CN117979518A

CN117979518A - Control method, device, equipment and storage medium for vehicle atmosphere lamp

Info

Publication number: CN117979518A
Application number: CN202410366155.0A
Authority: CN
Inventors: 陈敏毅
Original assignee: E Link Technology Co ltd
Current assignee: E Link Technology Co ltd
Priority date: 2024-03-28
Filing date: 2024-03-28
Publication date: 2024-05-03
Anticipated expiration: 2044-03-28
Also published as: CN117979518B

Abstract

The application relates to the technical field of deep learning, and discloses a control method, a device, equipment and a storage medium of a vehicle atmosphere lamp. The method comprises the following steps: acquiring passenger behavior data of a target vehicle and extracting behavior characteristics to obtain the behavior characteristics of the passenger; analyzing the control parameters of the atmosphere lamp of the vehicle through an RNN-PSO algorithm to obtain a first control parameter set; acquiring internal environment characteristics of a vehicle and external environment characteristics of the vehicle; performing vehicle atmosphere lamp action analysis through a dual-depth Q network to obtain a first vehicle atmosphere lamp action execution strategy; performing vehicle atmosphere lamp control, acquiring passenger feedback data of a target vehicle, and performing rewarding feedback data calculation through the passenger feedback data to obtain rewarding feedback parameters; and performing parameter optimization to obtain a second control parameter set, and generating a corresponding second action execution strategy according to the second control parameter set.

Description

Control method, device, equipment and storage medium for vehicle atmosphere lamp

Technical Field

The application relates to the technical field of deep learning, in particular to a control method, a device, equipment and a storage medium of a vehicle atmosphere lamp.

Background

In the current development of the automotive industry, research and development of intelligent vehicle systems is becoming a hotspot field, where vehicle mood light control systems are receiving a great deal of attention as an important component for enhancing the riding experience of passengers. The atmosphere lamp not only can improve the inner environment of the vehicle and increase the beauty and comfort, but also can influence the emotion and the safety of passengers to a certain extent. However, existing vehicle mood light control methods are mostly static or based on automatic adjustment of simple logic only, lacking in response and adaptability to individual demands of passengers.

With the advancement of artificial intelligence technology, particularly the successful application of machine learning and deep learning algorithms in various fields, a new solution is provided for intelligent vehicle atmosphere lamp control. By analyzing the behavior data of passengers and the internal and external environment data of the vehicle and combining advanced algorithms to process and learn the data, more intelligent and personalized atmosphere lamp control can be realized. However, how to efficiently integrate and process such complex data information, and how to design efficient algorithmic models to support real-time and accurate control decisions, remains a significant challenge in current research. In addition, the preferences of passengers for vehicle mood lights may vary over time, with environmental and personal mood changes, requiring that the vehicle mood light control system be capable of learning and adapting to continuously optimize control strategies to meet the dynamic needs of passengers. However, the prior art lacks an efficient mechanism to capture occupant feedback and a systematic approach to adjust and optimize the mood light control strategy based on these feedback.

Disclosure of Invention

The application provides a control method, a device, equipment and a storage medium for a vehicle atmosphere lamp, which are used for improving the control accuracy of the vehicle atmosphere lamp.

In a first aspect, the present application provides a control method of a vehicle atmosphere lamp, the control method of the vehicle atmosphere lamp comprising:

acquiring passenger behavior data of a target vehicle and extracting behavior characteristics to obtain the behavior characteristics of the passenger;

inputting the passenger behavior characteristics into an RNN-PSO algorithm to analyze the control parameters of the atmosphere lamp of the vehicle, and obtaining a first control parameter set;

Acquiring vehicle internal environment data and vehicle external environment data of the target vehicle, and respectively analyzing the environment characteristics to obtain vehicle internal environment characteristics and vehicle external environment characteristics;

inputting the first control parameter set, the vehicle internal environment characteristics and the vehicle external environment characteristics into a dual-depth Q network to perform vehicle atmosphere lamp execution action analysis, so as to obtain a first vehicle atmosphere lamp action execution strategy;

Performing vehicle atmosphere lamp control through the first vehicle atmosphere lamp action execution strategy, acquiring passenger feedback data of the target vehicle, and performing rewarding feedback data calculation through the passenger feedback data to obtain rewarding feedback parameters;

and carrying out parameter optimization on the first control parameter set according to the reward feedback parameter to obtain a second control parameter set, and generating a corresponding second action execution strategy according to the second control parameter set.

In a second aspect, the present application provides a control device for a vehicle atmosphere lamp, the control device comprising:

the acquisition module is used for acquiring the passenger behavior data of the target vehicle and extracting behavior characteristics to obtain the passenger behavior characteristics;

The analysis module is used for inputting the passenger behavior characteristics into an RNN-PSO algorithm to analyze the control parameters of the atmosphere lamp of the vehicle, so as to obtain a first control parameter set;

The analysis module is used for acquiring the vehicle internal environment data and the vehicle external environment data of the target vehicle and respectively analyzing the environment characteristics to obtain the vehicle internal environment characteristics and the vehicle external environment characteristics;

The processing module is used for inputting the first control parameter set, the vehicle internal environment characteristics and the vehicle external environment characteristics into a dual-depth Q network to perform vehicle atmosphere lamp execution action analysis, so as to obtain a first vehicle atmosphere lamp action execution strategy;

The calculation module is used for controlling the vehicle atmosphere lamp according to the first vehicle atmosphere lamp action execution strategy, acquiring passenger feedback data of the target vehicle, and calculating reward feedback data according to the passenger feedback data to obtain reward feedback parameters;

The generating module is used for carrying out parameter optimization on the first control parameter set according to the rewarding feedback parameters to obtain a second control parameter set, and generating a corresponding second action executing strategy according to the second control parameter set.

A third aspect of the present application provides a control apparatus of a vehicle atmosphere lamp, comprising: a memory and at least one processor, the memory having instructions stored therein; the at least one processor invokes the instructions in the memory to cause the control device of the vehicle atmosphere lamp to execute the control method of the vehicle atmosphere lamp described above.

A fourth aspect of the present application provides a computer-readable storage medium having instructions stored therein, which when run on a computer, cause the computer to perform the above-described method of controlling a vehicle atmosphere lamp.

According to the technical scheme provided by the application, the specific requirements and preferences of the passengers can be identified by acquiring and analyzing the behavior data of the passengers of the target vehicle, and the setting of the atmosphere lamp is further adjusted to match the moods and activities of the passengers. Such personalized lighting control not only improves passenger comfort and satisfaction, but also increases the added value of the vehicle. By combining an RNN-PSO algorithm and a dual-depth Q network, the control method can intelligently analyze the behavior characteristics of passengers and the internal and external environment data of the vehicle, automatically optimize control parameters and adjust atmosphere lamp settings in real time. The intelligent self-adaptive control strategy can ensure the optimal in-vehicle atmosphere under different conditions, and the experience of passengers is improved. And the feedback of the passengers to the atmosphere lamp setting is collected, and the rewarding feedback parameters are obtained according to the feedback calculation, so that the control parameters are further optimized. The feedback-based dynamic learning and optimizing mechanism can continuously improve the atmosphere lamp control strategy, ensure that the system continuously evolves along with the time and better meet the changing demands of passengers. The intelligent adjustment of the brightness and the color of the atmosphere lamp according to the internal and external environments of the vehicle and the behaviors of passengers is beneficial to optimizing the energy consumption. Compared with the traditional manual regulation or fixed mode, the intelligent control method can more effectively use energy, and unnecessary energy waste is reduced. Through considering vehicle external environment characteristic when adjusting the atmosphere lamp, can ensure the vision comfort level and the security of passenger under different driving environment, say at night or rainy day automatically regulated atmosphere lamp luminance, reduce the interference to driver's sight, promote whole driving safety, and then improved the control accuracy of vehicle atmosphere lamp.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings may be obtained based on these drawings without inventive effort for a person skilled in the art.

FIG. 1 is a schematic view of an embodiment of a control method of a vehicle atmosphere lamp according to the present application;

fig. 2 is a schematic diagram of an embodiment of a control device for a vehicle atmosphere lamp according to an embodiment of the present application.

Detailed Description

The embodiment of the application provides a control method, a device, equipment and a storage medium for a vehicle atmosphere lamp. The terms "first," "second," "third," "fourth" and the like in the description and in the claims and in the above drawings, if any, are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments described herein may be implemented in other sequences than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed or inherent to such process, method, article, or apparatus.

For ease of understanding, a specific flow of an embodiment of the present application is described below with reference to fig. 1, and an embodiment of a method for controlling an atmosphere lamp of a vehicle according to the embodiment of the present application includes:

Step S101, acquiring passenger behavior data of a target vehicle and extracting behavior characteristics to obtain passenger behavior characteristics;

It is to be understood that the execution body of the present application may be a control device of a vehicle atmosphere lamp, and may also be a terminal or a server, which is not limited herein. The embodiment of the application is described by taking a server as an execution main body as an example.

Specifically, image data of at least one target passenger is acquired through an image sensor of a device in the target vehicle, and meanwhile, atmosphere lamp interaction data of the passenger is acquired through an atmosphere lamp interaction sensor in the vehicle. And analyzing the passenger image data, and analyzing the sitting postures of the passengers and the activity frequency of the passengers to obtain passenger sitting posture data and activity frequency data. Through the recognition of the body gestures of the passengers and the quantification of the frequent degree of the activities of the passengers, the behavior habits and the activity characteristics of the passengers are obtained. Analyzing the interaction data of the atmosphere lamp to analyze the behavior intention of the passengers and generating behavior intention data. The motivation of the passenger to interact with the mood light is understood, such as to adjust the brightness of the light or change color, reflecting the passenger's needs and preferences. And combining the behavior intention data with the sitting posture data and the activity frequency data of the passengers to construct more comprehensive behavior data of the passengers. And classifying the passenger behavior data through a cluster analysis model, and identifying different passenger behavior categories. By grouping the behavior data, an understanding of the diversity and commonality of passenger behavior is facilitated. Principal component feature analysis is performed on each behavior category, identifying the initial behavior features of each category, which helps to accurately distinguish between different behavior patterns and habits. And comprehensively analyzing and fusing the initial behavior characteristics of each passenger behavior category through a characteristic fusion technology to obtain final passenger behavior characteristics.

Step S102, inputting the behavior characteristics of passengers into an RNN-PSO algorithm to analyze the control parameters of the atmosphere lamps of the vehicle, and obtaining a first control parameter set;

Specifically, the passenger behavior characteristics are input into an RNN-PSO algorithm, and the vehicle atmosphere lamp control parameters are deeply analyzed. The RNN-PSO algorithm includes a recurrent neural network and a particle swarm optimization algorithm, wherein the RNN specifically refers to a two-layer structure containing 256 gated loop units (GRUs) therein. This structural design enables the algorithm to effectively capture the characteristics of time series data, namely the dynamic characteristics of the passenger behavior over time. At the first floor of the RNN, 256 GRU units are responsible for preliminary hidden state feature extraction of the entered passenger behavior features. By analyzing the time dependence and pattern variations in the passenger behavior data, a more advanced feature representation is extracted. The 256 GRU units of the second tier further process these hidden state features, and predict control parameters of the vehicle mood lights based on the features, which determine the brightness, color, mode, etc. of the mood lights. The algorithm controls the parameter selection process through the particle swarm optimization algorithm. The PSO algorithm constructs a population of particles based on a plurality of predicted control parameter values according to a set of numerical control ranges for the target vehicle. Each particle represents a set of possible control parameter values. And the contribution of each particle in the particle population to the effect of realizing the ideal atmosphere lamp is evaluated by calculating the fitness of each particle in the particle population, and the PSO algorithm can guide the particles to move towards the direction with higher fitness. With the progress of iterative computation, the algorithm continuously adjusts the position of the particles, i.e., the control parameter values, to find the optimal solution. In the process, the combination of fitness calculation and iterative calculation ensures that the algorithm can find a control parameter set which is most suitable for the current passenger behavior characteristics in a complex parameter space. When the iterative calculation meets a preset condition, such as reaching a predetermined number of iterations or the fitness improvement is no longer significant, the algorithm terminates the iteration, and the particle position at this time represents the optimal control parameter value. And obtaining a final first control parameter set by correcting and converting the control parameters of the optimal solution. This set reflects the deep understanding and sophisticated control requirements of the passenger behavior feature analysis, enabling the vehicle mood lights to respond to the passenger's behavior and preferences in the most appropriate manner, thereby enhancing the passenger's experience and satisfaction.

Step S103, acquiring vehicle internal environment data and vehicle external environment data of a target vehicle, and respectively analyzing the environment characteristics to obtain vehicle internal environment characteristics and vehicle external environment characteristics;

Specifically, internal environmental data, such as the temperature and the illumination intensity in the vehicle cabin, and external environmental data, including weather conditions and road condition information, are acquired from a sensing system of the vehicle. And carrying out data standardization processing to remove possible deviation and inconsistency in the data acquisition process, ensure the accuracy and comparability of the data and obtain standardized in-vehicle and-out environment data. And classifying the standardized data to respectively identify key indexes of the in-vehicle environment, such as temperature and illumination intensity, and key factors of the out-vehicle environment, such as weather conditions and road condition information. And carrying out average value operation on the temperature and the illumination intensity in the environment data in the vehicle to respectively obtain the average temperature and the average illumination intensity in the vehicle. These two averages reflect the general state of the vehicle interior environment and provide a simplified and effective indicator for understanding the vehicle interior environment. And carrying out correlation analysis on the average value, and exploring whether a certain correlation exists between the temperature and the illumination intensity, wherein the analysis is helpful for revealing the interaction between the vehicle internal environment factors, and further carrying out feature extraction and fusion on the two factors according to the analysis result, so as to finally obtain the comprehensive index reflecting the vehicle internal environment features. And for the vehicle exterior environment data, the correlation between the weather conditions and the road condition information is obtained through correlation analysis, and the two external factors are understood to influence the running and riding experience of the vehicle together. Based on the analysis result, feature extraction and fusion are carried out to obtain an index comprehensively reflecting the external environment features of the vehicle.

Step S104, inputting the first control parameter set, the internal environment characteristics of the vehicle and the external environment characteristics of the vehicle into a dual-depth Q network to perform vehicle atmosphere lamp execution action analysis, and obtaining a first vehicle atmosphere lamp action execution strategy;

Specifically, the first control parameter set, the internal environment feature of the vehicle and the external environment feature of the vehicle are input into a dual-depth Q network, so as to accurately analyze and make a suitable vehicle atmosphere lamp execution action strategy, and the dual-depth Q network comprises: an input layer, a first deep neural network, and a second deep neural network. The first set of control parameters, the vehicle interior environmental characteristics, and the vehicle exterior environmental characteristics are encoded by the input layer. The encoded vectors are spliced into a unified input feature encoded vector. The input feature encoding vector is input into a first deep neural network specifically designed to predict potential performing actions of the vehicle mood lights. The first deep neural network is capable of outputting a series of initial vehicle atmosphere lamp action execution strategies by learning complex relationships between control parameters and environmental features. These strategies reflect a selection of actions that may achieve optimal ambiance effects given the environmental and control parameters. To ensure the validity and optimization of these policies, the dual deep Q network further evaluates and optimizes the initial action execution policies output by the first deep neural network through the second deep neural network. Through action value evaluation, the second deep neural network can identify action execution strategies which possibly bring about the best atmosphere effect in actual execution, and then generate a final first vehicle atmosphere lamp action execution strategy.

Step 105, performing vehicle atmosphere lamp control through a first vehicle atmosphere lamp action execution strategy, acquiring passenger feedback data of a target vehicle, and performing rewarding feedback data calculation through the passenger feedback data to obtain rewarding feedback parameters;

Specifically, the atmosphere lamp of the target vehicle is intelligently controlled according to the first vehicle atmosphere lamp action execution strategies, and the strategies aim to adjust the illumination intensity, the color, the mode and the like in the vehicle based on the results obtained through the dual-depth Q network analysis so as to create the optimal driving atmosphere. In order to evaluate the actual effect of these action execution strategies and passenger satisfaction, feedback data of the passenger is acquired through devices such as an image sensor and a microphone. The image sensor captures the expression and body language of the passenger and obtains feedback image data that can reflect the immediate reaction and emotional state of the passenger to the current mood light setting. At the same time, the microphone collects feedback voice data of the passengers, including their possible verbal comments and sensory expressions, which provide information on the passenger satisfaction. A first feedback index is generated from the feedback image data, and the index evaluates the satisfaction degree of passengers on the effect of the atmosphere lamp by analyzing the expression change and the body language of the passengers. And meanwhile, generating a second feedback index according to the feedback voice data, and analyzing the oral feedback of the passengers through voice recognition and emotion analysis technology to obtain quantitative evaluation of the influence of atmosphere lamp setting on the emotion of the passengers. And integrating the two feedback indexes to calculate the reward feedback data, thereby obtaining the reward feedback parameter. The multidimensional information fed back by the passengers is considered in the calculation process, and the rewarding feedback parameters obtained through algorithm analysis directly reflect the influence degree of the atmosphere lamp setting on the satisfaction degree of the passengers, so that a basis is provided for further optimizing the atmosphere lamp control strategy.

And S106, carrying out parameter optimization on the first control parameter set according to the rewarding feedback parameters to obtain a second control parameter set, and generating a corresponding second action execution strategy according to the second control parameter set.

Specifically, parameter optimization is performed on the first control parameter set according to the reward feedback parameters, and the first control parameter set is adjusted through an algorithm and an optimization technology to obtain an optimized second control parameter set. And (3) carrying out strategy gradient analysis on the second control parameter set, and evaluating the contribution degree of each parameter to the final control effect and how to influence the improvement of the satisfaction degree of the atmosphere lamp to the passengers by using a strategy gradient method. The policy gradient analysis helps determine in which directions adjusting the control parameters may lead to better ambiance effects, thereby providing directions and basis for policy updates. Based on the strategy gradient, calculating strategy updating parameters of the reward feedback parameters, and adjusting the original reward feedback parameters according to the result of the strategy gradient to generate new strategy updating parameters. These updated parameters reflect how to adjust the control strategy to maximize passenger satisfaction, which is derived from an assessment of the current control effect and an analysis of how to improve this effect by adjusting the control parameters. And carrying out strategy updating on the second control parameter set based on the strategy updating parameters to generate a corresponding second action executing strategy. The optimized control parameters and the information obtained through the strategy gradient analysis are applied to the actual atmosphere lamp control, so that the atmosphere lamp is ensured to be adjusted more finely and the requirements and preferences of passengers can be met better. By the method, the vehicle atmosphere lamp control system can perform self optimization according to feedback of passengers and continuously learn and adapt to changing requirements of the passengers, so that riding experience is improved, and meanwhile intelligence and individuation capacity of the system are enhanced. The iterative optimization process based on feedback embodies the advanced capability of the intelligent control system in real-time response and continuous improvement, and ensures that the atmosphere lamp setting can enhance the driving experience of passengers in the most appropriate mode.

According to the embodiment of the application, the specific requirements and preferences of the passengers can be identified by acquiring and analyzing the behavior data of the passengers of the target vehicle, and the setting of the atmosphere lamp is further adjusted to match the moods and activities of the passengers. Such personalized lighting control not only improves passenger comfort and satisfaction, but also increases the added value of the vehicle. By combining an RNN-PSO algorithm and a dual-depth Q network, the control method can intelligently analyze the behavior characteristics of passengers and the internal and external environment data of the vehicle, automatically optimize control parameters and adjust atmosphere lamp settings in real time. The intelligent self-adaptive control strategy can ensure the optimal in-vehicle atmosphere under different conditions, and the experience of passengers is improved. And the feedback of the passengers to the atmosphere lamp setting is collected, and the rewarding feedback parameters are obtained according to the feedback calculation, so that the control parameters are further optimized. The feedback-based dynamic learning and optimizing mechanism can continuously improve the atmosphere lamp control strategy, ensure that the system continuously evolves along with the time and better meet the changing demands of passengers. The intelligent adjustment of the brightness and the color of the atmosphere lamp according to the internal and external environments of the vehicle and the behaviors of passengers is beneficial to optimizing the energy consumption. Compared with the traditional manual regulation or fixed mode, the intelligent control method can more effectively use energy, and unnecessary energy waste is reduced. Through considering vehicle external environment characteristic when adjusting the atmosphere lamp, can ensure the vision comfort level and the security of passenger under different driving environment, say at night or rainy day automatically regulated atmosphere lamp luminance, reduce the interference to driver's sight, promote whole driving safety, and then improved the control accuracy of vehicle atmosphere lamp.

In a specific embodiment, the process of executing step S101 may specifically include the following steps:

(1) Acquiring passenger image data of at least one target passenger through an image sensor in the target vehicle, and acquiring atmosphere lamp interaction data of at least one target passenger through an atmosphere lamp interaction sensor in the target vehicle;

(2) Analyzing the passenger sitting posture and the activity frequency of the passenger image data to obtain passenger sitting posture data and activity frequency data;

(3) Analyzing the passenger behavior intention of the atmosphere lamp interaction data to obtain behavior intention data, and generating passenger behavior data of the target vehicle according to the behavior intention data, the passenger sitting posture data and the activity frequency data;

(4) Performing behavior category analysis on the passenger behavior data through a cluster analysis model to obtain a plurality of passenger behavior categories;

(5) Performing principal component feature analysis on the passenger behavior data according to the plurality of passenger behavior categories to obtain initial behavior features of each passenger behavior category;

(6) And carrying out feature fusion on the initial behavior features of each passenger behavior category to obtain the passenger behavior features.

Specifically, passenger image data of at least one target passenger is acquired through an image sensor in the target vehicle, and meanwhile, the atmosphere lamp interaction sensor records relevant data of interaction between the passenger and the atmosphere lamp, such as actions of adjusting light brightness or color. The passenger image data is analyzed to understand the sitting posture and the activity frequency of the passenger. By applying image recognition and behavior analysis techniques, passenger sitting posture information, such as sitting, tilting or other specific gestures, is extracted from the image data while the frequency of passenger activity, such as frequent movements or remaining stationary, is assessed. The comfort and current activity status of the passenger are revealed by their physical status and frequency of behavior. Then, the data obtained from the atmosphere lamp interaction sensor is analyzed, and the behavior intention of the passenger is analyzed. The preferences and demands of the passengers are deduced by considering specific actions of the passengers interacting with the atmosphere lights, such as adjusting the light intensity or changing the light color. And analyzing the comprehensive passenger behavior data through a cluster analysis model, and classifying the passenger behaviors into a plurality of categories. The similarity and variability of behavior patterns is automatically identified by algorithms, which categorize passenger behavior into several typical categories, such as patterns of behavior seeking relaxation, active communication, or need to be focused. Such classification helps the system more accurately identify the behavioral tendencies and needs of the passenger. A principal component feature analysis is applied to each passenger behavior category, and the behavior features of the core are extracted from each behavior category. Feature variables capable of most effectively representing each behavior category are identified through a dimension reduction technology, so that complexity of behavior data is simplified, and the most critical information is reserved. And carrying out feature fusion on the initial behavior features of each behavior category, and generating final behavior features of the passengers by comprehensively considering sitting posture data, activity frequency data and behavior intention data of the passengers. By integrating data and analysis results of different dimensions, a comprehensive behavior feature model which comprehensively reflects the behavior trend and the demand of passengers is formed.

In a specific embodiment, the process of executing step S102 may specifically include the following steps:

(1) Inputting the passenger behavior characteristics into an RNN-PSO algorithm, wherein the RNN-PSO algorithm comprises a cyclic neural network and a particle swarm optimization algorithm;

(2) Extracting hidden state features from passenger behavior features through 256 GRU units in a first layer in the cyclic neural network to obtain hidden state features;

(3) Carrying out vehicle atmosphere lamp control parameter prediction on the hidden state characteristics through 256 GRU units in a second layer of the cyclic neural network to obtain a plurality of prediction control parameter values;

(4) Acquiring a numerical control range set of a target vehicle, and constructing a particle population for a plurality of predicted control parameter values through a particle swarm optimization algorithm to obtain the particle population;

(5) Performing particle fitness calculation on the particle population to obtain a particle fitness set corresponding to the particle population, and performing iterative calculation on the particle fitness set until a preset condition is met, so as to generate an optimal solution corresponding to the particle population;

(6) And carrying out control parameter correction and integrated conversion on the plurality of predicted control parameter values through an optimal solution to obtain a first control parameter set.

Specifically, the passenger behavior characteristics are input to an RNN-PSO algorithm, which includes a recurrent neural network and a particle swarm optimization algorithm. The hidden state feature extraction is carried out on the passenger behavior features through 256 GRU units in the first layer in the cyclic neural network, and the characteristics of the sequence data, such as time dependence and long-term dependence, are processed. The GRU units effectively capture and maintain time-series information in the passenger behavior data through their internal mechanisms so that the extraction of hidden state features becomes more accurate. And predicting the vehicle atmosphere lamp control parameters of the hidden state features through 256 GRU units of a second layer in the cyclic neural network. According to the deep information extracted from the passenger behavior characteristics, atmosphere lamp settings, such as brightness, color or mode of light, etc., which can improve the passenger experience are predicted. The predicted values of the plurality of control parameters are then used to direct the adjustment of the atmosphere lamp in order to achieve the optimal atmosphere effect. And acquiring a numerical control range set of the target vehicle, wherein the set defines a possible value range of the atmosphere lamp control parameters, and ensures that the adjustment of the control parameters is carried out within the actual physical limit. And optimizing the control parameter value obtained through RNN prediction according to a PSO algorithm. The PSO algorithm builds a population of particles by modeling the behavior of a bird population looking for food, each particle representing a set of possible control parameter values. And calculating particle fitness of the particle population, wherein the fitness reflects the effectiveness of the set of control parameters in achieving the desired atmosphere effect. The fitness calculation is based on a predefined objective function quantifying the relation between the mood light setting and the passenger satisfaction. The PSO algorithm continuously updates the position of the particle (i.e., the control parameter value) in the iterative calculation process to find the optimal solution, and the process is continued until a preset stopping condition is met, such as the maximum number of iterations is reached or the fitness improvement is below a certain threshold. Finally, when the PSO algorithm finds the optimal solution, this solution represents an atmosphere lamp control parameter set that can maximize passenger satisfaction under the current passenger behavior characteristics. And correcting and integrating the original predicted control parameter values through the optimal solution to generate a final first control parameter set.

In a specific embodiment, the process of executing step S103 may specifically include the following steps:

(1) Acquiring vehicle interior environment data and vehicle exterior environment data of a target vehicle, and performing data standardization processing on the vehicle interior environment data and the vehicle exterior environment data to obtain standard vehicle interior environment data and standard vehicle exterior environment data;

(2) Classifying the data set of the standard in-vehicle environment data to obtain in-vehicle temperature and in-vehicle illumination intensity, and classifying the data set of the standard in-vehicle environment data to obtain weather conditions and road condition information;

(3) Carrying out average value operation on the temperature in the vehicle to obtain average temperature, and carrying out average value operation on the illumination intensity in the vehicle to obtain average illumination intensity;

(4) Carrying out correlation analysis on the average temperature and the average illumination intensity to obtain a first correlation analysis result, and carrying out feature extraction and feature fusion on the average temperature and the average illumination intensity according to the first correlation analysis result to obtain the internal environment feature of the vehicle;

(5) And carrying out correlation analysis on the weather condition and the road condition information to obtain a second correlation analysis result, and carrying out feature extraction and feature fusion on the weather condition and the road condition information according to the second correlation analysis result to obtain the external environment features of the vehicle.

Specifically, vehicle interior and exterior environmental data is collected from a plurality of sensors and external data sources of a target vehicle. These data sources may include temperature and lighting sensors built into the vehicle for collecting in-vehicle environmental data, and external data services connected to the vehicle information system for providing real-time weather conditions and road condition information. After collecting these data, a data normalization process is performed to ensure that data from different sources and sensors can be compared and analyzed on the same basis. The normalization process involves converting various data into a uniform format and unit of measure, while handling any missing or outliers to ensure accuracy and consistency of the data. And classifying the standardized data, namely subdividing the in-vehicle environment data into two major categories of in-vehicle temperature and illumination intensity, and classifying the in-vehicle environment data into weather conditions and road condition information. The complex environmental data sets are organized into more ordered and easily analyzed structures by predefined classification rules and algorithms. And (3) carrying out average value operation on the temperature and the illumination intensity in the vehicle, and calculating the average temperature and the average illumination intensity representing the general condition of the environment in the vehicle. And carrying out correlation analysis on the average temperature and the average illumination intensity to obtain whether a remarkable correlation exists between the average temperature and the average illumination intensity, and how the correlation affects the comfort of passengers and the regulation strategy of the environment in the vehicle. The result of the correlation analysis will reveal the interaction of temperature and illumination intensity in the in-vehicle environment, providing basis for further feature extraction and fusion. And based on the first correlation analysis result, carrying out feature extraction and fusion, and fusing the data of the average temperature and the average illumination intensity into a comprehensive vehicle internal environment feature. Features of the most representative and informative quantities are extracted from the raw data using statistical and machine learning techniques, and then combined into a single, comprehensive representation reflecting the overall condition of the in-vehicle environment. Meanwhile, correlation analysis is carried out on weather conditions and road condition information, and the mutual influence between the two external environment factors is explored, and how the external environment factors act on the vehicle running and the passenger experience together. The second correlation analysis results reveal correlations between external environmental conditions, providing a basis for feature extraction and fusion. And according to the analysis result, extracting and fusing the characteristics of the weather condition and the road condition information to generate a comprehensive external environment characteristic of the vehicle.

In a specific embodiment, the process of executing step S104 may specifically include the following steps:

(1) Inputting the first set of control parameters, the vehicle interior environmental characteristics, and the vehicle exterior environmental characteristics into a dual depth Q network, the dual depth Q network comprising: an input layer, a first deep neural network, and a second deep neural network;

(2) The method comprises the steps of performing control parameter coding on a first control parameter set through an input layer to obtain a control parameter coding vector, performing environment characteristic coding on vehicle internal environment characteristics and vehicle external environment characteristics to obtain an in-vehicle environment characteristic coding vector and an out-vehicle environment characteristic coding vector, and performing coding vector splicing on the control parameter coding vector, the in-vehicle environment characteristic coding vector and the out-vehicle environment characteristic coding vector to obtain an input characteristic coding vector;

(3) Inputting the input feature code vector into a first depth neural network to perform vehicle atmosphere lamp execution action prediction, so as to obtain an initial vehicle atmosphere lamp action execution strategy;

(4) And performing action value evaluation on the initial vehicle atmosphere lamp action execution strategy through the second deep neural network to obtain a first vehicle atmosphere lamp action execution strategy.

Specifically, a first set of control parameters, a vehicle interior environmental characteristic, and a vehicle exterior environmental characteristic are input into a dual depth Q network. The first set of control parameters is encoded and these parameters are converted into control parameter encoded vectors. Environmental features inside and outside the vehicle are also converted into an in-vehicle environmental feature code vector and an out-of-vehicle environmental feature code vector through the encoding process. By adopting a specific coding algorithm through the input layer, it is ensured that all input data can be effectively converted into a format which can be processed by the neural network. And splicing the three coding vectors to form a comprehensive input characteristic coding vector. The input feature encoding vector is input to a first deep neural network. According to the current control parameters and environmental characteristics, a series of possible vehicle atmosphere lamp execution actions are predicted. Through the learning and training of a large amount of historical data, the first deep neural network can identify which execution actions are most likely to improve the comfort and satisfaction of passengers, and a plurality of initial vehicle atmosphere lamp action execution strategies are generated. Thereafter, in order to further optimize these initial strategies, ensuring that they are able to achieve the best results in practical applications, the initial action execution strategies are evaluated and screened by the second deep neural network. Through the action value evaluation mechanism, the possible effects of each strategy are quantitatively analyzed, so that the most effective execution strategies in improving the experience of passengers are identified. And finally, obtaining a first vehicle atmosphere lamp action execution strategy. These strategies are not only based on current vehicle control parameters and environmental characteristics, but also have been subjected to deep analysis and optimization of the two-layer neural network, thereby ensuring their effectiveness and practicality.

In a specific embodiment, the process of executing step S105 may specifically include the following steps:

(1) Performing vehicle atmosphere lamp control on the target vehicle through a first vehicle atmosphere lamp action execution strategy;

(2) Acquiring feedback image data of at least one target passenger through an image sensor, and acquiring feedback voice data of at least one target passenger;

(3) Generating a first feedback index according to the feedback image data, and generating a second feedback index according to the feedback voice data;

(4) And calculating the reward feedback data through the first feedback index and the second feedback index to obtain the reward feedback parameter.

Specifically, the atmosphere lamp of the target vehicle is intelligently adjusted according to the first vehicle atmosphere lamp action execution strategy. The strategy is based on comprehensive analysis of passenger behavior characteristics and vehicle internal and external environment characteristics, and aims to create the most suitable vehicle internal environment by adjusting parameters such as brightness, color, mode and the like of atmosphere lamps. After this strategy is implemented, real-time feedback of the passengers is collected in order to evaluate their effect and further optimize the control logic. Image data of at least one target passenger is captured by an image sensor installed inside the vehicle, while voice feedback data of the same passenger is acquired using a microphone device in the vehicle. These two data types capture the passenger's response and feel to the current mood light setting from different dimensions. And analyzing the feedback image data, extracting information such as the expression and the gesture of the passenger from the feedback image data by utilizing an image processing and face recognition technology, and generating a first feedback index. This indicator can reflect the immediate mood and satisfaction of the passenger, such as by identifying the passenger's smile, eye lock, or other change in expression, inferring the passenger's acceptance of the current mood setting. And meanwhile, processing and analyzing the collected feedback voice data, extracting emotion and satisfaction information in the feedback voice data by adopting voice recognition and emotion analysis technology, and generating a second feedback index. This index provides the system with passenger satisfaction information of another dimension based on the speech characteristics of the passenger, such as speech tonality, speech speed, pauses, and possibly a direct verbal evaluation. And integrating the two feedback indexes to calculate the rewarding feedback data. The calculation process extracts key information from feedback indexes of two dimensions through data analysis and a machine learning algorithm, comprehensively evaluates the actual effect of the atmosphere lamp control strategy, and calculates a comprehensive rewarding feedback parameter which reflects the overall influence of the current atmosphere lamp setting on the satisfaction degree of passengers. The reward feedback parameter is the key of self-learning and optimization of the system and directly influences the adjustment direction of the future atmosphere lamp control strategy. And carrying out fine adjustment or great modification on the original atmosphere lamp control strategy according to the value of the reward feedback parameter so as to ensure that the setting of the atmosphere lamp can better meet the expectations of passengers and improve the experience of the passengers.

In a specific embodiment, the process of executing step S106 may specifically include the following steps:

(1) Performing parameter optimization on the first control parameter set according to the reward feedback parameter to obtain a second control parameter set;

(2) Performing strategy gradient analysis on the second control parameter set to determine strategy gradients of the second control parameter set;

(3) Carrying out strategy updating parameters on the reward feedback parameters through strategy gradients to obtain strategy updating parameters;

(4) And carrying out strategy updating on the second control parameter set based on the strategy updating parameters to generate a corresponding second action executing strategy.

Specifically, the first set of control parameters is analyzed and optimized according to the reward feedback parameters. By applying a machine learning and optimization algorithm, which control parameters most affect passenger satisfaction is identified, and the first control parameter set is adjusted accordingly, generating an optimized second control parameter set. And carrying out strategy gradient analysis on the second control parameter set, and calculating the influence degree of each control parameter on a final objective function (namely passenger satisfaction). Policy gradient analysis is a reinforcement learning technique that guides the direction of adjustment of parameters by evaluating the effect of parameter changes on an objective function. From this analysis, the direction of optimization of each control parameter, i.e. the strategy gradient, can be determined. And further analyzing the reward feedback parameters by using the determined strategy gradient to calculate strategy updating parameters. These updated parameters are derived from a combination of the strategy gradients and rewards feedback parameters that indicate how to adjust the control parameters to more effectively promote passenger satisfaction. The calculation of the policy update parameters is a complex optimization process, and the influence of different parameter adjustments on the objective function needs to be accurately evaluated to ensure that the updated control policy can achieve the best effect in practical application. And carrying out strategy updating on the second control parameter set based on the strategy updating parameters to generate a corresponding second action executing strategy. The policy update parameters are applied to the actual adjustment of the control parameters to form an improved action execution policy. The improved strategy not only reflects the actual feedback and preferences of the passengers, but also takes into account the interaction between the control parameters and the overall optimization objectives.

The method for controlling the vehicle atmosphere lamp in the embodiment of the present application is described above, and the device for controlling the vehicle atmosphere lamp in the embodiment of the present application is described below, referring to fig. 2, where an embodiment of the device for controlling the vehicle atmosphere lamp in the embodiment of the present application includes:

an acquisition module 201, configured to acquire passenger behavior data of a target vehicle and extract behavior features to obtain passenger behavior features;

The analysis module 202 is configured to input the passenger behavior feature into an RNN-PSO algorithm to perform analysis on the vehicle atmosphere lamp control parameters, so as to obtain a first control parameter set;

the analysis module 203 is configured to obtain vehicle internal environment data and vehicle external environment data of the target vehicle, and respectively analyze environmental features to obtain vehicle internal environment features and vehicle external environment features;

The processing module 204 is configured to input the first control parameter set, the internal environment feature of the vehicle, and the external environment feature of the vehicle into the dual-depth Q network to perform vehicle atmosphere lamp execution action analysis, so as to obtain a first vehicle atmosphere lamp action execution policy;

The calculation module 205 is configured to perform vehicle atmosphere lamp control according to a first vehicle atmosphere lamp action execution policy, obtain passenger feedback data of the target vehicle, and perform reward feedback data calculation according to the passenger feedback data, so as to obtain a reward feedback parameter;

the generating module 206 is configured to perform parameter optimization on the first control parameter set according to the reward feedback parameter, obtain a second control parameter set, and generate a corresponding second action execution policy according to the second control parameter set.

Through the cooperation of the components, the specific requirements and preferences of the passengers can be identified by acquiring and analyzing the behavior data of the passengers of the target vehicle, and the setting of the atmosphere lamp is further adjusted to match the moods and activities of the passengers. Such personalized lighting control not only improves passenger comfort and satisfaction, but also increases the added value of the vehicle. By combining an RNN-PSO algorithm and a dual-depth Q network, the control method can intelligently analyze the behavior characteristics of passengers and the internal and external environment data of the vehicle, automatically optimize control parameters and adjust atmosphere lamp settings in real time. The intelligent self-adaptive control strategy can ensure the optimal in-vehicle atmosphere under different conditions, and the experience of passengers is improved. And the feedback of the passengers to the atmosphere lamp setting is collected, and the rewarding feedback parameters are obtained according to the feedback calculation, so that the control parameters are further optimized. The feedback-based dynamic learning and optimizing mechanism can continuously improve the atmosphere lamp control strategy, ensure that the system continuously evolves along with the time and better meet the changing demands of passengers. The intelligent adjustment of the brightness and the color of the atmosphere lamp according to the internal and external environments of the vehicle and the behaviors of passengers is beneficial to optimizing the energy consumption. Compared with the traditional manual regulation or fixed mode, the intelligent control method can more effectively use energy, and unnecessary energy waste is reduced. Through considering vehicle external environment characteristic when adjusting the atmosphere lamp, can ensure the vision comfort level and the security of passenger under different driving environment, say at night or rainy day automatically regulated atmosphere lamp luminance, reduce the interference to driver's sight, promote whole driving safety, and then improved the control accuracy of vehicle atmosphere lamp.

The application also provides a control device of the vehicle atmosphere lamp, which comprises a memory and a processor, wherein the memory stores computer readable instructions, and the computer readable instructions, when executed by the processor, cause the processor to execute the steps of the control method of the vehicle atmosphere lamp in the above embodiments.

The present application also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium, or may be a volatile computer-readable storage medium, in which instructions are stored that, when executed on a computer, cause the computer to perform the steps of the method for controlling a vehicle atmosphere lamp.

It will be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, systems and units may refer to the corresponding processes in the foregoing method embodiments, which are not repeated herein.

The integrated units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application may be embodied essentially or in part or all of the technical solution or in part in the form of a software product stored in a storage medium, including instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a read-only memory (ROM), a random access memory (random acceS memory, RAM), a magnetic disk, or an optical disk, or other various media capable of storing program codes.

The above embodiments are only for illustrating the technical solution of the present application, and not for limiting the same; although the application has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A control method of a vehicle atmosphere lamp, characterized by comprising:

2. The method for controlling a vehicle atmosphere lamp according to claim 1, wherein the step of acquiring the passenger behavior data of the target vehicle and extracting the behavior feature to obtain the passenger behavior feature comprises:

Acquiring passenger image data of at least one target passenger through an image sensor in a target vehicle, and acquiring atmosphere lamp interaction data of the at least one target passenger through an atmosphere lamp interaction sensor in the target vehicle;

Analyzing the passenger sitting posture and the activity frequency of the passenger image data to obtain passenger sitting posture data and activity frequency data;

Analyzing the passenger behavior intention of the atmosphere lamp interaction data to obtain behavior intention data, and generating passenger behavior data of the target vehicle according to the behavior intention data, the passenger sitting posture data and the activity frequency data;

performing behavior category analysis on the passenger behavior data through a cluster analysis model to obtain a plurality of passenger behavior categories;

performing principal component feature analysis on the passenger behavior data according to the plurality of passenger behavior categories to obtain initial behavior features of each passenger behavior category;

And carrying out feature fusion on the initial behavior features of each passenger behavior category to obtain the passenger behavior features.

3. The method for controlling a vehicle atmosphere lamp according to claim 1, wherein inputting the passenger behavior feature into RNN-PSO algorithm for vehicle atmosphere lamp control parameter analysis, obtaining a first control parameter set, comprises:

Inputting the passenger behavior characteristics into an RNN-PSO algorithm, wherein the RNN-PSO algorithm comprises a cyclic neural network and a particle swarm optimization algorithm;

extracting hidden state features from the passenger behavior features through 256 GRU units in a first layer in the recurrent neural network to obtain hidden state features;

Predicting the control parameters of the atmosphere lamp of the vehicle to the hidden state features through 256 GRU units of a second layer in the cyclic neural network to obtain a plurality of prediction control parameter values;

Acquiring a numerical control range set of the target vehicle, and constructing a particle population for the plurality of prediction control parameter values through the particle swarm optimization algorithm to obtain the particle population;

Performing particle fitness calculation on the particle population to obtain a particle fitness set corresponding to the particle population, and performing iterative calculation on the particle fitness set until a preset condition is met, so as to generate an optimal solution corresponding to the particle population;

and carrying out control parameter correction and integrated conversion on the plurality of predicted control parameter values through the optimal solution to obtain a first control parameter set.

4. The control method of the vehicle atmosphere lamp according to claim 1, wherein the acquiring the vehicle interior environment data and the vehicle exterior environment data of the target vehicle and respectively performing the environmental feature analysis to obtain the vehicle interior environment feature and the vehicle exterior environment feature includes:

Acquiring vehicle interior environment data and vehicle exterior environment data of the target vehicle, and performing data standardization processing on the vehicle interior environment data and the vehicle exterior environment data to obtain standard vehicle interior environment data and standard vehicle exterior environment data;

Classifying the data set of the standard in-vehicle environment data to obtain in-vehicle temperature and in-vehicle illumination intensity, and classifying the data set of the standard in-vehicle environment data to obtain weather conditions and road condition information;

Performing average value operation on the temperature in the vehicle to obtain average temperature, and performing average value operation on the illumination intensity in the vehicle to obtain average illumination intensity;

Performing correlation analysis on the average temperature and the average illumination intensity to obtain a first correlation analysis result, and performing feature extraction and feature fusion on the average temperature and the average illumination intensity according to the first correlation analysis result to obtain vehicle internal environment features;

And carrying out correlation analysis on the weather condition and the road condition information to obtain a second correlation analysis result, and carrying out feature extraction and feature fusion on the weather condition and the road condition information according to the second correlation analysis result to obtain the external environment feature of the vehicle.

5. The method for controlling a vehicle atmosphere lamp according to claim 1, wherein inputting the first set of control parameters, the vehicle internal environment feature and the vehicle external environment feature into a dual-depth Q network to perform vehicle atmosphere lamp execution action analysis, and obtaining a first vehicle atmosphere lamp action execution strategy comprises:

inputting the first set of control parameters, the vehicle interior environmental feature, and the vehicle exterior environmental feature into a dual depth Q network comprising: an input layer, a first deep neural network, and a second deep neural network;

Performing control parameter coding on the first control parameter set through the input layer to obtain a control parameter coding vector, performing environment characteristic coding on the internal environment characteristic of the vehicle and the external environment characteristic of the vehicle to obtain an internal environment characteristic coding vector and an external environment characteristic coding vector, and performing coding vector splicing on the control parameter coding vector, the internal environment characteristic coding vector and the external environment characteristic coding vector to obtain an input characteristic coding vector;

Inputting the input feature coding vector into the first depth neural network to perform vehicle atmosphere lamp execution action prediction, so as to obtain an initial vehicle atmosphere lamp action execution strategy;

And performing action value evaluation on the initial vehicle atmosphere lamp action execution strategy through the second deep neural network to obtain a first vehicle atmosphere lamp action execution strategy.

6. The method for controlling a vehicle atmosphere lamp according to claim 2, wherein the performing vehicle atmosphere lamp control by the first vehicle atmosphere lamp action execution strategy and obtaining passenger feedback data of the target vehicle, and performing bonus feedback data calculation by the passenger feedback data, obtaining bonus feedback parameters, includes:

performing vehicle atmosphere lamp control on the target vehicle through the first vehicle atmosphere lamp action execution strategy;

Acquiring feedback image data of the at least one target passenger through the image sensor, and acquiring feedback voice data of the at least one target passenger;

Generating a first feedback index according to the feedback image data, and generating a second feedback index according to the feedback voice data;

And calculating the reward feedback data through the first feedback index and the second feedback index to obtain the reward feedback parameter.

7. The method for controlling an atmosphere lamp of a vehicle according to claim 6, wherein the performing parameter optimization on the first control parameter set according to the reward feedback parameter to obtain a second control parameter set, and generating a corresponding second action execution policy according to the second control parameter set, includes:

performing parameter optimization on the first control parameter set according to the reward feedback parameter to obtain a second control parameter set;

Performing strategy gradient analysis on the second control parameter set to determine strategy gradients of the second control parameter set;

Carrying out strategy updating parameters on the reward feedback parameters through the strategy gradient to obtain strategy updating parameters;

and carrying out strategy updating on the second control parameter set based on the strategy updating parameters to generate a corresponding second action executing strategy.

8. A control device for a vehicle atmosphere lamp, characterized by comprising:

9. A control apparatus of a vehicle atmosphere lamp, characterized in that the control apparatus of a vehicle atmosphere lamp comprises: a memory and at least one processor, the memory having instructions stored therein;

The at least one processor invokes the instructions in the memory to cause the control device of the vehicle atmosphere lamp to perform the control method of the vehicle atmosphere lamp according to any one of claims 1-7.

10. A computer-readable storage medium having instructions stored thereon, which when executed by a processor, implement the method of controlling a vehicle atmosphere lamp according to any one of claims 1-7.