System and method for operating system

Abstract

The invention relates to a system and an operation method of the system. And the convenience of the user is improved. This system is provided with: a first in-vehicle device that is mounted on a first vehicle and has a recommendation model that is generated by performing machine learning on a combination of a first input made by a first user of the first vehicle and a first operation of the first vehicle in accordance with the first input; and a second in-vehicle device that is mounted on a second vehicle having a vehicle type different from that of the first vehicle, and that uses the recommendation model acquired from the first in-vehicle device to determine a second operation of the second vehicle corresponding to a second input made by a second user of the second vehicle.

Description

Technical Field

[0001] This disclosure relates to a system and the methods of operating the system. Background Technology

[0002] Techniques for controlling vehicle actions based on the user's preferences have been proposed. For example, Patent Document 1 discloses an example of a system that associates in-vehicle environment settings with user identification information to set a user's preferred in-vehicle environment.

[0003] Existing technical documents

[0004] Patent documents

[0005] Patent Document 1: Japanese Patent Application Publication No. 2019-182096 Summary of the Invention

[0006] When the on-board devices installed vary depending on the vehicle model, the convenience enjoyed by users from the actions of the vehicle controlled by the on-board devices will differ, so there is room for further improvement in user convenience.

[0007] The following are systems that can improve user convenience.

[0008] This disclosure provides a system comprising: a first vehicle-mounted device mounted on a first vehicle, having a recommendation model generated by machine learning of a combination of a first input made by a first user of the first vehicle and a first action of the first vehicle corresponding to the first input; and a second vehicle-mounted device mounted on a second vehicle of a different model from the first vehicle, using the recommendation model obtained from the first vehicle-mounted device to determine a second action of the second vehicle corresponding to a second input made by a second user of the second vehicle.

[0009] Other aspects of this disclosure provide a method of operating a system having first and second vehicle-mounted devices respectively mounted on first and second vehicles of different models, wherein the first vehicle-mounted device has a proposal model generated by machine learning of a combination of a first input made by a first user of the first vehicle and a first action of the first vehicle corresponding to the first input, and the second vehicle-mounted device uses the proposal model obtained from the first vehicle-mounted device to determine a second action of the second vehicle corresponding to a second input made by a second user of the second vehicle.

[0010] The system and other features disclosed herein can improve user convenience. Attached Figure Description

[0011] Figure 1 This is a diagram illustrating an example of the structure of an information provision system.

[0012] Figure 2 This is a diagram showing an example of the structure of an on-board device.

[0013] Figure 3 This is a flowchart illustrating an example of the operation process of an on-board device.

[0014] Figure 4 This is a flowchart illustrating an example of the operation process of an on-board device. Detailed Implementation

[0015] The embodiments are described below with reference to the accompanying drawings.

[0016] Figure 1 This diagram illustrates a structural example of a vehicle control system in one embodiment. The vehicle control system 1 includes one or more server devices 10 connected to each other via a network 11 in a manner capable of information communication, and on-board units 12-1 and 12-2 respectively mounted on vehicles 13-1 and 13-2. The server devices 10 are, for example, cloud computing systems or other computing systems, and function as servers with various installed functions. Vehicles 13-1 and 13-2 are passenger cars, commercial vehicles, etc., of different models, such as internal combustion engine vehicles, hybrid electric vehicles (HEVs), and plug-in hybrid electric vehicles (PHEVs). On-board units 12-1 and 12-2 are computers with communication and information processing functions, respectively controlling the actions of vehicles 13-1 and 13-2. The network 11 includes, for example, mobile communication networks, the Internet, ad hoc networks, LANs (Local Area Networks), MANs (Metropolitan Area Networks), or other networks or any combination thereof.

[0017] In this embodiment, vehicle 13-1 is a higher-class vehicle than vehicle 13-2. For example, vehicle 13-1 is a so-called luxury car, while vehicle 13-2 is a so-called mass-market car that is cheaper than a luxury car. Vehicle 13-1 is equipped with features that provide a more refined user experience than vehicle 13-2, and the on-board unit 12-1 is configured to perform controls that are more conducive to improving user convenience than the on-board unit 12-2. In this embodiment, by reducing the imbalance in user convenience between vehicles 13-1 and 13-2, and improving the user convenience in vehicle 13-2, the overall user convenience of vehicles 13-1 and 13-2 is improved.

[0018] In this embodiment, the vehicle-mounted device 12-1 has a recommendation model obtained by machine learning from a combination of inputs made by users of vehicle 13-1 (including the driver and one or more passengers, hereinafter referred to as premium vehicle users) and actions of vehicle 13-1 corresponding to those inputs (hereinafter referred to as premium vehicle actions). Furthermore, the vehicle-mounted device 12-1 updates the recommendation model using a combination of premium vehicle actions and responses made by premium vehicle users to those actions. The vehicle-mounted device 12-2 uses the recommendation model obtained from the vehicle-mounted device 12-1 to determine actions of vehicle 13-2 corresponding to inputs made by users of vehicle 13-2 (including the driver and one or more passengers, hereinafter referred to as mass-market vehicle users) (hereinafter referred to as mass-market vehicle actions). Here, the inputs made by premium vehicle users or mass-market vehicle users are intentional inputs made by each user, and are vocal content corresponding to each user's expectations and preferences. Hereinafter, such inputs will be referred to as intentional inputs. Furthermore, the responses of luxury car users to luxury car actions are various actions performed by the luxury car user, such as physical states including body temperature or pulse, drowsy or uncomfortable behavior in camera images, and operations on the air conditioning or audio system. According to the vehicle control system 1, by using reinforcement learning based on the responses of luxury car users in the vehicle-mounted device 12-2 to update the proposed model in the vehicle-mounted device 12-1, it is possible to respond to the input of the mass-market car user to achieve further refined mass-market car actions. Therefore, overall, user convenience can be improved.

[0019] Figure 2 Examples of the structures of vehicle-mounted devices 12-1 and 12-2 are shown. Vehicle-mounted devices 12-1 and 12-2 are identically configured in the following ways: They include a communication unit 121, a storage unit 122, a control unit 123, a positioning unit 124, an input unit 125, an output unit 126, and a detection unit 127. These can constitute a single control device, or two or more control devices, or a control device and other devices such as a communication unit. Control devices may include, for example, an ECU (Electronic Control Unit). Communication units may include, for example, a DCM (Data Communication Module). Each part is connected to each other or to the equipment of the vehicle 15 via an in-vehicle network according to a standard such as CAN (Controller Area Network). Furthermore, vehicle-mounted devices 12-1 and 12-2 may also be configured to include an information processing device such as a smartphone or tablet in one part.

[0020] The communication unit 121 includes modules corresponding to mobile communication standards such as LTE (Long Term Evolution), 4G (4th generation), or 5G (5th generation), and modules corresponding to in-vehicle LANs such as CAN. The vehicle-mounted device 12 uses the communication unit 121 to connect to the network 11 via a nearby router or mobile communication base station to communicate with other devices via the network 11, or to communicate with various parts of vehicles 13-1 and 13-2 via the in-vehicle LAN.

[0021] The storage unit 122 includes one or more semiconductor memories, one or more magnetic memories, one or more optical memories, or a combination of at least two of these. Semiconductor memories are, for example, RAM (Random Access Memory) or ROM (Read Only Memory). RAM is, for example, SRAM (Static RAM) or DRAM (Dynamic RAM). ROM is, for example, EEPROM (Electrically Erasable Programmable ROM). The storage unit 122 functions as, for example, a main storage device, an auxiliary storage device, or a cache memory. The storage unit 122 stores information used in the operation of the control unit 123 and information obtained through the operation of the control unit 123. In this embodiment, the storage unit 122 stores a suggestion model 21 and a language model 22.

[0022] A suggestion model 21 is pre-generated and saved to the vehicle-mounted device 12-1 by using machine learning as training data a combination of inputs intentionally made by a premium vehicle user and premium vehicle actions. In vehicle 13-1, suggestion model 21 suggests premium vehicle actions corresponding to the premium vehicle user's intentional inputs. Here, the suggestions include instructions to perform the premium vehicle actions for vehicle 13-1. Vehicle-mounted device 12-1 updates suggestion model 21 using reinforcement learning based on the premium vehicle user's responses through a process described later. The updated suggestion model 21 is then sent from vehicle-mounted device 12-1 to vehicle-mounted device 12-2 via server device 10 and saved there. In vehicle 13-2, suggestion model 21 suggests Volkswagen vehicle actions corresponding to the Volkswagen user's intentional inputs.

[0023] Language model 22 is a model generated by machine learning using a large amount of text data to study the patterns, structure, and meaning of natural language. It is used for text generation, extraction, and parsing. For example, language model 22 is a relatively small-scale language model based on the Transformer architecture with millions to hundreds of millions of parameters. In the vehicle-mounted devices 12-1 and 12-2, the form and scale of language model 22 can also be different.

[0024] The control unit 123 includes one or more processors, one or more dedicated circuits, or a combination thereof. The processor is a general-purpose processor such as a CPU (Central Processing Unit) or a dedicated processor such as a GPU (Graphics Processing Unit) specifically designed for particular processing. The dedicated circuit is, for example, a FPGA (Field-Programmable Gate Array) or an ASIC (Application-Specific Integrated Circuit). The control unit 123 controls the various parts of the vehicle-mounted device 12 while performing information processing related to the operation of the vehicle-mounted device 12.

[0025] The functions of the control unit 123 are implemented by a control / processing program executed by a processor included in the control unit 123. The control / processing program is a program that enables the computer to perform a function corresponding to a step of an action included in the control unit 123 by causing the computer to execute that step. In other words, the control / processing program is a program that enables the computer to function as the control unit 123. Furthermore, some or all of the functions of the control unit 123 may also be implemented using dedicated circuitry included in the control unit 123.

[0026] The positioning unit 124 includes one or more GNSS (Global Navigation Satellite System) receivers. Among GNSS receivers, this includes, for example, at least one of GPS (Global Positioning System), QZSS (Quasi-Zenith Satellite System), BeiDou, GLONASS (Global Navigation Satellite System), and Galileo. The positioning unit 124 sends the positioning results to the control unit 123, where the control unit 123 calculates the position information of the vehicle-mounted device 12.

[0027] The input unit 125 includes one or more input interfaces. These input interfaces may include, for example, a microphone for accepting voice input, physical keys, electrostatic capacitive keys, indicator devices, or a touchscreen integrated with the display. Additionally, the input interface may include a camera for capturing images inside the vehicle. The input unit 125 accepts various input information, including the user's voice, and sends the input information to the control unit 123, or sends the captured image to the control unit 123.

[0028] The output unit 126 includes one or more output interfaces. The output interface may be, for example, a speaker or a display. The display may be, for example, an LCD (Liquid Crystal Display) or an organic EL (Electro-Luminescence) display. The output unit 126 outputs information obtained through the operation of the control unit 123.

[0029] The detection unit 127 has an interface with one or more sensors for detecting the status of various parts of the vehicles 13-1 and 13-2, or one or more sensors. These sensors include, for example, sensors for vehicle speed and acceleration, and sensors for indoor and outdoor temperature and humidity. The sensors in the detection unit 127 of the vehicle-mounted device 12-1 also include sensors that use infrared light to measure and detect the body temperature, pulse, and other physical conditions of the luxury car user, as well as sensors that detect the amount of operation performed by the luxury car user on in-vehicle equipment such as air conditioning and audio systems.

[0030] Figure 3 This is a flowchart illustrating the operation of the onboard device 12-1 related to the reinforcement learning of the proposed model 21. Figure 3 Each step in the process is an information processing step performed by the control unit 123 of the vehicle-mounted device 12-1.

[0031] In step S31, the control unit 123 acquires intentional input. The control unit 123 acquires the vocalizations of the luxury car user from the input unit 125, and, for example, transcribes the vocalizations into text using speech recognition processing and inputs it into the language model 22. The language model 22 then acquires intentional input corresponding to the vocalizations. Intentional inputs may correspond to vocalizations such as "I want to go to ○○," "I want to listen to ○○'s music," "I want to concentrate on driving," or "I want to change my mood," which suggest states such as anxiety, tension, drowsiness, boredom, lack of focus, or discomfort for the luxury car user.

[0032] In S32, the control unit 123 obtains user attribute information. For example, the control unit 123 obtains attribute information about the luxury car user from the server device 10. Attribute information includes, for example, the luxury car user's age, gender, place of residence, occupation, etc. For example, when purchasing vehicle 13-1, the attribute information is pre-sent from any information processing device to the server device 10 and stored. The attribute information can also be stored on the server device 10 in any form, such as text. Alternatively, the attribute information can be obtained from content submitted to the luxury car user's SNS (Social Network Service). The control unit 123 can collect various texts and submissions from various servers, parse these using language model 22 or a large-scale language model possessed by the server device 10, and obtain the attribute information.

[0033] In S33, the control unit 123 determines the actions of vehicle 13-1. The control unit 123 inputs the input intentionally made by the premium vehicle user to the suggestion model 21, and uses the suggestion model 21 to determine the corresponding actions of vehicle 13-1, i.e., premium vehicle actions. The determined premium vehicle actions include, for example, adjusting the air conditioning temperature / humidity, adjusting the audio system (music selection, volume adjustment), and recommending parking / stopping locations (sound output) corresponding to the premium vehicle user's intentional input. Furthermore, the control unit 123 can also incorporate the premium vehicle user's attribute information to determine the actions of vehicle 13-1. For example, the control unit 123 sets the air conditioning and selects music based on the premium vehicle user's age and gender. Air conditioning settings and music information suitable for each age group and gender are pre-stored in the storage unit 122. Additionally, the control unit 123 selects a parking / stopping location near the premium vehicle user's residence or workplace based on map information. Map information is pre-stored in the storage unit 122. Furthermore, for example, if the user of a luxury car is a male office worker in his 50s and the time is in the afternoon, he is more likely to be wearing a suit, so the air conditioner is set to lower the temperature and increase the airflow. Or, if the user is a female in her 60s and it is summer, she is more likely to be wearing light clothing, so the air conditioner is set to not lower the temperature excessively and to suppress the airflow.

[0034] In S34, the control unit 123 instructs the operation of the vehicle 13-1. The control unit 123 sends instructions to the equipment, actuators, etc. of the vehicle 13-1 used to perform the high-end vehicle operation. Instructions include, for example, instructions for adjusting the temperature / humidity of the air conditioning unit, instructions for selecting music and adjusting the volume of the audio system, and instructions for outputting a recommended parking / stopping sound through the output unit 126.

[0035] In step S35, the control unit 123 acquires and analyzes the user's reaction. The control unit 123 obtains the user's reaction from the sensor unit 127 (monitoring the user's physical condition), operations on the vehicle 13-1, and camera images acquired by the input unit 125. Furthermore, the control unit 123 analyzes the user's reaction and determines whether it was a positive or negative reaction to the previous action. For example, if the temperature / humidity changes, a positive reaction is classified if an appropriate body temperature / pulse rate corresponding to the changed temperature / humidity is obtained; otherwise, a negative reaction is classified. Information regarding the appropriate body temperature / pulse rate for each temperature / humidity level is arbitrarily preset and stored in the storage unit 122. Additionally, in the camera image, a positive reaction is determined when a user exhibiting drowsy or uncomfortable behavior becomes alert or comfortable; otherwise, a negative reaction is determined. The control unit 123 uses image processing on consecutive frames to export patterns of the user's actions, expressions, and gestures, and matches these exported patterns with preset patterns to determine whether the behavior is that of a sober or comfortable person. Alternatively, if the control unit detects that the user has stopped or reversed the change in response to changes in temperature / humidity or audio, it determines that the response is negative.

[0036] In step S36, the control unit 123 performs reinforcement learning on the proposal model 21. The control unit 123 executes steps S31-S34 an arbitrary number of times, for example, at an arbitrary period of approximately a few seconds, to obtain a combination of intentional input and the corresponding advanced vehicle action, as well as a combination of the advanced vehicle action and the response to that action. The control unit 123 uses positive responses as rewards and performs reinforcement learning on the proposal model 21 using an arbitrary algorithm. Step S36 can be executed at any frequency. For example, step S36 can be executed for each trip of vehicle 13-1, or at any interval, such as every few days.

[0037] In S37, the control unit 123 updates the proposed model 21 with the result of reinforcement learning and saves it to the storage unit 122.

[0038] In step S38, the control unit 123 copies the proposed model 21 and sends the information of the copied proposed model 21 to the vehicle-mounted device 12-2. The vehicle-mounted device 12-2 receives the information of the copied proposed model 21 via the server device 10 and saves it to its own storage unit 122, thereby sharing the updated proposed model 21 in the vehicle-mounted devices 12-1 and 12-2. Furthermore, step S38 can be executed either every time the proposed model 21 is updated in step S37, or it can be executed any number of times the proposed model 21 is updated.

[0039] Figure 4This is a flowchart illustrating the operation of the vehicle-mounted device 12-2 after obtaining the suggested model 21. Figure 4 Each step in the process is an information processing step performed by the control unit 123 of the vehicle-mounted device 12-2.

[0040] In step S41, the control unit 123 acquires intentional input. The control unit 123 acquires the vocalizations of the Volkswagen user from the input unit 125, and, for example, transcribes the vocalizations into text using speech recognition processing and inputs it into the language model 22. The language model 22 then acquires intentional input corresponding to the vocalizations. Intentional inputs may correspond to vocalizations such as "I want to go to ○○," "I want to listen to ○○'s music," "I want to concentrate on driving," or "I want to change my mood," which suggest states such as anxiety, tension, drowsiness, boredom, lack of focus, or discomfort for the Volkswagen user.

[0041] In step S42, the control unit 123 obtains user attribute information. For example, the control unit 123 obtains attribute information about the Volkswagen user from the server device 10. Attribute information includes, for example, the Volkswagen user's age, gender, place of residence, occupation, etc. For example, when purchasing vehicle 13-2, the attribute information is pre-sent from any information processing device to the server device 10 and stored. The attribute information can also be stored on the server device 10 in any form, such as text. Alternatively, the attribute information can be obtained from submissions to the Volkswagen user's SNS. The control unit 123 can collect various texts and submissions from various servers, parse these using language model 22 or the large-scale language model possessed by the server device 10, and obtain the attribute information.

[0042] In step S43, the control unit 123 determines the actions of vehicle 13-2. The control unit 123 inputs the input intentionally made by the Volkswagen user to the suggestion model 21, and uses the suggestion model 21 to determine the corresponding actions of vehicle 13-2, i.e., Volkswagen actions. The determined Volkswagen actions include, for example, adjusting the air conditioning temperature / humidity, adjusting the audio system (music selection, volume adjustment), and recommending parking / stopping locations (sound output) according to the Volkswagen user's intentional input. Furthermore, the control unit 123 can also incorporate the Volkswagen user's attribute information to determine the actions of vehicle 13-2. For example, the control unit 123 sets the air conditioning and selects music according to the Volkswagen user's age and gender. Air conditioning settings and music information suitable for each age group and gender are pre-saved in the storage unit 122. Additionally, the control unit 123 selects parking / stopping locations near the Volkswagen user's residence or workplace based on map information. Map information is pre-saved in the storage unit 122.

[0043] In S44, the control unit 123 instructs the operation of the vehicle 13-2. The control unit 123 sends instructions to the equipment, actuators, etc. of the vehicle 13-2 used to perform the vehicle's operation. Instructions include, for example, instructions for adjusting the temperature / humidity of the air conditioning unit, instructions for selecting music and adjusting the volume of the audio system, and instructions for recommending parking / stopping sound output using the output unit 126.

[0044] Thus, in vehicle 13-2, the suggested model 21 obtained from vehicle 13-1 is used to execute Volkswagen actions based on the intentional input of the Volkswagen user. Therefore, according to this embodiment, user convenience can be improved.

[0045] In the foregoing, embodiments have been described with reference to the accompanying drawings, but it is intended that those skilled in the art can readily make various modifications and alterations based on this disclosure. Therefore, it is intended that these modifications and alterations be included within the scope of this disclosure. For example, the functions of each unit, step, etc., can be reconfigured logically without contradiction, and multiple units, steps, etc., can be combined into one or divided.

[0046] (Symbol Explanation)

[0047] 1: Information providing system; 10: Server device; 11: Network; 12-1, 12-2: Vehicle-mounted device; 13-1, 13-2: Vehicle; 121: Communication unit; 122: Storage unit; 123: Control unit; 124: Positioning unit; 125: Input unit; 126: Output unit; 127: Detection unit.

Claims

1. A system having: A first vehicle-mounted device, installed in a first vehicle, generates a suggestion model by performing machine learning on a combination of a first input made by a first user of the first vehicle and a first action of the first vehicle corresponding to the first input, and a combination of the first action and the user's response to the first action; and A second vehicle-mounted device, installed in a second vehicle of a different model than the first vehicle, uses the suggested model obtained from the first vehicle-mounted device to determine a second action of the second vehicle corresponding to a second input made by a second user of the second vehicle.

2. A system having: A first vehicle-mounted device, mounted on a first vehicle, includes a suggestion model generated through machine learning of a combination of a first input made by a first user of the first vehicle and a first action of the first vehicle corresponding to the first input; and A second vehicle-mounted device, installed in a second vehicle of a different model than the first vehicle, uses the suggested model obtained from the first vehicle-mounted device to determine a second action of the second vehicle corresponding to a second input made by a second user of the second vehicle.

3. The system according to claim 2, wherein, The first vehicle-mounted device further performs machine learning on the combination of the first action and the first user's response to the first action to update the proposed model.

4. The system according to claim 2, wherein, The first vehicle-mounted device also uses the attribute information of the first user to update the suggested model.

5. The system according to claim 4, wherein, The first vehicle-mounted device uses a language model to obtain the attribute information of the first user.

6. The system according to claim 2, wherein, The second vehicle-mounted device also uses the attribute information of the second user to determine the second action.

7. The system according to claim 6, wherein, The second vehicle-mounted device uses a language model to obtain the attribute information of the second user.

8. The system according to claim 2, wherein, The first input contains vocal content emitted by the first user.

9. The system according to claim 2, wherein, The second input contains vocal content emitted by the second user.

10. A method of operating a system, the system comprising a first vehicle-mounted device and a second vehicle-mounted device respectively mounted on a first vehicle and a second vehicle of different vehicle types, wherein, The first vehicle-mounted device has a recommendation model generated by machine learning of a combination of a first input made by a first user of the first vehicle and a first action of the first vehicle corresponding to the first input. The second vehicle-mounted device uses the proposed model obtained from the first vehicle-mounted device to determine the second action of the second vehicle corresponding to the second input made by the second user of the second vehicle.

11. The method of operating the system according to claim 10, wherein, The first vehicle-mounted device further performs machine learning on the combination of the first action and the first user's response to the first action to update the proposed model.

12. The method of operating the system according to claim 10, wherein, The first vehicle-mounted device also uses the attribute information of the first user to update the suggested model.

13. The method of operating the system according to claim 12, wherein, The first vehicle-mounted device uses a language model to obtain the attribute information of the first user.

14. The method of operating the system according to claim 10, wherein, The second vehicle-mounted device also uses the attribute information of the second user to determine the second action.

15. The method of operating the system according to claim 14, wherein, The second vehicle-mounted device uses a language model to obtain the attribute information of the second user.

16. The method of operating the system according to claim 10, wherein, The first input contains vocal content emitted by the first user.

17. The method of operating the system according to claim 10, wherein, The second input contains vocal content emitted by the second user.

Images