships.

[0082] Step S132: input the user operation habit features in the target association feature set to a behavior entity matching unit of the knowledge association graph, and match user behavior pattern entity nodes stored in the knowledge association graph.

[0083] The user operation habit features in the target association feature set are input to a behavior entity matching unit of the knowledge association graph. The behavior entity matching unit searches for user behavior pattern entity nodes matching the input user operation habit features in the knowledge association graph. For example, the user operation habit features include user adjustment frequency and single adjustment duration of the air conditioning system and other information, and the behavior entity matching unit finds corresponding user operation behavior pattern entity nodes of the air conditioning system in the knowledge association graph, which record operation habits and related association information of the user.

[0084] Step S133: input the user physiological feedback features in the target association feature set to a physiological entity matching unit of the knowledge association graph, and match physiological state entity nodes stored in the knowledge association graph.

[0085] The user physiological feedback features in the target association feature set are input to a physiological entity matching unit of the knowledge association graph. The physiological entity matching unit searches for physiological state entity nodes matching the input user physiological feedback features in the knowledge association graph. For example, the user physiological feedback features include peak-valley difference of heart rate fluctuation, rising edge slope of skin electric response and other information, and the physiological entity matching unit finds corresponding heart rate, skin electric response and other physiological state entity nodes in the knowledge association graph, which store physiological state information and association relationships of the user.

[0086] Step S134: input the user semantic demand features in the target association feature set to a demand entity matching unit of the knowledge association graph, and match demand intention entity nodes stored in the knowledge association graph.

[0087] The user semantic demand features in the target association feature set are input to a demand entity matching unit of the knowledge association graph. The demand entity matching unit finds demand intent entity nodes in the knowledge association graph that match the input user semantic demand features. For example, the user semantic demand features include specific demand keywords and emotional tendency words for the cockpit environment, and the demand entity matching unit finds corresponding demand intent entity nodes in the knowledge association graph, which reflect the user's demand and preference.

[0088] Step S135: A relationship reasoning module of the knowledge association graph is called to analyze the historical association relationship between the environmental parameter entity nodes and the user behavior pattern entity nodes, analyze the causal relationship between the user behavior pattern entity nodes and the physiological state entity nodes, and analyze the mapping relationship between the physiological state entity nodes and the demand intent entity nodes.

[0089] After the matching of each feature with the entity nodes in the knowledge association graph is completed, a relationship reasoning module of the knowledge association graph is called for in-depth analysis.

[0090] Step S1351: Historical association edges between the environmental parameter entity nodes and the user behavior pattern entity nodes are extracted from the knowledge association graph, and the historical association edges include association times and association confidence.

[0091] In the knowledge association graph, there are historical association edges between the environmental parameter entity nodes and the user behavior pattern entity nodes. These historical association edges are extracted from the knowledge association graph, and each historical association edge includes information such as association times and association confidence. The association times represent the number of times the environmental parameter entity nodes and the user behavior pattern entity nodes appear together in historical data, and the association confidence reflects the reliability and credibility of the above association.

[0092] Step S1352: The co-occurrence times of different environmental parameter values and user behavior patterns are counted, and the conditional probability of the environmental parameter values and the user behavior patterns is calculated, which is the probability of the user exhibiting the behavior pattern under the given environmental parameter values.

[0093] The co-occurrence times of different environmental parameter values and user behavior patterns are counted. For example, the number of times the user adjusts the air conditioning temperature under different cockpit temperature environments is counted. Then, the conditional probability of the environmental parameter values and the user behavior patterns is calculated according to the co-occurrence times. The conditional probability represents the likelihood of the user exhibiting a specific behavior pattern given a certain environmental parameter value. By calculating the conditional probability, the influence of the environmental parameter on the user behavior pattern can be more accurately understood.

[0094] Step S1353: generating a historical association strength between the environment parameter entity node and the user behavior pattern entity node according to the co-occurrence number and the conditional probability.

[0095] The co-occurrence number and the conditional probability are combined to generate a historical association strength between the environment parameter entity node and the user behavior pattern entity node. The co-occurrence number and the conditional probability can be considered comprehensively by a certain calculation method (such as weighted combination) to obtain a strength value that reflects the historical association tightness between them. The higher the historical association strength value is, the tighter the association between the environment parameter and the user behavior pattern is.

[0096] Step S1354: taking the historical association strength as a weight parameter of relationship reasoning, which is used for weight distribution in the subsequent multi-hop reasoning process.

[0097] The generated historical association strength is taken as a weight parameter of relationship reasoning. In the subsequent multi-hop reasoning process, different association relationships are assigned weights according to the weight parameter. For example, when reasoning the user's potential configuration demand intention, if the historical association strength between a certain environment parameter entity node and a user behavior pattern entity node is high, the weight of this association relationship in the decision-making process will be larger, and it is more likely to affect the final reasoning result.

[0098] Step S1355: analyzing the causal relationship between the user behavior pattern entity node and the physiological state entity node.

[0099] In the knowledge association graph, the causal relationship between the user behavior pattern entity node and the physiological state entity node is analyzed. For example, the user's frequent adjustment of the air conditioner temperature may cause changes in physiological states such as heart rate fluctuations. Through the mining and analysis of relevant information in the knowledge association graph, the internal logic of the above causal relationship is found out. This may involve analysis of a large amount of historical data to observe the changes in physiological states corresponding to the user's behavior patterns. For example, the change trend of physiological indicators such as heart rate and skin electric response after the user adjusts the seat position multiple times is counted to determine the causal relationship between the behavior pattern and the physiological state.

[0100] Step S1356: analyzing the mapping relationship between the physiological state entity node and the demand intention entity node.

[0101] The mapping relationship between the physiological state entity node and the demand intention entity node is analyzed. In the knowledge association graph, different physiological states may correspond to different demand intentions. For example, when the user's heart rate fluctuates greatly, the skin electric response is enhanced, and the breathing frequency is accelerated, it may reflect that the user has the demand intention of adjusting the temperature, ventilation, etc. of the cabin environment. By sorting and analyzing the large amount of data stored in the knowledge association graph, the corresponding relationship between the physiological state and the demand intention is determined. The common demand intention of the user under different physiological state combinations can be found out through the classification statistics of the historical data, so as to establish the above mapping relationship.

[0102] For the relationship between the user behavior mode entity node and the physiological state entity node, the causal relationship between them is analyzed. For example, the user's frequent behavior of adjusting the air conditioner temperature may cause changes in physiological states such as heart rate fluctuations. By analyzing the related information in the knowledge association graph, the above causal relationship is found out. For the relationship between the physiological state entity node and the demand intention entity node, the mapping relationship between them is analyzed. For example, the user's heart rate fluctuation may reflect that the user has the demand intention of adjusting the temperature of the cabin environment, and the above mapping relationship is determined through the information in the knowledge association graph.

[0103] Step S136: Based on the historical association relationship, the causal relationship and the mapping relationship, multi-hop reasoning processing is performed to generate potential configuration demand intentions of the user under the current environmental state through historical operation behaviors, physiological feedback and semantic demands, the potential configuration demand intentions including temperature adjustment priority, seat position adjustment direction, sound effect preference type and window opening degree suggestion range.

[0104] Based on the historical association relationship, the causal relationship and the mapping relationship obtained through the above analysis, multi-hop reasoning processing is performed. By comprehensively considering the information of environmental parameters, user behavior modes, physiological states and demand intentions, potential configuration demand intentions of the user under the current environmental state are generated. These potential configuration demand intentions include temperature adjustment priority, i.e. the importance and priority of temperature adjustment of the user; seat position adjustment direction, i.e. the direction in which the user may wish to adjust the seat in front and back, up and down, angle, etc.; sound effect preference type, i.e. the type of sound effect preferred by the user; and window opening degree suggestion range, i.e. the suggested range of window opening degree according to the current environment and user demand.

[0105] Step S137: The potential configuration demand intentions are subjected to structured description processing to generate a configuration demand intention feature set including demand type identification, demand intensity level and demand associated equipment.

[0106] The generated potential configuration requirement intention is processed by structured description. A requirement type identifier is assigned to each potential configuration requirement intention, indicating whether the requirement intention belongs to temperature adjustment, seat adjustment, sound effect adjustment, or window opening adjustment, etc. At the same time, the requirement intensity level is determined, reflecting the intensity of the requirement intention. In addition, the requirement associated equipment is labeled, that is, which cabin equipment is related to the requirement intention. Through the above structured description processing, a configuration requirement intention feature set containing requirement type identifier, requirement intensity level and requirement associated equipment is generated, which effectively reflects the user's potential configuration requirement intention.

[0107] Step S140: Based on the configuration requirement intention feature set, a candidate configuration scheme set is generated, and the candidate configuration scheme set is prioritized according to the feature association strength in the target association feature set, to generate a final personalized configuration recommendation result.

[0108] After obtaining the configuration requirement intention feature set, a candidate configuration scheme set is generated based on the set, and prioritized.

[0109] Step S141: Extract the cabin equipment type corresponding to each requirement intention in the configuration requirement intention feature set, which includes air conditioning system, seat adjustment system, car audio system and window control system.

[0110] The cabin equipment type corresponding to each requirement intention is extracted from the configuration requirement intention feature set. Since the configuration requirement intention contains temperature adjustment, seat adjustment, sound effect adjustment and window opening adjustment, these requirements correspond to different cabin equipment types such as air conditioning system, seat adjustment system, car audio system and window control system. By extracting this information, it can be determined which cabin equipment is related to each requirement intention.

[0111] Step S142: For each cabin equipment type, retrieve the historical effective configuration scheme related to the equipment type from the knowledge association graph, which contains device parameter adjustment value, user feedback record after adjustment and environmental adaptability record.

[0112] For each cockpit device type, the relevant historical effective configuration scheme is retrieved from the knowledge association graph. The knowledge association graph stores a large amount of historical configuration scheme information, and the historical effective configuration scheme contains device parameter adjustment values, user feedback records after adjustment, and environmental adaptability records. For the air conditioning system, the historical effective configuration scheme may contain different temperature adjustment values, air speed adjustment values, and other device parameter adjustment values, as well as user feedback records on comfort after adjustment, and adaptability records of the configuration scheme under different environmental temperature, humidity, and other conditions. For the seat adjustment system, the historical effective configuration scheme will have adjustment values of the seat position such as front and back, up and down, and angle, as well as user feedback on the comfort of the adjusted seat, and the environmental adaptability of the configuration in different driving scenarios. For the car audio system, the historical effective configuration scheme includes device parameter adjustment values such as volume and sound effect mode, user feedback on different sound effects, and adaptability in different noise environments. For the window control system, the historical effective configuration scheme covers the adjustment values of the opening degree of the window, the user feedback on the ventilation effect, and the adaptability in different light and noise intensity environments.

[0113] Step S143: The historical effective configuration schemes that match the environmental state statistical features and user operation habit features in the current target association feature set are screened out, and the candidate configuration sub-scheme set corresponding to each device type is generated.

[0114] From the retrieved historical effective configuration schemes, the schemes that match the environmental state statistical features and user operation habit features in the current target association feature set are screened out. For the environmental state statistical features, the environmental adaptability records in the historical effective configuration schemes can be compared. If a historical configuration scheme performs well under similar environmental conditions to the current environmental state statistical features, and the user feedback is positive, then this scheme is likely to be selected. For the user operation habit features, it can be checked whether the device parameter adjustment values in the historical configuration scheme are consistent with the user's operation habits. For example, if the user often adjusts the air conditioner temperature to a certain range, then the historical effective configuration scheme with device parameter adjustment values in this range will be screened out. Through the above screening process, a candidate configuration sub-scheme set corresponding to each cockpit device type is generated, and each candidate configuration sub-scheme set contains historical effective configuration schemes that match the current environment and user operation habits.

[0115] Step S144: The candidate configuration sub-scheme sets of each device type are combined to generate a candidate configuration scheme set containing multi-device collaborative adjustment.

[0116] The candidate configuration sub-scheme sets corresponding to each cabin equipment type are combined. Since the equipment in the cabin is interrelated, adjustment of one equipment can affect the use effect of other equipment, so the coordinated adjustment of multiple equipment needs to be considered. For example, when adjusting the air conditioner temperature, the opening degree of the window may need to be adjusted at the same time to achieve better ventilation and comfort effect. In the combination process, the candidate configuration sub-schemes of different equipment types can be arranged and combined to generate a candidate configuration scheme set containing coordinated adjustment of multiple equipment. These candidate configuration schemes consider the mutual influence and synergy between different equipment, and can more comprehensively meet the user's needs.

[0117] Step S145: Extracting a feature correlation strength value corresponding to each demand intention from the target correlation feature set, the feature correlation strength value representing the correlation closeness of the demand intention to the current environment state and user behavior.

[0118] The feature correlation strength value corresponding to each demand intention is extracted from the target correlation feature set. In the previous process of generating the target correlation feature set, the correlation degree value between each pair of environment features and user features has been calculated. Here, according to the demand intention, the feature correlation strength value related to the demand intention is extracted from these correlation degree values. The feature correlation strength value reflects the correlation closeness of the demand intention to the current environment state and user behavior. For example, if a demand intention is to adjust the air conditioner temperature, then the correlation degree value between the environment feature (such as the current environment temperature) and the user feature (such as the user's historical air conditioner temperature adjustment habit) related to the air conditioner temperature adjustment is extracted as the feature correlation strength value of the demand intention.

[0119] Step S146: Assigning a weight parameter to each candidate configuration scheme according to the feature correlation strength value, the weight parameter being positively correlated with the feature correlation strength value of the demand intention.

[0120] According to the extracted feature correlation strength value, a weight parameter is assigned to each candidate configuration scheme. Since the feature correlation strength value represents the correlation closeness of the demand intention to the current environment state and user behavior, the weight parameter is positively correlated with the feature correlation strength value of the demand intention. That is, the closer the correlation between the demand intention and the current environment and user behavior, the higher the weight parameter assigned to the candidate configuration scheme corresponding to the demand intention. For example, a candidate configuration scheme closely related to the user's current demand intention for adjusting the air conditioner temperature can be assigned a higher weight parameter, while a candidate configuration scheme less related to the demand intention is assigned a lower weight parameter.

[0121] Step S147: The candidate configuration scheme set is subjected to a weighted sorting process, and the candidate configuration schemes with weight parameters greater than a set threshold are preferentially retained. Then, the sorted candidate configuration scheme set is subjected to a conflict detection process to check whether there is a conflict of different adjustment parameters of the same device or a conflict of environmental impact of multi-device adjustment.

[0122] The candidate configuration scheme set is subjected to a weighted sorting process. The candidate configuration schemes are sorted according to the assigned weight parameters, and the candidate configuration schemes with weight parameters greater than a set threshold are preferentially retained. The set threshold is determined in advance according to actual conditions and is used to screen out candidate configuration schemes closely associated with the current environment and user demand. After retaining these candidate configuration schemes, the sorted candidate configuration scheme set is subjected to a conflict detection process. For the conflict of different adjustment parameters of the same device, for example, one candidate configuration scheme requires the air conditioner temperature to be raised, while another candidate configuration scheme requires the air conditioner temperature to be lowered, which is a conflict. For the conflict of environmental impact of multi-device adjustment, for example, the air conditioner temperature is adjusted while the window is opened, and if the above combination will cause energy waste or comfort reduction under certain environmental conditions, there is a conflict of environmental impact. Through the conflict detection process, candidate configuration schemes with conflicts can be excluded to ensure that the recommended configuration scheme is feasible and reasonable.

[0123] Step S148: According to the check result, the candidate configuration schemes without conflicts are subjected to a user adaptability verification process to obtain the verified candidate configuration schemes. The verification content includes whether the configuration adjustment amplitude conforms to the user's historical operation habit and whether the physiological feedback prediction after adjustment is better than the current state.

[0124] After the conflict detection obtains the candidate configuration schemes without conflicts, the schemes are subjected to a user adaptability verification process to ensure that the recommended scheme truly meets the user's demand.

[0125] Step S1481: The adjustment parameter values of each device in the candidate configuration scheme are extracted, including the air conditioner temperature adjustment value, the seat angle adjustment value, the sound volume adjustment value, and the window opening degree adjustment value.

[0126] The adjustment parameter values of each cabin device are extracted from the candidate configuration schemes without conflicts. For the air conditioning system, the temperature adjustment value is extracted, which represents the air conditioner temperature adjustment amplitude and direction suggested by the candidate configuration scheme; for the seat adjustment system, the seat angle adjustment value is extracted to clarify the adjustment of the seat angle; for the vehicle sound system, the sound volume adjustment value is extracted to understand the sound volume adjustment size suggested by the scheme; for the window control system, the window opening degree adjustment value is extracted to master the adjustment suggestion of the window opening degree.

[0127] Step S1482: Extract the user's adjustment parameter preference range for each device from the user's historical configuration operation records, which contains the minimum parameter value and the maximum parameter value of the user's historical adjustment.

[0128] Extract the user's adjustment parameter preference range for each cabin device from the user's historical configuration operation records. For air conditioning temperature, find the minimum temperature value and the maximum temperature value in the user's historical adjustment process, which constitute the user's adjustment parameter preference range for air conditioning temperature; for seat angle, determine the minimum angle value and the maximum angle value adjusted by the user in the historical operation as the adjustment parameter preference range for seat angle; for audio volume, obtain the minimum volume value and the maximum volume value of the user's historical adjustment to form the adjustment parameter preference range for audio volume; for window opening degree, extract the minimum opening degree value and the maximum opening degree value of the user's historical adjustment as the adjustment parameter preference range for window opening degree.

[0129] Step S1483: Verify whether the adjustment parameter value of the candidate configuration scheme falls within the user's preference range, and generate a parameter adaptability verification result.

[0130] Compare the extracted adjustment parameter values of each device in the candidate configuration scheme with the user's adjustment parameter preference range. If the air conditioning temperature adjustment value is within the user's adjustment parameter preference range for air conditioning temperature, it means that the air conditioning adjustment parameter conforms to the user's historical operation habit; similarly, the adjustment parameter values for seat angle, audio volume and window opening degree are compared with the corresponding user preference range. According to the comparison result, a parameter adaptability verification result is generated, which indicates whether the candidate configuration scheme matches the user's historical operation habit in terms of device adjustment parameters. If the adjustment parameter values of all devices fall within the user's preference range, the parameter adaptability verification result is passed; as long as the adjustment parameter value of one device exceeds the user's preference range, the parameter adaptability verification result is failed.

[0131] Step S1484: Extract the user's physiological feedback features after historical configuration adjustment from the user's physiological signal data, which include heart rate fluctuation amplitude, skin galvanic response intensity and respiratory frequency change rate.

[0132] Extract the user's physiological feedback features after historical configuration adjustment from the user's physiological signal data. For heart rate fluctuation amplitude, analyze the difference between the maximum value and the minimum value of the user's heart rate after each configuration adjustment to obtain the feature value of heart rate fluctuation amplitude; for skin galvanic response intensity, extract the intensity value of the user's skin galvanic response after configuration adjustment; for respiratory frequency change rate, calculate the change rate of the user's respiratory frequency before and after configuration adjustment. These physiological feedback features can reflect the user's physiological state changes after historical configuration adjustment.

[0133] Step S1485: predict the physiological feedback characteristics of the user after adjustment of the candidate configuration scheme, compare the difference between the predicted value and the current physiological feedback characteristics, and generate a physiological adaptability verification result.

[0134] The physiological feedback characteristics of the user after adjustment of the candidate configuration scheme are predicted through a prediction mechanism such as an AI model trained based on historical data. The AI model can learn the physiological feedback rules of the user under different configuration adjustments, and predict physiological feedback characteristics such as heart rate fluctuation amplitude, skin electrical response intensity, and respiratory frequency change rate that the user may have after adjustment according to the device adjustment parameter values in the candidate configuration scheme. The predicted physiological feedback characteristic values are compared with the current physiological feedback characteristics of the user. If the predicted physiological feedback characteristics show that the physiological state of the user is more comfortable, such as smaller heart rate fluctuation amplitude, lower skin electrical response intensity, and more stable respiratory frequency change rate, it indicates that the physiological feedback prediction after adjustment is better than the current state, and the physiological adaptability verification result is passed; otherwise, if the predicted physiological feedback characteristics show that the physiological state of the user is deteriorated, the physiological adaptability verification result is failed.

[0135] Step S1486: integrate the parameter adaptability verification result and the physiological adaptability verification result to determine whether the candidate configuration scheme passes the user adaptability verification, and eliminate the candidate configuration scheme that fails the verification from the recommendation result.

[0136] The parameter adaptability verification result and the physiological adaptability verification result are integrated to determine whether the candidate configuration scheme passes the user adaptability verification. Only when the parameter adaptability verification result is passed and the physiological adaptability verification result is also passed, the candidate configuration scheme passes the user adaptability verification. For the candidate configuration scheme that fails the verification, it is eliminated from the recommendation result to ensure that the final recommended scheme provided to the user not only conforms to the user's historical operation habits, but also makes the user feel more comfortable in physiology.

[0137] Step S149: sort the candidate configuration schemes that pass the verification in descending order according to the weight parameter, and generate a final personalized configuration recommendation result containing a main recommended scheme and alternative recommended schemes.

[0138] The candidate configuration schemes that pass the user adaptability verification are sorted in descending order according to the weight parameter. The candidate configuration scheme with a higher weight parameter is ranked in the front. After sorting, the candidate configuration scheme ranked in the front is selected as the main recommended scheme, which is the scheme that best matches the current environment and user needs and best meets the user's individual needs. The remaining candidate configuration schemes that pass the verification are alternative recommended schemes, thus generating a final personalized configuration recommendation result containing a main recommended scheme and alternative recommended schemes, providing the user with diversified choices.

[0139] Step S150: updating the semantic association rules and feature association strength parameters of the knowledge association graph according to the actual interaction feedback information of the user on the final personalized configuration recommendation result.

[0140] After providing the final personalized configuration recommendation result to the user, the knowledge association graph needs to be updated according to the actual interaction feedback information of the user.

[0141] Step S151: collecting actual interaction feedback information of the user on the final personalized configuration recommendation result, the actual interaction feedback information including user acceptance / rejection operation records on the main recommendation scheme, selection operation records on the alternative recommendation scheme, and voice evaluation content after configuration adjustment.

[0142] The actual interaction feedback information of the user on the final personalized configuration recommendation result is collected. The user's operation on the main recommendation scheme is recorded through the interactive device in the cockpit (such as touch screen, voice interaction system, etc.). If the user chooses to accept the main recommendation scheme, the acceptance operation can be recorded; if the user rejects the main recommendation scheme, the rejection operation can be recorded. At the same time, the user's selection operation on the alternative recommendation scheme is recorded to understand the user's preference among multiple recommendation schemes. In addition, the user's voice evaluation content after configuration adjustment is collected through the voice interaction system. These evaluation contents may include the user's satisfaction with the configuration scheme, opinions on certain device adjustments, etc.

[0143] Step S152: feedback effect quantification processing of the actual interaction feedback information, converting the acceptance operation into a positive feedback value, the rejection operation into a negative feedback value, and the voice evaluation content into an emotional tendency score.

[0144] The actual interaction feedback information collected is subjected to feedback effect quantification processing. For the user's acceptance operation on the main recommendation scheme, it is converted into a positive feedback value, which indicates the user's approval of the recommendation scheme. For the rejection operation, it is converted into a negative feedback value, indicating that the user is not satisfied with the recommendation scheme. For the voice evaluation content, it is converted into an emotional tendency score through semantic analysis. For example, if the user's voice evaluation contains positive words such as "very comfortable" and "very good", it can be converted into a higher emotional tendency score; if it contains negative words such as "uncomfortable" and "not good", it is converted into a lower emotional tendency score.

[0145] Step S153: extracting the configuration demand intention features corresponding to the main recommendation scheme and the alternative recommendation scheme from the final personalized configuration recommendation result, the configuration demand intention features including demand type identification, demand intensity level, and demand associated device.

[0146] The configuration demand intention features corresponding to the main recommendation scheme and the alternative recommendation scheme are extracted from the final personalized configuration recommendation result. These configuration demand intention features have been determined in the process of generating the configuration demand intention feature set in the foregoing, including demand type identification (such as temperature adjustment, seat adjustment, etc.), demand intensity level, and demand associated equipment (such as air conditioning system, seat adjustment system, etc.). By extracting these features, the demand intention corresponding to each recommendation scheme can be determined.

[0147] Step S154: Perform association storage processing on the configuration demand intention features and the feedback effect quantitative values to generate a target association record.

[0148] The extracted configuration demand intention features are associated with the values obtained after the feedback effect quantification processing and are stored. The configuration demand intention features of each recommendation scheme are associated with the corresponding positive feedback value, negative feedback value, or sentiment tendency score to form a target association record. The target association record records the relationship between each configuration demand intention and user feedback.

[0149] Step S155: Perform statistical analysis processing on the historically stored target association records to calculate the positive feedback rate and the negative feedback rate corresponding to each demand type identification.

[0150] The historically stored target association records are statistically analyzed. For each demand type identification, the number of positive feedback values and negative feedback values corresponding thereto are counted. The proportion of the number of positive feedback values in the total number of feedback values is calculated to obtain the positive feedback rate corresponding to the demand type identification, and the proportion of the number of negative feedback values in the total number of feedback values is calculated to obtain the negative feedback rate. For example, for the demand type identification of temperature adjustment, the number of positive feedback values and negative feedback values of all recommendation schemes related to temperature adjustment are counted, and then the positive feedback rate and the negative feedback rate are calculated.

[0151] Step S156: Adjust the semantic association rules between the demand intention entity nodes and the user behavior mode entity nodes in the knowledge association graph according to the positive feedback rate and the negative feedback rate, and the adjustment direction of the semantic association rules is to enhance the association rules with a positive feedback rate greater than a first feedback rate and weaken the association rules with a negative feedback rate greater than a second feedback rate.

[0152] The semantic association rules between the corresponding demand intention entity nodes and the user behavior mode entity nodes in the knowledge association graph are adjusted according to the calculated positive feedback rate and negative feedback rate. Two thresholds, the first feedback rate and the second feedback rate, are preset. If the corresponding positive feedback rate of a demand type identifier is greater than the first feedback rate, it indicates that the association between the demand intention and the user behavior mode is positive and effective, and the semantic association rules between the corresponding demand intention entity nodes and the user behavior mode entity nodes in the knowledge association graph can be enhanced, so that the above association is more easily identified and utilized in the subsequent semantic reasoning process. If the corresponding negative feedback rate of a demand type identifier is greater than the second feedback rate, it indicates that the association effect between the demand intention and the user behavior mode is poor, and the semantic association rules between the corresponding demand intention entity nodes and the user behavior mode entity nodes in the knowledge association graph can be weakened, and the dependence on the above association in the subsequent reasoning is reduced.

[0153] Step S157: Extracting the feature association strength values from the target association feature set, and calculating the adjustment coefficient of the feature association strength in combination with the feedback effect quantitative value. The adjustment coefficient is positively correlated with the positive feedback effect and negatively correlated with the negative feedback effect.

[0154] The feature association strength values are extracted from the target association feature set. These feature association strength values have been calculated in the process of generating the target association feature set. Then, the adjustment coefficient of the feature association strength is calculated in combination with the feedback effect quantitative value (positive feedback value, negative feedback value, sentiment tendency score, etc.). Since the adjustment coefficient is positively correlated with the positive feedback effect and negatively correlated with the negative feedback effect, when the feedback effect is positive, the adjustment coefficient will increase; when the feedback effect is negative, the adjustment coefficient will decrease. For example, if the user's feedback on a recommended solution is positive, the positive feedback value is high, then the adjustment coefficient of the feature association strength related to the recommended solution will increase; if the user's feedback is negative, the negative feedback value is high, then the adjustment coefficient will decrease.

[0155] Step S158: Using the adjustment coefficient to perform weighted update processing on the feature association strength parameters between the environmental features and the user features in the knowledge association graph, and performing storage replacement processing on the knowledge association graph after the weighted update.

[0156] The calculated adjustment coefficient is used to update the feature association strength parameter between each environmental feature and user feature in the knowledge association graph. The original feature association strength parameter is multiplied by the adjustment coefficient to obtain the updated feature association strength parameter. For example, the original feature association strength parameter between an environmental feature and a user feature is A, and the adjustment coefficient is B, so the updated feature association strength parameter is A multiplied by B. Through the above weighting update processing, the feature association strength parameter in the knowledge association graph can reflect the latest feedback information of the user. Finally, the weighted updated knowledge association graph is stored and replaced, and the updated knowledge association graph is used to replace the original knowledge association graph, so that more accurate and more user demand knowledge association graph can be used for semantic reasoning and configuration recommendation in the subsequent recommendation process.

[0157] Figure 2 A schematic diagram of exemplary hardware and software components of the personalized configuration recommendation system based on adaptive cockpit environment 100 that can implement the idea of the present application is shown. For example, the processor 120 can be used in the personalized configuration recommendation system based on adaptive cockpit environment 100 and used to perform the functions in the present application.

[0158] The personalized configuration recommendation system based on adaptive cockpit environment 100 can be a general server or a special-purpose server, both of which can be used to implement the personalized configuration recommendation method based on adaptive cockpit environment of the present application. Although only one server is shown in the present application, for the sake of convenience, the functions described in the present application can be implemented in a distributed manner on multiple similar platforms to balance the processing load.

[0159] For example, the personalized configuration recommendation system based on adaptive cockpit environment 100 can include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and different forms of storage media 140, such as a disk, a ROM, or a RAM, or any combination thereof. The personalized configuration recommendation system based on adaptive cockpit environment 100 can also include program instructions stored in the ROM, the RAM, or other types of non-transitory storage media, or any combination thereof. The method of the present application can be implemented according to these program instructions. The personalized configuration recommendation system based on adaptive cockpit environment 100 also includes an input / output (I / O) interface 150 between the computer and other input / output devices.

[0160] For the convenience of description, only one processor is described in the adaptive cockpit environment-based personalized configuration recommendation system 100. However, it should be noted that the adaptive cockpit environment-based personalized configuration recommendation system 100 in the present application can also include multiple processors, and therefore the steps performed by one processor described in the present application can also be jointly performed or separately performed by multiple processors. For example, if the processor of the adaptive cockpit environment-based personalized configuration recommendation system 100 performs steps A and B, it should be understood that steps A and B can also be jointly performed by two different processors or separately performed in one processor. For example, a first processor performs step A, a second processor performs step B, or the first processor and the second processor jointly perform steps A and B.

[0161] In addition, the embodiment of the present application further provides a readable storage medium, wherein computer executable instructions are preset, and when a processor executes the computer executable instructions, the adaptive cockpit environment-based personalized configuration recommendation method is realized.

[0162] It should be noted that, in order to simplify the description of the present application and to help the understanding of one or more embodiments of the present application, in the foregoing description of the embodiments of the present application, various features are sometimes combined into one embodiment, figure or description thereof.

Claims

1. A personalized configuration recommendation method based on an adaptive cockpit environment, characterized in that, The method includes: Acquire a real-time cockpit environment perception data set and a user behavior data set. The real-time cockpit environment perception data set includes physical environment parameter information collected by environmental monitoring equipment and cockpit equipment operating status information. The user behavior data set includes user historical configuration operation records, physiological signal data, and voice interaction content. The real-time perception data set of the cockpit environment and the user behavior data set are subjected to feature fusion processing to generate a target-related feature set, which includes a description of the correspondence between environmental state features and user behavior features. The pre-built knowledge association graph is invoked to perform semantic reasoning processing on the target association feature set to generate a configuration requirement intent feature set, which contains descriptions of the user's potential cockpit function configuration requirements. A set of candidate configuration schemes is generated based on the configuration requirement intent feature set, and the set of candidate configuration schemes is prioritized according to the feature association strength in the target association feature set to generate the final personalized configuration recommendation result. Based on the actual interactive feedback from users regarding the final personalized configuration recommendation results, update the semantic association rules and feature association strength parameters of the knowledge association graph; The step of invoking a pre-built knowledge graph to perform semantic reasoning processing on the target associated feature set, generating a configuration requirement intent feature set, includes: The environmental state statistical features in the target associated feature set are input into the environmental entity matching unit of the knowledge association graph to match the environmental parameter entity nodes stored in the knowledge association graph; The user operation habit features in the target associated feature set are input into the behavior entity matching unit of the knowledge association graph to match the user behavior pattern entity nodes stored in the knowledge association graph; The user physiological feedback features in the target associated feature set are input into the physiological entity matching unit of the knowledge association graph to match the physiological state entity nodes stored in the knowledge association graph. The user semantic demand features in the target association feature set are input into the demand entity matching unit of the knowledge association graph to match the demand intent entity nodes stored in the knowledge association graph. The relational reasoning module of the knowledge association graph is invoked to analyze the historical association between environmental parameter entity nodes and user behavior pattern entity nodes, analyze the causal relationship between user behavior pattern entity nodes and physiological state entity nodes, and analyze the mapping relationship between physiological state entity nodes and demand intention entity nodes. Based on the historical relationships, causal relationships and mapping relationships, multi-hop reasoning is performed to generate potential configuration needs intentions reflected by the user in the current environment state through historical operation behavior, physiological feedback and semantic needs. The potential configuration needs intentions include temperature adjustment priority, seat position adjustment direction, audio sound effect preference type and window opening degree suggested range. The potential configuration requirement intents are processed into a structured description to generate a set of configuration requirement intent features that include requirement type identifier, requirement intensity level, and requirement associated devices.

2. The method according to claim 1, characterized in that, The acquisition of the real-time perception data set of the cockpit environment and the user behavior data set includes: Physical environment parameter information is collected by temperature sensors, humidity sensors, light sensors and noise sensors installed in the cabin. The physical environment parameter information includes continuous time series of temperature measurement values, humidity measurement values, light intensity measurement values ​​and noise intensity measurement values. The cockpit equipment operating status information is collected by the status monitoring module of the cockpit equipment control system. The cockpit equipment operating status information includes the current working mode of the air conditioning system, the current position parameters of the seat adjustment system, the current volume level of the vehicle audio system, and the current opening and closing degree of the window control system. The user's historical configuration operation records are extracted from the cockpit interaction log database. These historical configuration operation records include timestamp information and adjustment range information of the user's adjustments to air conditioning temperature, seat angle, audio effects, and window opening degree. The user's physiological signal data is collected by biosensors installed in the cabin. The physiological signal data includes heart rate fluctuation sequence, skin conductance intensity sequence and respiratory rate change sequence. The user's voice interaction content is extracted through the dialogue recording storage module of the voice interaction system. The voice interaction content includes the user's explicit demand statements and implicit preference evaluation statements regarding the cabin environment. The physical environment parameter information, cockpit equipment operating status information, historical configuration operation records, physiological signal data, and voice interaction content are aligned according to timestamps to generate a real-time cockpit environment perception data set and a user behavior data set with time synchronization.

3. The method according to claim 1, characterized in that, The step of performing feature fusion processing on the real-time perception data set of the cockpit environment and the user behavior data set to generate a target-related feature set includes: The physical environment parameter information in the real-time sensing data set of the cabin environment is subjected to time window statistical processing to calculate the average temperature, humidity variance, median light intensity and maximum noise intensity within each time window, and to generate environmental state statistical features. The cabin equipment operating status information in the real-time cabin environment perception data set is processed by pattern encoding, the air conditioning system working mode is converted into discrete encoded values ​​of cooling / heating / ventilation, and the seat adjustment system position parameters are converted into continuous coordinate values ​​of front / back / up / down / angle, generating equipment status pattern features; Frequency analysis is performed on the historical configuration operation records in the user behavior data set to statistically analyze the frequency of user adjustments to the configuration of each cockpit device and the duration of each adjustment, thereby generating user operation habit characteristics. Feature extraction processing is performed on the physiological signal data in the user behavior dataset to calculate the peak-to-trough difference of heart rate fluctuations, the rising slope of skin conductance response, and the periodic interval of respiratory rate, thereby generating user physiological feedback features. Semantic parsing is performed on the voice interaction content in the user behavior data set to extract specific demand keywords from explicit demand statements and sentiment words from implicit preference evaluation statements to generate user semantic demand features. Feature association calculations are performed on the environmental state statistical features, device state pattern features, user operation habit features, user physiological feedback features, and user semantic demand features to determine the co-occurrence frequency and conditional probability between each pair of environmental features and user features, and to generate a target associated feature set containing feature association degree values.

4. The method according to claim 3, characterized in that, The frequency analysis of historical configuration operation records in the user behavior data set, and the statistical analysis of the frequency of user adjustments to the configuration of each cockpit device and the duration of each adjustment, include: Arrange the historical configuration operation records in chronological order to generate an operation time sequence; The time interval between two adjacent adjustments to the same device configuration is calculated based on the operation time sequence, and the number of adjustments per unit time is counted as the adjustment frequency. The start and end timestamps of each adjustment operation are extracted based on the operation time series, and the time difference is calculated as the duration of a single adjustment. The adjustment frequency and duration of a single adjustment for different equipment types were statistically analyzed to generate the adjustment frequency of the air conditioning system, seat adjustment system, car audio system, and window control system, as well as the corresponding distribution of the duration of a single adjustment. The adjustment frequency and duration of a single adjustment are standardized to generate standardized user operation habit characteristics.

5. The method according to claim 4, characterized in that, The process of calling the relationship reasoning module of the knowledge graph to analyze the historical association relationships between environmental parameter entity nodes and user behavior pattern entity nodes includes: Historical association edges between environmental parameter entity nodes and user behavior pattern entity nodes are extracted from the knowledge association graph. The historical association edges include the association count and association confidence. The number of times different environmental parameter values ​​and user behavior patterns co-occur is counted, and the conditional probability of environmental parameter values ​​and user behavior patterns is calculated. The conditional probability is the probability that a user will exhibit a certain behavior pattern under a set environmental parameter value. The strength of the historical association between environmental parameter entity nodes and user behavior pattern entity nodes is generated based on the co-occurrence frequency and conditional probability. The strength of the historical association is used as a weight parameter for relational reasoning, and is used for weight allocation in subsequent multi-hop reasoning processes.

6. The method according to claim 1, characterized in that, The process of generating a candidate configuration scheme set based on the configuration requirement intent feature set, and prioritizing the candidate configuration scheme set according to the feature association strength in the target association feature set to generate a final personalized configuration recommendation result includes: Extract the cabin equipment type corresponding to each requirement intent in the configuration requirement intent feature set, wherein the cabin equipment type includes air conditioning system, seat adjustment system, vehicle audio system and window control system; For each cockpit equipment type, retrieve historical valid configuration schemes related to that equipment type from the knowledge association graph. The historical valid configuration schemes include equipment parameter adjustment values, user feedback records after adjustment, and environmental adaptability records. Filter out historically valid configuration schemes that match the environmental status statistical features and user operation habit features in the feature set associated with the current target, and generate a set of candidate configuration sub-schemes for each device type; The candidate configuration sub-schemes for each device type are combined and processed to generate a candidate configuration scheme set that includes multi-device collaborative adjustment; Extract the feature association strength value corresponding to each demand intent from the target association feature set. The feature association strength value indicates the degree of association between the demand intent and the current environmental state and user behavior. Each candidate configuration scheme is assigned a weight parameter based on the feature association strength value, and the weight parameter is positively correlated with the feature association strength value of the demand intention. The candidate configuration scheme set is weighted and sorted. Candidate configuration schemes with weight parameters greater than a set threshold are retained first. Then, conflict detection is performed on the sorted candidate configuration scheme set to check whether there are conflicts between different adjustment parameters of the same device or conflicts between environmental influences of multiple device adjustments. Based on the inspection results, user adaptability verification is performed on the candidate configuration schemes without conflicts to obtain the candidate configuration schemes that pass the verification. The verification content includes whether the configuration adjustment range is in line with the user's historical operating habits and whether the physiological feedback prediction after the adjustment is better than the current state. The verified candidate configuration schemes are sorted in descending order according to the weight parameters to generate the final personalized configuration recommendation result, which includes the main recommended scheme and the alternative recommended schemes.

7. The method according to claim 6, characterized in that, The process of performing user compatibility verification on conflict-free candidate configuration schemes to obtain verified candidate configuration schemes includes: Extract the adjustment parameter values ​​of each device in the candidate configuration scheme. The adjustment parameter values ​​include air conditioning temperature adjustment value, seat angle adjustment value, audio volume adjustment value, and window opening degree adjustment value. Extract the user's preferred range of adjustment parameters for each device from the user's historical configuration operation records. The preferred range includes the minimum and maximum parameter values ​​that the user has historically adjusted. Verify whether the adjustment parameter values ​​of the candidate configuration schemes fall within the user's preference range, and generate parameter adaptability verification results; Extract physiological feedback features of the user after historical configuration adjustments from user physiological signal data. The physiological feedback features include heart rate fluctuation amplitude, skin conductance intensity and respiratory rate change rate. Predict the physiological feedback characteristics of users after the candidate configuration scheme is adjusted, compare the difference between the predicted value and the current physiological feedback characteristics, and generate physiological adaptability verification results. Based on the combined results of parameter compatibility verification and physiological compatibility verification, it is determined whether the candidate configuration scheme has passed the user compatibility verification. Candidate configuration schemes that have not passed the verification are removed from the recommendation results.

8. The method according to claim 1, characterized in that, The step of updating the semantic association rules and feature association strength parameters of the knowledge association graph based on the user's actual interaction feedback information on the final personalized configuration recommendation result includes: Collect actual interactive feedback information from users regarding the final personalized configuration recommendation results. The actual interactive feedback information includes user acceptance / rejection records of the main recommendation scheme, selection records of alternative recommendation schemes, and voice evaluation content after configuration adjustment. The actual interactive feedback information is quantified by converting the acceptance operation into a positive feedback value, the rejection operation into a negative feedback value, and the voice evaluation content into an emotional tendency score. From the final personalized configuration recommendation results, extract the configuration requirement intent features corresponding to the main recommendation scheme and the alternative recommendation scheme. The configuration requirement intent features include the requirement type identifier, the requirement intensity level, and the requirement-related devices. The configuration requirement intent features are associated with the feedback effect quantification value and stored together to generate a target association record. Perform statistical analysis on the historically stored target-related records to calculate the positive and negative feedback rates corresponding to each type of demand. The semantic association rules between the corresponding demand intent entity nodes and user behavior pattern entity nodes in the knowledge association graph are adjusted according to the positive feedback rate and the negative feedback rate. The adjustment direction of the semantic association rules is to strengthen the association rules with a positive feedback rate greater than the first feedback rate and weaken the association rules with a negative feedback rate greater than the second feedback rate. Extract the correlation strength values ​​of each feature from the target correlation feature set, and calculate the adjustment coefficient of the feature correlation strength by combining it with the feedback effect quantification value. The adjustment coefficient is positively correlated with the positive feedback effect and negatively correlated with the negative feedback effect. The adjustment coefficients are used to perform weighted update processing on the feature association strength parameters between each environmental feature and user feature in the knowledge association graph, and the weighted updated knowledge association graph is then stored and replaced.

9. A personalized configuration recommendation system based on an adaptive cockpit environment, characterized in that, The system includes a processor and a memory, the memory being connected to the processor. The memory is used to store programs, instructions, or code, and the processor is used to execute the programs, instructions, or code in the memory to implement the personalized configuration recommendation method based on the adaptive cockpit environment as described in any one of claims 1-8.

Images