Guidance device, guidance method, program, and guidance system
The guidance device optimizes power consumption in LLM-based systems by generating prompts tailored to the power margin, addressing high computational demands and reducing energy usage.
Patent Information
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- DENSO TEN LTD
- Filing Date
- 2024-12-25
- Publication Date
- 2026-07-07
AI Technical Summary
The increased power consumption in guiding systems using Large Language Models (LLM) poses a challenge due to the high computational demands, necessitating effective power management solutions.
A guidance device that generates prompts based on the power margin of a response generation device, including an LLM, to adjust computation and reduce power consumption by controlling the amount of information and images requested, thereby optimizing power usage.
This approach efficiently manages power consumption by adjusting the computational load of the LLM based on available power, reducing energy usage while maintaining effective guidance functionality.
Smart Images
Figure 2026112661000001_ABST
Abstract
Description
Technical Field
[0006] , , ,
[0005] , , ,
[0001] The present invention relates to a guiding device, a guiding method, a program, and a guiding system.
Background Art
[0002] Conventionally, there has been a technology of reading out guiding text regarding a landmark within the arrival time to a predetermined landmark (see, for example, Patent Document 1 and Patent Document 2).
Prior Art Documents
Patent Documents
[0003]
Patent Document 1
Patent Document 2
Summary of the Invention
Problems to be Solved by the Invention
[0004] Due to the recent development of LLM (Large Language Models), the spread of a guiding system using LLM is expected. However, in a guiding system using LLM, since the amount of computation of LLM is large, the power consumption becomes larger compared with a conventional guiding system. Therefore, in a guiding system using LLM, management of power consumption becomes important.
[0005] In view of the above circumstances, an object of the present invention is to provide a technology capable of achieving good management of power consumption.
Means for Solving the Problems
[0006] An exemplary guidance device of the present invention is a guidance device that outputs guidance information, which generates a prompt for acquiring the guidance information based on the power margin of a response generation device including an LLM, transmits the generated prompt to the response generation device, and outputs the guidance information using the response of the response generation device. [Effects of the Invention]
[0007] According to an exemplary version of the present invention, a prompt is generated to obtain guidance information on the amount of information based on the power margin for the answer generation device. As a result, the amount of computation of the LLM is adjusted according to the power margin for the answer generation device, enabling efficient power consumption management. [Brief explanation of the drawing]
[0008] [Figure 1] A diagram showing an example configuration of the guidance system according to the embodiment. [Figure 2] Diagram showing an example configuration of a guidance device. [Figure 3] A diagram showing an example of a prompt table. [Figure 4] A diagram showing an example of a power margin table. [Figure 5] Diagram showing an example configuration of the response generation device. [Figure 6] Flowchart for guidance processing [Figure 7] Diagram showing other configuration examples of the guidance system. [Modes for carrying out the invention]
[0009] Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the drawings.
[0010] [Information System] Figure 1 is a diagram showing an example configuration of the guidance system SYS1 according to an embodiment. As shown in Figure 1, the guidance system SYS1 includes a guidance device 1 and an answer generation device 2.
[0011] The guidance device 1 and the response generation device 2 are used inside the vehicle V1 in the present embodiment. The guidance device 1 may be a device installed in the vehicle V1. The response generation device 2 is a device installed in the vehicle V1. In the present embodiment, the vehicle V1 includes a microphone 3, a speaker 4, a drive recorder 5, and a navigation device 6.
[0012] The guidance device 1 outputs guidance information using the response generated by the response generation device 2.
[0013] The response generation device 2 includes an LLM. The response generation device 2 inputs the prompt sent from the guidance device 1 into the LLM and generates a response according to the prompt.
[0014] The microphone 3 collects the voice emitted from the passengers in the vehicle V1 and converts it into voice data, which is an electrical signal.
[0015] The speaker 4 converts the voice data, which is an electrical signal, into voice and outputs it.
[0016] The drive recorder 5 records the surrounding situation of the vehicle V1 (especially the image showing the front of the vehicle V1).
[0017] The navigation device 6 notifies the current position of the vehicle V1 and guides the route to the destination.
[0018] [Guidance Device] FIG. 2 is a diagram showing a configuration example of the guidance device 1. The guidance device 1 includes a communication unit 11, a storage unit 12, and a controller 13.
[0019] The communication unit 11 transmits and receives arbitrary signals to and from the response generation device 2. The communication between the guidance device 1 and the response generation device 2 may be wired communication or wireless communication. The communication unit 11 receives arbitrary signals from the microphone 3. The communication unit 11 transmits arbitrary signals to the speaker 4. The communication unit 11 transmits and receives arbitrary signals to and from the drive recorder 5. The communication unit 11 transmits and receives arbitrary signals to and from the navigation device 6. Note that the controller 13 can transmit and receive arbitrary information to and from the response generation device 2 etc. using the communication unit 11, but in the following, the description of the communication unit 11 may be omitted.
[0020] The storage unit 12 is configured to include a non-volatile memory such as a ROM (Read only memory) or a flash memory, and a volatile memory such as a RAM (Random access memory). The storage unit 12 stores a prompt table 121, a power margin table 122, and a program 123.
[0021] As shown in FIG. 3 for example, the prompt table 121 is a table in which the degree of power margin (power margin for the response generation device 2) is linked to the structure of the prompt. The main question sentence included in the prompt is the voice data corresponding to the user's question converted into text data. The image attached to the prompt is an image taken by the drive recorder 5. Note that in the prompt table 121 of FIG. 3, the power margin is classified into three levels: "high", "medium", and "low", but the number of classifications of the power margin is not limited to three and may be plural.
[0022] In the prompt table 121 of Figure 3, the prompt when the power margin is "high" has a structure in which the main question is followed by the conditional statement "Please also provide three pieces of trivia," and two images to be attached (two types of zoom (different shooting ranges)). Also in the prompt table 121 of Figure 3, the prompt when the power margin is "medium" has a structure in which the main question is followed by the conditional statement "Please also provide one piece of trivia," and one image to be attached. Also in the prompt table 121 of Figure 3, the prompt when the power margin is "low" has a structure in which the main question is followed by the conditional statement "Please answer concisely. Your current location is as follows," and location information (location information of vehicle V1 obtained from navigation device 6) (no attached image).
[0023] In the prompt table 121 in Figure 3, the content of the condition statement is the number of trivia facts, but the content of the condition statement is not limited to the number of trivia facts. For example, it could be a constraint on the number of characters in the answer, or a constraint on the number of vowels in the answer.
[0024] Furthermore, in the prompt table 121 in Figure 3, a single prompt may contain two, one, or zero images, but the number of images in a single prompt is not limited to two, one, or zero.
[0025] By including the conditional statement described above in the prompt, the guidance device 1 can obtain guidance information (the answer from the answer generation device 2) based on the power margin for the answer generation device 2. Specifically, the lower the power margin for the answer generation device 2, the less information the guidance device 1 can obtain (the answer from the answer generation device 2). The less information the guidance information (the answer from the answer generation device 2) contains, the less computation is required for the LLM222 (see Figure 6) included in the answer generation device 2, which reduces the power consumption of the answer generation device 2 and enables good power consumption management.
[0026] As described above, the power consumption of the answer generation device 2 can be controlled by the guidance device 1 adjusting the number of images included in the prompt. Specifically, the guidance device 1 reduces the number of images included in the prompt as the power margin for the answer generation device 2 decreases. The fewer images included in the prompt, the less computation is required for the LLM222 (see Figure 6) included in the answer generation device 2, thereby suppressing the power consumption of the answer generation device 2 and enabling good power consumption management.
[0027] The power margin table 122 is a table that links the degree of power margin (power margin for the answer generation device 2) with the parameters used to determine the power margin, as shown in Figure 4, for example. In the power margin table 122 of Figure 4, the only parameter used to determine the power margin is the "remaining charge of the auxiliary battery." The auxiliary battery is the battery that supplies power to the answer generation device 2. In addition to the answer generation device 2, the auxiliary battery also supplies power to other devices such as the navigation device 6, headlights, and air conditioner.
[0028] If vehicle V1 is an HEV (Hybrid Electric Vehicle) or BEV (Battery Electric Vehicle), vehicle V1 is equipped with a drive battery in addition to an auxiliary battery. The drive battery supplies power to the drive motor. If vehicle V1 is an HEV or BEV, the battery that supplies power to the air conditioner compressor may be the drive battery instead of the auxiliary battery.
[0029] There may be multiple parameters for determining the power margin in the power margin table 122. If there are multiple parameters for determining the power margin in the power margin table 122, the power margin table 122 becomes a multidimensional table. The specific contents of the power margin table 122 differ depending on the type of vehicle V1.
[0030] The parameters for determining the power margin may include the type of road on which vehicle V1 travels (for example, two types: general roads and highways). This is because the regenerative energy generation differs depending on the type of road on which vehicle V1 travels, and therefore the type of road on which vehicle V1 travels is likely to affect the power margin (power margin for the response generation device 2).
[0031] The parameters used to determine the power margin may include the remaining charge of the drive battery. For example, if the generator, powered by the power output from the drive battery, charges the auxiliary battery, then the remaining charge of the drive battery will affect the power margin (the power margin for the answer generation device 2).
[0032] The parameters used to determine the power margin may include whether or not the generator is operating. For example, if the generator is configured to charge the auxiliary battery, then whether or not the generator is operating will affect the power margin (the power margin for the answer generation device 2).
[0033] The parameters used to determine the power margin may include the remaining fuel level of vehicle V1. For example, if the generator (alternator) operated by the engine's rotation charges the auxiliary battery, the remaining fuel level of vehicle V1 will affect the power margin (power margin for the response generation device 2).
[0034] Alternatively, instead of the power margin table 122, a calculation formula represented by parameters for determining the power margin may be used, and the power margin quantified by this formula may be calculated.
[0035] Program 123 is a program that should be executed by controller 13.
[0036] The controller 13 comprehensively controls the operation of each part of the guide device 1. The controller 13 is equipped with a processing unit including a CPU (Central Processing Unit) as a hardware resource. The controller 13 has functional blocks 131 to 134.
[0037] Controller 13 is a program execution device (computer) capable of executing any program. The controller 13 executes program 123, thereby realizing each function of the controller 13 (including the functions of function blocks 131 to 134). All operations of the controller 13 described in this embodiment may be operations realized by the controller 13 executing program 123. Program 123 may consist of multiple programs.
[0038] The following describes each functional block of the controller 13. Functional blocks 131 to 134 are the text conversion unit, image acquisition unit, prompt generation unit, and guidance unit, respectively.
[0039] The text conversion unit 131 corresponds to the voices emitted by the occupants of the vehicle V1 and converts the voice data output from the microphone 3 into text data.
[0040] The image acquisition unit 132 acquires images captured by the drive recorder 5.
[0041] The prompt generation unit 133 generates prompts based on the power margin for the answer generation device 2 according to the prompt table 121.
[0042] The guidance unit 134 converts the response (text data) into audio data and transmits it to the speaker 4. In this embodiment, the guidance unit 134 outputs guidance information (response) by voice, but output by display may be performed instead of voice output. That is, the guidance unit 134 may transmit the response (text data) to a display device installed in the vehicle V1 or a display device brought into the vehicle V1. Furthermore, the response from the response generation device does not have to be text data, but may be image data, video data, or audio data. In that case, the guidance unit 134 transmits and outputs the acquired guidance information (response) to the display device or speaker 4 of the vehicle V1.
[0043] [Answer generation device] Figure 5 shows an example of the configuration of the answer generation device 2. The answer generation device 2 comprises a communication unit 21, a storage unit 22, and a controller 23.
[0044] The communication unit 21 transmits and receives arbitrary signals with the guidance device 1. The controller 23 can also transmit and receive arbitrary information with the guidance device 1 using the communication unit 21; however, the description of the communication unit 21 may be omitted below.
[0045] The storage unit 22 is configured to include non-volatile memory such as ROM or flash memory, and volatile memory such as RAM. The storage unit 22 stores the program 221 and LLM 222.
[0046] Program 221 is a program that should be executed by controller 23. LLM222 is a multimodal LLM capable of image recognition.
[0047] The controller 23 comprehensively controls the operation of each part in the answer generation device 2. The controller 23 is equipped with a processing unit including a CPU as a hardware resource. The controller 23 has an LLM execution unit 231.
[0048] When the LLM execution unit 231 receives a prompt from the guidance device 1, it uses the prompt as input data to execute the LLM 222 and causes the LLM 222 to generate a response. Once the LLM 222 generates a response, the controller 23 transmits the response generated by the LLM 222 to the guidance device 1.
[0049] [Information processing] Figure 6 is a flowchart of the guidance process executed by the controller 13. This guidance process is achieved when the controller 13 executes the program 123 described above. The guidance process shown in Figure 6 starts when power is turned on to the guidance device 1.
[0050] First, in step S10, the controller 13 analyzes the audio data sent from microphone 3 to determine whether the audio collected by microphone 3 contains a keyword that instructs the system to begin accepting questions from the user. If the audio collected by microphone 3 contains the keyword and the keyword is detected, the system proceeds to step S20.
[0051] In step S20, the controller 13 analyzes the audio data sent from the microphone 3 to determine whether the audio collected by the microphone 3 contains the user's voice (a question from the user). If the audio collected by the microphone 3 contains a question from the user and a question from the user is detected, the process proceeds to step S30. If no question from the user is detected, the process proceeds to step S30, where the controller 13 determines whether a certain timeout period has elapsed since the keyword was detected. If a certain timeout period has not elapsed since the keyword was detected, the process returns from step S30 to step S20. If a certain timeout period has elapsed since the keyword was detected, the process returns from step S30 to step S10.
[0052] In step S40, the controller 13 obtains information on the remaining charge of the auxiliary battery from, for example, the ECU (Electronic Control Unit) installed in the vehicle V1 that manages the auxiliary battery, and uses the power margin table 122 to determine the power margin from the remaining charge of the auxiliary battery. Once the power margin is determined, the process proceeds to step S50.
[0053] In step S50, the controller 13 (text conversion unit 131) converts the user's question into text data. The controller 13 (image acquisition unit 132) also acquires images taken by the drive recorder 5 at the time the user's question is detected in step S20, in a number corresponding to the power margin, and uses these as prompt attachment images. Then, the controller 13 (prompt generation unit 133) generates a prompt according to the prompt table 121 based on the power margin. Once prompt generation is complete, the process proceeds to step S60. For example, if the user's question is "What is that red tower?", and the power margin is "high", the controller 13 generates the prompt "What is that red tower? Please also provide three fun facts."
[0054] In step S60, the controller 13 sends the prompt and its attached image to the answer generation device 2. Once the prompt has been sent, the process proceeds to step S70. For example, if the power margin is "high", the prompt and image generated in step 50 are sent to the answer generation device 2.
[0055] In step S70, the controller 13 (guidance unit 134) determines whether it has received a response from the response generation device 2. If a response is received, the process proceeds to step S80.
[0056] In step S80, the controller 13 (guidance unit 134) converts the response (text) received in step S70 into audio data and transmits the converted audio data to the speaker 4. After step S80, the process returns to step S10. For example, it may receive a response such as, "This is Kobe Port Tower. It is a tower-type observation facility located in Chuo-ku, Kobe City, Hyogo Prefecture, Japan. It is 108 meters tall and opened in 1995." If the received response is image data, video data, or audio data, the acquired response (guidance information) is transmitted and output to the display device or speaker 4 of the vehicle V1.
[0057] <Notes, etc.> The various technical features disclosed in the embodiments for carrying out the invention as specified herein can be modified in various ways without departing from the spirit of the technical creation. Furthermore, the multiple embodiments and modifications disclosed in the embodiments for carrying out the invention as specified herein may be combined to the extent possible.
[0058] In the embodiment described above, the guidance device 1 controlled the power consumption of the answer generation device 2 by adjusting the number of images included in the prompt. However, the guidance device 1 may also control the power consumption of the answer generation device 2 by adjusting the resolution of the images included in the prompt. Specifically, the lower the power margin for the answer generation device 2, the lower the resolution of the images included in the prompt may be.
[0059] Furthermore, if vehicle V1 is a BEV, the parameters for determining the power margin may include the presence or absence of charging facilities along the route guided by the navigation device 6.
[0060] The guidance system SYS1 may have a configuration such as that shown in Figure 7. In the configuration example shown in Figure 7, the guidance system SYS1 includes a server SV1. Server SV1 has a configuration that is basically the same as that of the answer generation device 2, but differs from the answer generation device 2 in that it communicates with the guidance device 1 via the network NT1. The LLM included in server SV1 may have the same structure as the LLM included in the answer generation device 2, or it may have a different structure.
[0061] In the SYS1 guidance system with the configuration example shown in Figure 7, the guidance device 1 can send a prompt to the server SV1 instead of the response generation device 2 when a large amount of informational guidance is required, even when the power margin is "medium" or "low". Specifically, for example, by analyzing the voice data sent from microphone 3, if the voice collected by microphone 3 contains words such as "details" or "in detail", and the power margin is "medium" or "low", then a prompt can be generated using the prompt structure corresponding to the case of a "high" power margin in the prompt table 121, and the generated prompt can be sent to the server SV1. This makes it possible to obtain a large amount of informational guidance without increasing the power consumption of the response generation device 2. [Explanation of Symbols]
[0062] 1. Guidance device 2...Answer generation device SYS1... Guidance System
Claims
1. A guidance device that outputs guidance information, A prompt is generated to obtain the guidance information for the amount of information based on the power margin for the answer generation device including LLM, The generated prompt is sent to the response generation device. A guidance device that outputs the guidance information using the response from the response generation device.
2. The aforementioned guidance device and the aforementioned response generation device are mounted on the vehicle. The guidance device according to claim 1, wherein the parameter for determining the power margin includes the type of road on which the vehicle travels.
3. The aforementioned guidance device, the aforementioned response generation device, and the battery are mounted on the vehicle. The guidance device according to claim 1, wherein the parameter for determining the power margin includes the remaining charge of the battery.
4. The aforementioned guidance device, the aforementioned response generation device, and the generator are mounted on the vehicle. The guidance device according to claim 1, wherein the parameter for determining the power margin includes whether or not the generator is operating.
5. The aforementioned guidance device and the aforementioned response generation device are mounted on the vehicle. The guide device according to claim 1, wherein the parameter for determining the power margin includes the remaining fuel amount of the vehicle.
6. The aforementioned LLM is a multimodal LLM capable of image recognition. The guidance device according to any one of claims 1 to 5, wherein the number of images attached to the prompt is changed based on the power margin.
7. A prompt is generated to obtain guidance information on the amount of information based on the power margin for the answer generation device, including LLM. The generated prompt is input to the response generation device. A guidance method that outputs the guidance information using the response from the response generation device.
8. To generate prompts for obtaining guidance information on the amount of information based on the power margin for the answer generation device including LLM, Sending the generated prompt to the response generation device, The guidance information is output using the response from the response generation device. A program that causes a computer to execute something.
9. A guidance system including a guidance device that outputs guidance information and a response generation device that includes an LLM, The aforementioned guide device, A prompt is generated to obtain the guidance information based on the power margin for the response generation device, The generated prompt is sent to the response generation device. A guidance system that outputs the guidance information using the response from the response generation device.