Computer-implemented method for calibrating an engine control unit of a motor vehicle

Abstract

The present invention relates to a computer-implemented method for calibrating an engine control system of a motor vehicle (1), comprising the following steps: - receiving a calibration target, wherein the calibration target comprises a target difference between a target torque to be exerted on wheels of the motor vehicle (1) and a torque fictitiously requested by the engine control unit from an engine of the motor vehicle; - Calibration of the engine control system taking into account the calibration target, whereby at least one parameter of the engine control system is changed during the calibration; thereafter - Determination of a calibration difference between the target torque and a torque currently requested by the engine control unit from the engine of the motor vehicle (1); - Output of the calibration difference as a calibration difference image (16); - generating a description of the calibration difference image (16) in human language; and evaluating the calibration using the description.

Description

The present invention relates to a computer-implemented method for calibrating an engine control of a motor vehicle according to claim 1.It is known from the prior art to calibrate an engine control of a motor vehicle. In this case, different parameters of the engine control are usually changed, so that the torque requested by the engine control by the engine changes under different circumstances. The aim is often to achieve a certain driving feel for a user.US 2022 / 0126864 A1 discloses a method in which sensor data are received from a plurality of sensors. The plurality of sensors includes a first set of sensors and a second set of sensors. At least a portion of the plurality of sensors is connected to a vehicle. Control of the vehicle is automatically based on at least a portion of the sensor data generated by the first group of sensors. Passenger attributes of one or more passengers in the autonomous vehicles are determined from sensor data generated by the second group of sensors. Attributes of the vehicle are changed based on the passenger attributes and the sensor data generated by the first group of sensors.DE 10 2022 132 803 A1 discloses a computer-implemented method for calibrating an active system. Complex measurement data from an operation of an active system are provided by means of a measurement device. A first human-understood phrase having computer-understood information is generated. A software agent is used to determine a similarity of the computer-understood information and a predetermined target specification. By means of the software agent, a reward for a machine learning algorithm for calibrating the active system is generated as a function of the determined similarity.DE 10 2022 117 623 A1 discloses a computer-implemented method for adapting calibration data of a control unit for an electric drive of a motor vehicle. It comprises using parameters and the calibration data by the control unit for calculating calculated values and measuring measured values during operation of the drive. The measured values depend on the parameters. A difference between the calculated values and the measured values is detected. An algorithm is used to reduce the difference by changing the calibration data. An instruction for a procedure for changing the calibration data is created by an artificial intelligence. The instruction is transmitted from the artificial intelligence to the algorithm. A description of the instruction is issued in human language by the artificial intelligence. The calibration data is modified using the instruction through the algorithm.DE 10 2022 123 578 A1 discloses a method for tuning motor vehicles. A computer-linguistic model is used to formulate a driving behavior-related objective. A technical agent takes a calibration action with a goal-directed strategy that places the motor vehicle in a following state. The driving behavior of the motor vehicle is subjected to a judgment in the subsequent state. Based on the assessment, an appropriate reward for the action taken is determined by means of a statistical model. Depending on the reward, the agent changes its strategy.In contrast, the present invention is based on the object of making possible an evaluation of the calibration using human speech.This object is achieved by a method according to claim 1. Embodiments of the invention are set forth in the dependent claims.The method includes receiving a calibration target. The calibration target includes a target difference between a target torque to be applied to wheels of the motor vehicle and a fictitious torque requested by the engine controller from an engine of the motor vehicle. In this case, the target torque can be, for example, a torque desired under the current circumstances. The torque demanded by the motor controller may differ therefrom, as the demanded torque is dependent on several parameters such as the state of charge, temperature and current of the battery from which the motor draws its energy, the speed and the torque of the motor. In a perfectly calibrated engine control, there would be no difference between the target torque and the requested torque. In practice, however, this is achieved most rarely.The engine controller is calibrated in consideration of the calibration target. At least one parameter of the engine control is changed. For example, the calibration target can be taken into account by attempting to reach the calibration target by changing the at least one parameter or at least coming closer to it.After calibration, a calibration difference is determined between the target torque and the torque currently requested by the engine controller from the engine of the motor vehicle. The calibration difference differs from the target difference in particular in that the calibration difference is the current difference between the target torque and the requested torque, whereas the target difference is the difference between the target torque and the requested torque that is sought after completion of the method. It should be noted that although the target difference is sought, it does not necessarily have to be reached. The calibration difference is output as a calibration difference image. The calibration difference image may include, for example, a graph. A description of the calibration difference image is generated in human speech. The calibration is then calibrated using the description.It is advantageous to use the human language description for the assessment, since then further human language criteria can be used and compared with the description of the calibration difference image during the assessment. For example, a statement of a human being or another criterion formulated in human speech can also be used for the evaluation.According to an embodiment of the invention, the calibration target may be received as a calibration target image. For example, the calibration target image may include or be a graphical representation of the calibration target. In this embodiment, it is possible that a description of the calibration target image is generated in human speech. This description of the calibration target image can also be used in evaluating calibration.According to one embodiment of the invention, a user rating of the calibrated motor controller may be received from a user in human speech. The user assessment may be used in assessing calibration. The user assessment can relate, for example, to a driving feeling that he had during the travel with the motor vehicle. For receiving the user assessment, a microphone of the motor vehicle and / or a microphone of a portable device of the user, for example a smartphone or a smart watch, can be used, for example. For example, the user assessment can comprise a change request. The feeling of driving can be described as comfortable or sporting, for example. The change request can be, for example, that a more comfortable or more sportier driving feel is to be achieved.According to an embodiment of the invention, the calibration target can be adjusted depending on the user's evaluation. If the user assessment comprises, for example, a change request, the calibration target can be adapted with respect to the change request. For example, if the user desires a more sporting drive feel, the calibration target may be adjusted to achieve the more sporting drive feel at the next calibration.According to an embodiment of the invention, a mood state of the user can be determined. The mood state may be used in evaluating the calibration. Within the scope of this description, the mood is understood to mean, for example, fun, stress, anxiety, food, longwell or relaxation. For example, the mood state can be determined from a statement of the user. The statement can describe or name, for example, the mood state. The statement can be received, for example, via a microphone of the motor vehicle and / or a microphone of a portable device of the user, for example a smartphone or a smart watch. For example, the user may say that he has fun or anxiety or is drowsy.According to an embodiment of the invention, the mood may be determined based on a facial expression of the user, a blood pressure of the user, a body temperature of the user, and / or a heart rate of the user. The facial expression can be captured by a camera, for example. The blood pressure, the body temperature, and the heart rate may be detected via a smart watch.According to an embodiment of the invention, the mood state and / or the user assessment can be summarized in human speech by an artificial intelligence. This allows the mood state and / or user assessment to be used also in assessing calibration in a similar manner to the description of the calibration difference image or other assessment criteria formulated in human language.According to an embodiment of the invention, the calibration may be performed by an algorithm using reinforcement learning. The algorithm can use an agent which changes the at least one parameter of the engine control as an action.According to an embodiment of the invention, a similarity between the calibration target, the calibration difference, the user score and / or the mood state may be determined. This is particularly advantageous if information about the calibration target, the calibration difference, the user assessment and / or the mood state is present in human speech. For the calibration target, this may be, for example, a description of the calibration target image. For the calibration difference, this may be a description of the calibration difference image.From the similarity, a reward for the agent may be determined. The reward may be output to the agent in response to the action. For example, the reward may be positive if the similarity has increased compared to a previous calibration. The reward may be negative if the similarity has decreased compared to a previous calibration. Thus, the similarity between the above-mentioned criteria is increased when the method is carried out a plurality of times. This calibrates the engine controller in a lensed manner and is less dependent on the user rating. This is particularly advantageous when the user is a skilled person, for example an engineer, who calibrates the engine control for a series vehicle.The similarity can be, for example, a numerical value of the metrics for lexical and contextual similarity recognition. Examples of this are dot product, cosine similarity and Jaccard similarity.According to one embodiment of the invention, the calibration can be carried out several times. It is also possible that the determination of the calibration difference, the output of the calibration difference, the generation of the description of the calibration difference image and the evaluation of the calibration are carried out a plurality of times. The method is ended when the similarity exceeds a similarity threshold value. The similarity threshold value can be 0.9, for example, if the similarity can assume a value between 0 and 1.Further features and advantages of the present invention will become apparent from the following description of preferred embodiments with reference to the accompanying drawings. The same reference numerals are used for identical or similar components, features or elements and for components, features or elements having identical or similar functions. This shows FIG. 1 is a schematic illustration of a method according to an embodiment of the invention.A user 2 uses a motor vehicle 1 to calibrate the engine control of the motor vehicle 1. The user 2 outputs a user rating 10 of the engine control in human speech, which is received via a microphone. The user assessment can, for example, state that the engine control has been perceived as being too uncomfortable or too uncomfortable. It is also possible that the user rating is that the engine control has been perceived as too sporting.In addition, a facial expression 9 of the user is detected via a camera 3 and an electronic device 4, for example a smart watch, vital parameters 8, for example a blood pressure of the user, a body temperature of the user and / or a heart rate of the user, are detected. From the facial expression 9 and the vital parameters 8, a mood state 12 of the user is determined and described in human speech. An algorithm for this is already known from: H. Z. Wijasena, R. Ferdiana and S. Wibirama, "A Survey of Emotion Recognition using Physiological Signal in Wearable Devices," 2021 International Conference on Artificial Intelligence and Mechatronics Systems (AIMS), Banding, Indonesia, 2021, pp. 1-6, doi: 10.1109 / AIMS52415.2021.946609 2A calibration target is graphically depicted in a calibration target image 5. The calibration target includes a target difference between a target torque to be applied to wheels of the motor vehicle and a torque requested by the engine controller from an engine of the motor vehicle. The target difference may be represented depending on various parameters such as the state of charge and temperature of the battery that supplies the motor with electric power, a speed, and a torque.A description 7 of the calibration target image 5 in human speech is generated. This can be carried out, for example, by a correspondingly trained and already known artificial intelligence 6.The artificial intelligence 6 also creates a human-language description of the calibration difference image 16. The calibration difference image 16 depicts a difference between the target torque and the torque requested by the engine controller at the current settings. The difference may be represented, for example, depending on various parameters such as the state of charge and the temperature of the battery that supplies the motor with electrical energy, a rotational speed and a torque. The calibration difference image 16 differs from the calibration target image 5 in particular in that the calibration difference image 16 relates to the current calibration of the motor controller, whereas the calibration target image 5 relates to a desired calibration of the motor controller, which, however, does not necessarily have to be achieved, although it can be achieved.The human language user rating 10, the human language mood state 12 and the human language description 6 of the calibration difference image 16 are merged into a summary 13 using, for example, a large language model such as GPT 3 or GPT 4, and fed to a reward generator 14 of an algorithm that uses reinforcement learning. In addition, the description 7 is supplied in human language to the reward generator 14. The algorithm was trained to determine the similarity of the input data supplied to it. This can be carried out, for example, using Bert, GEMINI, GPT 3, GPT 4 or other algorithms.The reward generator 14 determines a similarity between the description of the calibration target image 5, the calibration difference image 16, the user score 10, and the mood condition 12, and generates a reward for the agent 15 of the algorithm calibrating the engine controller of the motor vehicle 1 depending on the similarity. With a high similarity, a positive reward is output to agent 15. With a low similarity, a negative reward may be output to agent 15. After calibration, the method begins again with the re-calibrated engine controller. The similarity may be, for example, a cosine similarity or a Jaccard similarity. The method is performed until the similarity exceeds a similarity threshold.A high similarity can be determined, for example, between the user assessment "The driving feel is comfortable" and the calibration target of an engine control system that is perceived as comfortable. A low similarity can be determined between the user rating "The driving feel is too sporting" and the calibration target of an engine control that is perceived as comfortable.Thus, the description 7 of the calibration difference image 16 in human speech, among other things, is used to evaluate the calibration of the motor controller. The assessment is carried out in the form of the reward of the agent 15.

Claims

Computer-implemented method for calibrating an engine control of a motor vehicle (1), comprising the following steps: - receiving a calibration target, wherein the calibration target comprises a target difference between a target torque which is to be exerted on wheels of the motor vehicle (1) and a torque which is notionally requested by the engine control from an engine of the motor vehicle; - calibrating the engine control taking into account the calibration target, wherein at least one parameter of the engine control is changed during the calibration; subsequently - determining a calibration difference between the target torque and a torque which is currently requested by the engine control from the engine of the motor vehicle (1); - outputting the calibration difference as a calibration difference image (16); - generating a description of the calibration difference image (16) in human speech; and - evaluating the calibration using the description.Method according to claim 1, characterized in that the calibration target is received as a calibration target image (5).Method according to one of the preceding claims, characterized in that a user rating (10) of the calibrated motor control is received by a user in human speech, wherein the user rating (10) is used in the rating of the calibration.Method according to the preceding claim, characterized in that the calibration target is adjusted as a function of the user evaluation (10).Method according to one of the preceding claims, characterized in that a mood state (12) of the user is determined, wherein the mood state (12) is used in the evaluation of the calibration.Method according to the preceding claim, characterized in that the mood state (12) is determined on the basis of a facial expression of the user, a blood pressure of the user, a body temperature of the user and / or a heart rate of the user.Method according to one of the preceding four claims, characterized in that the mood state (12) and / or the user assessment (10) in human speech are combined by an artificial intelligence.Method according to one of the preceding claims, characterized in that the calibration is carried out by an algorithm which uses reinforcement learning, the algorithm using an agent (15) which changes the at least one parameter of the engine control as an action.Method according to the preceding claim, characterized in that a similarity between the calibration target, the calibration difference, the user assessment (10) and / or the mood state (12) is determined, wherein a reward for the agent (15) is determined from the similarity, wherein the reward is output to the agent (15) in response to the action.Method according to the preceding claim, characterized in that the calibration is carried out a plurality of times, the method being ended if the similarity exceeds a similarity threshold value.

Images