Procedure for adjusting an air conditioner

Abstract

The invention relates to a method (1) for user-adapted adjustment of a vehicle's air conditioning system (4) for a user (3) using a control architecture. In this method, data relating to the air conditioning settings of all users (3) and all vehicles (4) are collected, and a mean base model (BM) is trained in an AI unit based on the collected data (2) of all users (3) and all vehicles (4). Then, a comfort model (KM) is created in the AI ​​unit, at least from the base model (BM), and user-adapted settings (6) for adjusting the air conditioning system for a specific user (3) are predicted based on the comfort model (KM). The invention also relates to the control architecture for carrying out the method (1).

Description

The invention relates to a method for user-adjusted setting of an air conditioning system of a vehicle for a user by means of a control architecture according to the preamble of claim 1.A user of an air conditioner can set it depending on his thermal feeling by changing a plurality of parameters. By means of characteristic-map-based climate control, the setting can be taken away from the user and the operation of the climate control system can thereby be simplified. The thermal feeling of comfort of the individual users is however very individual. The conventional characteristic map-based climate controls are disadvantageously designed for an "average" person and cannot adapt to the individual user in an automated manner. If the user feels uncomfortable in the characteristic-map-based climate control, he must re-set the air conditioning system laboriously by hand.Another solution provides AI-based comfort (AI) models that can adapt to the individual preferences of the individual user. In this case, the settings of the air conditioning system carried out by the user are correlated with context data-for example external temperature or internal temperature-and used for training the comfort model. The trained comfort model can then predict the optimum settings of the air conditioning system for the individual user at the predefined point in time from the context data. Unfortunately, this solution does not provide for the comfort model to be transferred to other users and also to other vehicles.The object of the invention is therefore to specify an improved or at least alternative embodiment for a method of the generic type, in which the described disadvantages are overcome. The object of the invention is also to provide a corresponding control architecture for carrying out the method.This object is achieved according to the invention by the subject matter of the independent claims. Advantageous embodiments are the subject of the dependent claims.The present invention is based on the general idea of creating the comfort model by a superposition, in order to be able to transmit the comfort model to different vehicles and different users.The method is provided for user-adjusted setting of an air conditioning system of a vehicle for a user by means of a control architecture. By means of the control architecture, data relating to the setting of the air conditioning system of all users and / or of all vehicles are collected and a medium basic model is trained in a AI unit on the basis of the collected data of all users and of all vehicles. Then, a comfort model is created at least from the basic model in the AI unit, and user-adjusted settings for setting the air conditioner for a specific user are predicted based on the comfort model.The comfort model thus contains at least the middle basic model, which takes into account the collected data of all users and all vehicles. The trained basic model can therefore be used to predict the user-adjusted settings for all users in all vehicles or for any users in any vehicles. Accordingly, the comfort model based on the base model can be transmitted to any vehicles and / or any users. As a result, the comfort model can on the one hand reduce the number of details to be carried out manually by the user and on the other hand be transmitted to any vehicles and / or any users with reduced outlay. The basic model and the comfort model can be data-based models, whereby the development effort is significantly reduced compared to conventional characteristic map-based models. The basic model can alternatively also be mapped by a characteristic map model.Advantageously, the collected data of all users and all vehicles can be transmitted to a cloud and the middle basic model can be trained with the collected data in the AI unit of the cloud. Accordingly, the comfort model can also be created in the cloud at least from the middle basic model. As a result, the middle basic model and / or a submodel described in more detail below can be trained in a centralized manner with the collected data of all users and all vehicles. Subsequently, the trained comfort model can be transmitted from the cloud to a specific vehicle using the air conditioning system to be set, and the user-adjusted settings of the air conditioning system to be set can be ascertained directly in the vehicle for a specific user. Alternatively, the user-adjusted settings of the air conditioning system of a specific vehicle to be adjusted can be determined for a specific user in the cloud and transmitted to the vehicle. The cloud is a network that provides computer resources such as servers and / or data storage and / or applications.In the context of the present invention, vehicles are intended to mean all vehicles having an air conditioning system, regardless of their properties. The vehicles may be divided into a plurality of different vehicle types, wherein one vehicle type includes a plurality of vehicles having identical characteristics. A specific vehicle can be assigned to only one vehicle type. The vehicles of a vehicle type are, for example, vehicles of a specific manufacturer or vehicles of a specific model of a specific manufacturer.The data related to the setting of the air conditioner may include context data of the vehicle such as interior temperature and / or exterior temperature and / or humidity and / or sun altitude and / or light intensity and / or air pressure and / or a vehicle type. The data relating to the setting of the air conditioning system can comprise setting data entered by the user, such as desired temperature and / or blower level to be set and / or seat heating to be set and / or heating surface to be set and / or heating intensity to be set. The data relating to the setting of the air conditioning system can comprise user data with physiological user information of the user, such as, for example, user identity and / or seat position and / or skin temperature and / or pulse and / or breathing rate and / or age and / or sex and / or clothing condition and / or emotional condition and / or biometric inputs and / or physiological inputs. The data related to the setting of the air conditioner may include user-specific sensation optimization data such as "head too cold", "right leg too warm", "the wind is unpleasant".In an advantageous embodiment of the method, at least one submodel can additionally be trained or created in the AI unit. The comfort model can then be created by a superposition of the base model and the at least one submodel. The respective sub-model can be trained based on the total collected data of all users and of all vehicles or based on a subset of the collected data of all users and / or of all vehicles. Accordingly, the respective partial model can correct or adapt the mean basic model and thereby individualize specific aspects or specific parameters of the comfort model. In particular, the respective partial model can adapt or individualize the comfort model based on the average basic model for a specific user and / or for a specific vehicle and / or for a specific user of a specific vehicle.Advantageously, a plurality of sub-models can be trained or created in the AI unit, which sub-models reflect respectively deviating aspects or deviating parameters of the comfort model. For this purpose, individual data relating to the setting of the air conditioning system can be excluded during the training of the one submodel and taken into account during the training of the other submodel. For example, when training the sub-model, the user data and / or the context data and / or the setting data and / or the optimization data can be partially or completely excluded. Thus, for example, in one of the submodels, the user data or the context data can be partially or completely excluded and the submodel can be trained independently of the user or independently of the vehicle. As already explained above, the respective submodels can then adapt or individualize the comfort model based on the average basic model for a specific user and / or for a specific vehicle and / or for a specific user of a specific vehicle.The respective submodel can be created, for example, in the form of a deviation from the basic model. In other words, the respective submodel can specify how strongly the comfort model deviates from the middle base model if an aspect taken into account by the submodel or a parameter taken into account by the submodel is adapted or individualized in the comfort model. The respective submodel can be mapped, for example, by a characteristic map model.The user-adjusted settings of the air conditioning system can comprise values of the parameters to be set in the air conditioning system-for example, target temperature to be set and / or blower level to be set-for user-adjusted setting of the air conditioning system. The basic model may output or contain at least an absolute value of at least one associated parameter of the air conditioning system to be set, such as target temperature equal to 23.3° C.. The respective submodel can output or contain at least one differential value of at least one associated parameter of the air conditioning system to be set-for example difference from the target temperature equal to -0.5° C.Advantageously, the control architecture can be used to train the submodel in the form of a vehicle model. The vehicle model can be a model based on the collected data of all users of a specific vehicle type of all vehicles. The vehicle model accordingly takes into account the collected data of all users of a specific vehicle type and thereby properties of a specific vehicle type independently of the user. The vehicle model may then be used to predict the customized settings in a particular vehicle type for all users.Advantageously, the control architecture can be used to train the submodel in the form of a user model. The user model can be a model based on the collected data of a specific user of a plurality of vehicles. The trained user model accordingly takes into account the collected data of a specific user of a plurality of vehicles and thus preferences of a specific user independently of vehicles or vehicle types. The user model can thus be used to predict the user-adjusted settings for a particular user in all vehicles.Advantageously, the control architecture can be used to train the submodel in the form of a user vehicle model. The user vehicle model may be a model based on the collected data of a specific user of a specific vehicle type of all vehicles. In the trained user vehicle model, the collected data of a specific user in a specific vehicle type and thereby preferences of a specific user and properties of a specific vehicle type are accordingly taken into account. The user vehicle model may be used to predict the customized settings for a particular user in a particular vehicle type.The vehicle model and / or the user model and / or the user vehicle model can deviate from one another. The individual sub-models take account of deviating subsets of the collected data and accordingly have a deviating accuracy and a deviating usability. In particular, sub-models can be related to a specific user and / or to a specific vehicle type.The comfort model can be created by superimposing the basic model with the vehicle model and / or the user model and / or the user vehicle model. Based on the superimposed comfort model, the user adjusted settings for adjusting the air conditioner for a particular user can then be predicted with a high accuracy.The invention also relates to a control architecture for carrying out the method described above. The control architecture is designed to collect the data relating to the setting of the air conditioning system of all users and all vehicles; to train a medium basic model on the basis of the collected data of all users and all vehicles in an AI unit; to create a comfort model at least from the basic model in the AI unit; to predict user-adjusted settings for setting the air conditioning system for a specific user on the basis of the comfort model. It is understood that the control architecture may include the hardware necessary to execute the method and the software necessary to execute the method. The hardware necessary for executing the method and the software necessary for executing the method can be a part of a cloud and / or a part of the vehicle. In order to avoid repetitions, reference is made at this point to the above explanations.Further important features and advantages of the invention are evident from the dependent claims, from the drawings and from the associated description of the figures with reference to the drawings.It is understood that the features mentioned above and those still to be explained below can be used not only in the respectively specified combination, but also in other combinations or alone, without departing from the scope of the present invention.Preferred exemplary embodiments of the invention are illustrated in the drawings and are explained in more detail in the following description, wherein the same reference numerals refer to the same or similar or functionally the same components.They show, in each case schematically FIG. 1 shows a data structure in a method according to the invention; FIG. 2 shows a sequence of the method according to the invention.FIG. 1 shows a data structure in a method 1 according to the invention. In the method 1, data 2 relating to the setting of the air conditioning system of all users 3 and of all vehicles 4 are collected. Based on the collected data 2, a basic model BM and submodels 5 are trained in a AI unit. The trained submodels 5 in this exemplary embodiment are a vehicle model FM, a user model NM and a user vehicle model NFM. A comfort model KM can be created from the basic model BM, optionally superimposed with the vehicle model FM and / or the user model NM and / or the user vehicle model NFM.The basic model BM is a middle model that is trained or created with the collected data 2 of all users 3 and all vehicles 4. The basic model BM mirrors the user-adjusted settings for all users 3 and all vehicles 4. if the comfort model KM is created solely from the basic model BM, the comfort model KM can predict the user-adjusted settings for any or all users 3 and for any or all vehicles 4.The vehicle model FM is a deviation from the basic model BM, which is trained or created with the collected data 2 of all users 3 of a specific vehicle type of the vehicles 4. The vehicle model FM thus indicates how tract the user-adjusted settings for the specific vehicle type distinguish from the user-adjusted settings of all further vehicle types. If the comfort model MK is created by the superposition of the base model BM and the vehicle model FM, the comfort model MK can predict the user-adjusted settings for all users 3 of a specific vehicle type.The user model NM is a deviation from the basic model BM, which is trained or created with the collected data 2 of a specific user 3 of all vehicles 4. The user model NM specifies how tract the user-adjusted settings for the specific user 3 differ from the user-adjusted settings of all further users 3. If the comfort model MK is created by the superposition of the basic model BM and the user model NM, the comfort model KM can predict the user-adjusted settings for a specific user 3 of all or any vehicles 4.The user vehicle model NFM is a deviation from the basic model BM, which is trained or created with the collected data 2 of a specific user 3 of a specific vehicle type. The user vehicle model NFM indicates how tract the user-adjusted settings for the specific user 3 in the specific vehicle type distinguish from the user-adjusted settings of all further users in all further vehicle types. If the comfort model MK is created by the superposition of the base model BM and the user vehicle model NFM, the comfort model MK can predict the user-adjusted settings for a specific user 3 of a specific vehicle type.The comfort model KM can be created in particular by the superposition of the basic model MB with all the mentioned submodels 5.FIG. 2 partially shows a sequence of method 1 according to the invention. As already explained above, data 2 relating to the setting of the air conditioning system of all users 3 and all vehicles 4 are collected in method 1. Based on the data 2, the basic model BM and submodels 5-vehicle model FM, user model NM, user vehicle model NFM-are trained or created. Subsequently, the basic model BM and the submodels 5 are superimposed to form the comfort model KM. User-adjusted settings 6 for setting the air conditioning system of a specific user 3 in a specific vehicle 4 are then predicted on the basis of the comfort model KM.

Claims

Method (1) for user-adjusted setting of an air conditioning system of a vehicle (4) for a user (3), wherein by means of a control architecture: - data of all users (3) and / or of all vehicles (4) relating to the setting of the air conditioning system are collected; - a medium basic model (BM) is trained in an AI unit based on the collected data (2) of all users (3) and of all vehicles (4), - a comfort model (KM) is created at least from the basic model (BM) in the AI unit; - user-adjusted settings (6) for setting the air conditioning system for a specific user (3) are predicted based on the comfort model (KM).Method according to Claim 1, characterized in that the average basic model (BM) is mapped by a characteristic map model.Method according to Claim 1 or 2, characterized in that at least one submodel (5) is additionally trained in the AI unit and the comfort model (KM) is created by superimposing the basic model (BM) and the at least one submodel (5).Method according to Claim 3, characterized in that the respective partial model (5) is created in the form of a deviation from the basic model (BM).Method according to one of Claims 3 to 4, characterized in that the control architecture is used to train the submodel (5) in the form of a vehicle model (FM), the vehicle model (FM) being a model based on the collected data (2) of all users (3) of a specific vehicle type of all vehicles (4).Method according to one of Claims 3 to 5, characterized in that the control architecture is used to train the submodel (5) in the form of a user model (NM), the user model (NM) being a model based on the collected data (2) of a specific user (3) of a plurality of vehicles (4).Method according to one of Claims 3 to 6, characterized in that the control architecture is used to train the submodel (5) in the form of a user vehicle model (NFM), the user vehicle model (NFM) being a model based on the collected data (2) of a specific user (3) of a specific vehicle type of all vehicles (4).Method according to at least two of Claims 5 to 7, characterized in that the vehicle model (FM) and / or the user model (NM) and / or the user vehicle model (NFM) deviate from one another.Method according to at least two of Claims 5 to 7, characterized in that the comfort model (KM) is created by superimposing the basic model (BM) with the vehicle model (FM) and / or the user model (NM) and / or the user vehicle model (NFM).Control architecture for carrying out the method (1) according to one of the preceding claims, wherein the control architecture is designed: - for collecting the data (2) relating to the setting of the air conditioning system of all users (3) and of all vehicles (4); - for training a medium basic model (BM) based on the collected data (2) of all users (3) and of all vehicles (4) in a AI unit; - for creating a comfort model (KM) at least from the basic model (BM) in the AI unit; - for predicting user-adjusted settings (6) for setting the air conditioning system for a specific user (3) based on the comfort model (KM).

Images