Method for optimising a function of a type of motor vehicle

Abstract

The invention relates to a method for optimising a function of a type of motor vehicle, which function has an impact on the behaviour of the vehicle perceived by a driver, the method being implemented by a processor and comprising the steps of: • implementing (203) the function in a plurality of vehicles of the same type; • collecting (205) scores associated with comments from the drivers regarding the behaviour of their vehicle with regard to the function and transmitting (207) the scores and comments to a remote device; • establishing (209) an average score for the function by aggregating the scores, and classifying (211) the comments using a large language model, or LLM, within a predefined list of criteria; • if the average score is below a predefined score, determining, using the LLM (213), the criteria receiving negative comments in order to improve said criteria. The invention further relates to a system for optimising a function.

Description

[0001] Description

[0002] Title of the invention: Method for optimizing a function of a type of motor vehicle

[0003] technical field

[0004]

[0001] The present invention claims priority from French application 2500097 filed on January 7, 2025, the content of which (text, drawings and claims) is incorporated herein by reference.

[0005] [2]The present invention relates to a method for optimizing a function of a type of motor vehicle, a computer program product and a system for optimizing a function of a type of motor vehicle.

[0006] State of the art

[0007] [3]Contemporary vehicles are made up of a large number of components, organs, and embedded systems configured to assist the driver in driving the vehicle or to accompany the driver and passengers during journeys. To this end, each embedded system or component implements one or more functions or services in the form of software modules. When developing a new function, it must be tested during the development phase before being deployed in series production on vehicles and before the marketing of vehicles equipped with this new function. These tests allow for the detection of malfunctions or bugs.

[0008] [4] The testing or evaluation of a new function is carried out internally by the car manufacturer, which sometimes uses a panel of selected users, for example, from among its customers or professional drivers. Such a testing campaign is therefore costly in terms of time and energy. Indeed, a sufficiently large panel must be defined to obtain statistically significant results, but this implies analyzing a large amount of feedback. This feedback most often consists of ratings associated with comments. The ratings can be processed using standard statistical tools. However, the comments, often written in free text, require reading and then classification by individuals, which is lengthy, tedious, and often prone to error. [5] There is therefore a real need for a process and a system for optimizing a function of a type of motor vehicle that resolves all or part of the aforementioned drawbacks.

[0009] Description of the invention

[0010] [6] To resolve one or more of the aforementioned drawbacks, according to a first embodiment, a process for optimizing a function of a type of motor vehicle, said function having an impact on the vehicle's behavior as perceived by a driver of said vehicle, is implemented by at least one processor and comprises the steps of:

[0011] • implementation of the function on a plurality of vehicles of the same type;

[0012] • collection of notes associated with driver comments on the behavior of their vehicle associated with the function and transmission of said notes and comments to a remote device via a network;

[0013] • establishment of an average function score by aggregating scores and classifying comments by a large language model, called LLM, within a predefined list of criteria;

[0014] • if the average grade is lower than a predefined grade, determination by the LLM, of the criteria collecting negative comments for improvement of said criteria.

[0015] [7]Thus, the use of an LLM advantageously makes it possible to significantly increase the capacity for processing comments, and therefore the number of testers, which makes it possible to improve the statistical quality of the tests.

[0016] [8]Special features or embodiments, usable alone or in combination, are:

[0017] • the steps are repeated after each modification of the function that improves at least one of the determined criteria;

[0018] • The LLM is configured to classify comments into positive or negative comments for a given criterion;

[0019] • When the LLM cannot classify a comment as positive or negative, the comment is passed to a human-machine interface for analysis by a human; and / or

[0020] • if the average grade is lower than the predefined grade and the LLM cannot determine a criterion for negative comments because the number of negative comments is less than a predefined number, the steps are repeated until the number of negative comments is greater than or equal to the predefined number.

[0021] [9]In a second embodiment, a computer program product includes program code instructions for implementing the above process when the program product is run on a computer.

[0022]

[0010] In a third embodiment, a system for optimizing a function of a type of motor vehicle, said function having an impact on the behavior of the vehicle as perceived by a driver of said vehicle, includes at least one processor configured to implement the above method.

[0023] Brief description of the figures

[0024]

[0011] The invention will be better understood upon reading the following description, given solely by way of example, and with reference to the figures in the appendix in which:

[0025] • [Fig 1] represents a communication and optimization system for a function according to one embodiment; and

[0026] • [Fig 2] represents a flowchart of a process for optimizing a function according to one embodiment.

[0027] Methods of implementation

[0028]

[0012] The embodiments presented below refer to a motor vehicle, a car. However, those skilled in the art understand that these are also applicable to other types of vehicles, such as vans, trucks, etc.

[0029]

[0013] A method and system for optimizing a function of a type of motor vehicle will now be described in what follows with joint reference to Figures 1 to 4. The same elements are identified with the same reference signs throughout the description that follows.

[0030]

[0014] The terms "first," "second" (or "firsts," "seconds"), etc., are used in this document by arbitrary convention to allow for the identification and distinction of different elements (such as operations, means, etc.) implemented in the embodiments described below. Such elements may be distinct or correspond to a single element, depending on the embodiment.

[0031]

[0015] Fig. 1 schematically illustrates a communication environment 101.

[0032]

[0016] The communication environment 101 includes a motor vehicle 103.

[0033] Vehicle 103 corresponds, for example, to a vehicle with a combustion engine, with electric motor(s) or even to a hybrid vehicle with a combustion engine and one or more electric motors.

[0034]

[0017] The vehicle 103 corresponds to a so-called connected vehicle in that it carries a communication system configured to communicate with one or more remote devices 105 via a wireless communication network infrastructure 107. The remote device 105 corresponds for example to a server or a computer in the "cloud".

[0035]

[0018] The communication system of a connected vehicle includes, for example, one or more communication antennas connected to a telematic control unit (TCU), which is itself connected to one or more computers of the connected vehicle's embedded system. The antenna(s), the TCU, and the computer(s) form, for example, a multiplexed architecture for providing various services useful for the proper functioning of the connected vehicle and for assisting the driver and / or passengers of the connected vehicle in controlling the vehicle and / or for diagnosing the operation of one or more components of the connected vehicle.The computer(s) and the TCU communicate and exchange data with each other via one or more computer buses, for example, a CAN (Controller Area Network) data bus, CAN FD (Controller Area Network Flexible Data-Rate), FlexRay (according to ISO 17458) or Ethernet (according to ISO / IEC 802-3).

[0036]

[0019] The environment 101 includes a data processing device 109 corresponding, for example, to a mobile communication device, such as a smartphone or tablet. According to one embodiment, the data processing device 109 corresponds to a laptop or computer.

[0037]

[0020] The data processing device 109 is advantageously configured to communicate with the remote device(s) 105 via the wireless communication network infrastructure 107.

[0038]

[0021] The environment 101 further includes an assembly 111 comprising one or more connected vehicles, such as vehicle 103 and / or one or more data processing devices, such as device 109. The connected vehicles of the assembly 111 are identical or similar to vehicle 103 in that they are of the same type, i.e. they belong to the same model, or they carry one or more embedded systems identical or similar to those embedded in vehicle 103 which are the subject of the optimization.

[0039]

[0022] The mobile communication infrastructure enabling wireless data communication between, on the one hand, the vehicle 103 and / or the data processing device 109 (as well as the vehicle(s) and / or the data processing device(s) of the assembly 111) and, on the other hand, each of the remote devices 105, comprising, for example, one or more communication devices 113 of the relay antenna type (cellular network). In a communication mode using such a network architecture, the data is, for example, transmitted by the vehicle connected to the remote device 105 of the "cloud" via a relay antenna 113 (the antenna 113 being, for example, connected to the "cloud" via a wired link and the remote device 105 being itself connected to the network infrastructure of the "cloud" via a wired and / or wireless network).

[0040]

[0023] The wireless communication system enabling the exchange of data between, on the one hand, the connected vehicle 103 and / or the data processing device 109, and on the other hand, the remote device 105, corresponds, for example, to:

[0041] • a vehicle-to-infrastructure (V2I) communication system, for example, based on the 3GPP LTE-V or IEEE 802.11p standards of ITS G5; or

[0042] • a cellular network communication system, for example an LTE (Long-Term Evolution) network, LTE-Advanced (also called LTE 3G, 4G or 5G); or

[0043] • a Wifi type communication system according to IEEE 802.11, for example according to IEEE 802.11 n or IEEE 802.11 ac.

[0044]

[0024] The optimization process of a function is implemented by the system 101 will now be described with reference to Fig. 2.

[0045]

[0025] In this context, we will focus on a function that has an impact on the behavior of the vehicle as perceived by the vehicle user, and is then able to make a judgment on it.

[0046]

[0026] Suppose, for example, that, in developing a "sport" mode, it becomes necessary to optimize several functions, such as pedal resistance, steering wheel rotation resistance, and dashboard ambient lighting. The first two functions will result in neuromuscular sensations in the driver's legs or arms, and the third in a visual sensation.

[0027] Another example is lane keeping assist (also known in English by the acronym LKA for "Lane Keeping Assist"). The functions to be optimized are then the steering wheel response, the intensity of the alert, and the number of times LKA is activated before a message appears on the dashboard. The first function will result in a neuromuscular sensation in the driver's arms, the second in an audible and / or visual sensation, and the third in a visual sensation.

[0047]

[0028] The function to be optimized (or the set of functions for the previous examples) is installed, step 201, in vehicle 103 and all vehicles of the same type in set 111.

[0048]

[0029] This function is then implemented and tested, step 203, for a period determined by the users of these vehicles.

[0049]

[0030] At the end of the testing period, users complete, in step 205, a questionnaire about their experience with the function. This questionnaire includes at least a rating associated with a comment.

[0050]

[0031] The rating is, for example, established on a scale of 1 to 10, with 1 corresponding to "does not work, or works poorly" and 10 to "works perfectly". The comment is a free text allowing each tester to express their feelings about the function.

[0051]

[0032] The questionnaire is presented and answered using the vehicle's human-machine interfaces or using the data processing devices 109.

[0052]

[0033] All completed questionnaires are transmitted, step 207, via the network to device 105.

[0053]

[0034] In device 105, all the questionnaires are aggregated by two parallel steps.

[0054]

[0035] In step 209, an average score for the function is established by statistical calculation of the provided scores. This step 209 may also include other statistical operations on the scores, such as distribution, standard deviation, median...

[0055]

[0036] In step 211, the comments are classified by a large language model, called LLM (from the English "Large language model"), within a list of predefined criteria.

[0056]

[0037] Such an LLM, also called a generative LLM agent, is, for example, implemented in the form of a neural network that has undergone deep learning, as known to a person skilled in the art.

[0038] Such an LLM is configured beforehand by giving it instructions in the form of a text called a "prompt".

[0057]

[0039] An example of a prompt is described below:

[0058] • You are an expert in functions;

[0059] • You will receive feedback from customer reviews of features;

[0060] • Based on these comments and depending on the function tested, you will classify these comments;

[0061] • You will then identify whether these comments are positive or negative, and then, for the negative comments, you will classify all or part of these comments according to criteria. These criteria will be provided to you depending on the function being tested;

[0062] • During the classification stage, you will take into account the following guidelines from the experts:

[0063] • You will give an average score to each criterion;

[0064] • You will identify the criteria that have a score below a predefined score and that need to be improved;

[0065] • If you don't know how to classify, you will identify these comments as "To be classified by an expert";

[0066] • You will define how to modify the parameters to try to make the comment positive during the next test iteration.

[0067]

[0040] Using the previous instructions, the LLM therefore tries to classify the comments, step 211-1.

[0068]

[0041] If classification is possible, the LLM will identify positive comments, for example by noting them "1", and negative comments, for example by noting them "0"; step 211-3.

[0069]

[0042] Then, the LLM will classify the comments according to criteria predefined by experts and provided in the prompt, step 211-5.

[0070]

[0043] If classification is not possible for any reason, the comment is provided to an expert in step 211-7 for manual classification. In this case, the expert's classification is fed back into the LLM for use in a subsequent iteration.

[0071]

[0044] The LLM then performs, in step 213, a synthesis of the classified comments and proposes a modification of the function to optimize it in the direction of minimizing negative ratings and comments.

[0045] Depending on the function, it is modified directly by the LLM within limits defined by experts or the modification is transferred to development teams.

[0072]

[0046] For example, in LKA detection, 60% of testers found the steering wheel response too slow. The LLM will then modify the embedded software settings by increasing the steering wheel response speed by two levels.

[0073]

[0047] More specifically, the result of the analysis leads to three situations:

[0074] • The average rating from testers and the rating generated by the LLM for comments are positive. The function is satisfactory and no optimization action is required;

[0075] • The average tester rating is negative, but the rating generated by the LLM for the comments is positive. It is not possible to draw any conclusions about the necessary changes, and a new round of testing must be initiated; or

[0076] • The average rating from testers and the rating generated by the LLM for the comments are negative. The function is optimized based on tester feedback, as summarized by the LLM.

[0077]

[0048] The system and method described thus advantageously allow a large quantity of comments to be manipulated and therefore improve the results of the vehicle's field tests.

[0078]

[0049] Fig. 1 illustrates a system according to certain embodiments. The breakdown presented is for pedagogical purposes to highlight the different functions. However, it is understood that each block can be implemented using different means or combinations thereof, such as hardware components, software, one or more computers, and / or electronic circuits. Each component can include at least one computer or a control unit. At least one memory can be included in each component. The memory can include computer program instructions or software code.

[0079]

[0050] The computers can be implemented by any type of data processing device, such as a central processing unit, a signal processing unit, a specific application integrated circuit, a programmable gate network, etc. The computers can be implemented in the form of a single controller, or a plurality of controllers or computers.

[0080]

[0051] The various modules are connected to each other by data links adapted to the environment. These can be wired or wireless.

[0052] For the software, the implementation can include modules or units distributed in the form of procedures, functions, etc. The memories can be any type of storage circuit. They can be part of the processor circuit, or separate from it and connected via electrical data links. These can be non-volatile memories, hard drives, RAM, flash memory, etc.

[0081]

[0053] The software product can be downloaded from a communication network and / or stored on a computer-readable medium. It can be directly executable by a processor or be in the form of a high-level language requiring one or more intermediate operations to be executable.

[0082]

[0054] Thus, the program instructions stored in memory and processed by the computers can be any type of program code, for example, a compiled or interpreted program written in a suitable programming language.

[0083]

[0055] The computer program instructions stored in memory are such that, when executed by the computer, the latter carries out one or more of the steps of the processes described above.

[0084]

[0056] The invention has been illustrated and described in detail in the drawings and the preceding description. This description is to be considered illustrative and given by way of example and not as limiting the invention to this single description. Numerous embodiments are possible.

[0085] In particular, it is understood that the test and analysis of results can be generalized to a plurality of independent or correlated functions. In the latter case, the LLM's analysis of comments takes this correlation into account, for example in the form of prompt instructions.

Claims

Demands 1. A method for optimizing a function of a type of motor vehicle, said function having an impact on the vehicle's behavior as perceived by a driver of said vehicle, and said method being implemented by at least one processor and comprising the steps of: • implementation (203) of the function on a plurality of vehicles of the same type; • collection (205) of notes associated with comments from drivers on the behaviour of their vehicle associated with the function and transmission (207) of said notes and comments to a remote device via a network; • establishment (209) of an average score of the function by aggregation of the scores and classification (211) of the comments by a large language model, called LLM, within a predefined list of criteria; • if the average score is less than a predefined score, determination by the LLM (213) of the criteria collecting negative comments for improvement of said criteria.

2. A method according to claim 1, wherein the steps are repeated after each modification of the function improving at least one of the determined criteria.

3. Method according to claim 1 or 2, wherein the LLM is parameterized to classify, for a given criterion, the comments into positive comments or negative comments.

4. A method according to claim 3, wherein, when the LLM cannot classify a comment as positive or negative, the comment is transmitted to a human-machine interface for analysis by a human.

5. A method according to claim 1, 2, 3 or 4, wherein, if the average score is less than the predefined score and the LLM cannot determine a criterion with negative comments because the number of negative comments is less than a predefined number, the steps are repeated until the number of negative comments is greater than or equal to the predefined number.

6. Computer program product characterized in that it includes program code instructions for implementing the method according to any one of claims 1 to 5 when the program product is executed on a computer.

7. System for optimizing a function of a type of motor vehicle, said function having an impact on the behavior of the vehicle as perceived by a driver of said vehicle, said system comprising at least one processor configured to implement the method according to any one of claims 1 to 5.

Images