Control device, control method, and storage medium

Abstract

The problem to be solved by the present invention is to provide a control device capable of suppressing the influence of noise or defects of input information and achieving control with little delay. In order to solve the problem, a vehicle control device (1) is provided with a target control amount calculation unit (11) that calculates a target control amount with respect to a control amount prescribed as a control target on the basis of time-series data of input information. A target control amount calculation unit (11) is provided with: an input information acquisition unit (2) that acquires input information at M (M is an integer of 2 or more) different times from the current time to the time before a specific time; a control quantity predictor (30,..., 3M-1) that calculates a predicted control quantity on the basis of the input information; and a weighted average calculator (4) that calculates, as a target control quantity, a weighted average of the M predicted control quantities calculated by the control quantity predictors. An i-th (i is an integer between 0 and M-1) control quantity predictor (3i) calculates a predicted control quantity by using a control quantity prediction model that associates input information before a prediction time [tau] i from a current time with a control quantity at the current time.

Description

Technical Field

[0001] The present invention relates to a control device, a control method, and a storage medium. More specifically, the present invention relates to a control device, a control method, and a storage medium for controlling a control variable specified for a control object. Background Art

[0002] In recent years, efforts have been underway to provide access to sustainable transportation systems that take into account vulnerable groups among road users. To achieve this goal, research and development related to autonomous driving technology is being pursued to further improve traffic safety and convenience.

[0003] For example, Patent Document 1 describes an autonomous driving technology in which a control device automatically controls the steering angle without relying on steering operations performed by the driver. The control device described in Patent Document 1 determines a target steering angle relative to the vehicle's steering angle so that the vehicle travels along an arc passing through the vehicle's current location and a target location specified on a target path.

[0004] [Prior Art Literature]

[0005] (Patent Document)

[0006] Patent Document 1: Japanese Patent Application Laid-Open No. 2021-126925 Summary of the Invention

[0007] [Problems to be solved by the invention]

[0008] However, many autonomous driving technologies calculate a target controlled variable (a target steering angle in the example of Patent Document 1) relative to the controlled variable based on information acquired by external sensors such as cameras and radars mounted on the vehicle. However, because the information acquired by these external sensors contains observation noise or defects, the target controlled variable may fluctuate due to the influence of noise or defects from the external sensors.

[0009] As a method for smoothing the vibration of the target control amount, a finite impulse response (FIR) filter is known. However, when an FIR filter is applied, although smoothing can be achieved, a delay will also occur (see the following). Figure 4 ).

[0010] The object of the present invention is to provide a control device, a control method and a storage medium that can suppress the influence of noise or defects in input information while achieving control with less delay, further contributing to the development of sustainable transportation systems.

[0011] [Technical means to solve the problem]

[0012] (1) A control device of the present invention (e.g., the vehicle control device 1, 1A, 1B described below) calculates a target control amount relative to a control amount prescribed for a control object (e.g., the braking device 7, the power equipment 8, and the electric power steering device 9 described below) based on input information, and operates the control object based on the target control amount. The control device is characterized in that it includes: a target control amount calculation unit (e.g., the target control amount calculation unit 11, 11A, 11B described below) that calculates the target control amount based on time series data of the input information; and an automatic operation unit (e.g., the motor drive unit 12 described below) that operates the control object based on the target control amount; and the target control amount calculation unit includes: an input information acquisition unit (e.g., the input information acquisition unit 2 described below) that acquires the input information at M different times (M is an integer greater than or equal to 2) between the current time and a specific time before; M control amount prediction units (e.g., the control amount predictors 30, 31, ..., 3 M-1 ), calculates the predicted control amount based on the input information obtained by the input information obtaining unit; and, a weighted average calculation unit (for example, the weighted average calculator 4, 4A described below), calculates the weighted average of the M predicted control amounts calculated by the M control amount prediction units as the target control amount; and, the i-th (i is an integer between 0 and M-1) control amount prediction unit calculates the predicted control amount by using a control amount prediction model, wherein the control amount prediction model associates the input information i hours before the current moment with the control amount at the current moment.

[0013] (2) At this time, it is preferred that the aforementioned control quantity prediction model is constructed using machine learning, and the aforementioned machine learning uses input sample data and ideal output data as training data, the aforementioned input sample data is the time series data of the aforementioned input information, and the aforementioned ideal output data is the time series data of the ideal control quantity relative to the input sample data.

[0014] (3) At this time, it is preferred that the i-th aforementioned control quantity prediction model is constructed using the aforementioned training data, and the aforementioned training data is the aforementioned input sample data and the aforementioned ideal output data with the time advanced i hours relative to the input sample data as a data set.

[0015] (4) At this time, it is preferred that the weighted average calculation unit sets the i-th weight corresponding to the predicted control quantity calculated by the i-th control quantity prediction unit to a value greater than the j-th weight corresponding to the predicted control quantity calculated by the j-th (j is an integer greater than i) control quantity prediction unit.

[0016] (5) At this time, it is preferable that the weighted average calculation unit sets the value of the kth (k is an integer between 0 and M-1) weight to decrease exponentially with respect to the value k.

[0017] (6) At this time, it is preferred that the aforementioned target control amount calculation unit also has a credibility acquisition unit (for example, the credibility acquisition unit 2A described below), which acquires the credibility of the aforementioned input information i hours ago from the aforementioned current moment, and the aforementioned weighted average calculation unit sets the value of the i-th aforementioned weight based on the i-th aforementioned credibility acquired by the aforementioned credibility acquisition unit.

[0018] (7) In this case, it is preferred that the control object is a steering mechanism of the vehicle (for example, the electric power steering device 9 described below), the control amount is a steering angle generated by the steering mechanism, and the input information includes external information related to the surroundings of the vehicle.

[0019] (8) At this time, it is preferred that the aforementioned control object is the vehicle's driving device (for example, the power device 8 described below), the aforementioned control amount is the driving force generated by the aforementioned driving device, and the aforementioned input information includes external information related to the surroundings of the aforementioned vehicle.

[0020] (9) In this case, it is preferred that the control object is a braking device of the vehicle (for example, the braking device 7 described below), the control amount is a braking force generated by the braking device, and the input information includes external information related to the surroundings of the vehicle.

[0021] (10) At this time, it is preferred that the control device further includes a learning unit that learns the control quantity prediction model based on the time series data of the control quantity and the input information during manual driving by the operating subject with the driver of the vehicle as the control object.

[0022] (11) The control method of the present invention utilizes a computer to control a control quantity specified for a control object, and the control method is characterized in that it includes the following steps: obtaining input information at M different moments (M is an integer greater than 2) between the current moment and a specific time; calculating M predicted control quantities based on the aforementioned input information at M different moments; calculating a weighted average value of the M aforementioned predicted control quantities as a target control quantity relative to the aforementioned control quantity; and, based on the aforementioned target control quantity, operating the aforementioned control object; and, in the step of calculating the aforementioned predicted control quantity, calculating the i-th aforementioned predicted control quantity by using a control quantity prediction model, wherein the control quantity prediction model establishes an association between the aforementioned input information i hours ago (i is an integer between 0 and M-1) from the aforementioned current moment and the aforementioned control quantity at the aforementioned current moment.

[0023] (12) The storage medium of the present invention stores a program, which enables a computer to control a control quantity specified for a control object. The storage medium is characterized in that the computer program enables the computer to execute the following steps: obtaining input information at M different times (M is an integer greater than 2) between the current moment and a specific time; calculating M predicted control quantities based on the input information at M different moments; calculating a weighted average of the M predicted control quantities as a target control quantity relative to the control quantity; and operating the control object based on the target control quantity; and, in the step of calculating the predicted control quantity, calculating the i-th predicted control quantity by using a control quantity prediction model, the control quantity prediction model is to establish an association between the input information i hours ago (i is an integer between 0 and M-1) from the current moment and the control quantity at the current moment.

[0024] (Effects of the Invention)

[0025] (1) In the present invention, a target control amount calculation unit calculates a target control amount relative to a control amount specified for a control object based on time series data of input information, and an automatic operation unit operates the control object based on the calculated target control amount. In addition, the target control amount calculation unit includes: an input information acquisition unit that acquires input information at M different moments between the current moment and a specific time before; M control amount prediction units that calculate a predicted control amount based on input information at each of the M moments; and a weighted average calculation unit that calculates a weighted average of the M predicted control amounts calculated by the M control amount prediction units as the target control amount. According to the present invention, by calculating the weighted average of the M predicted control amounts as the target control amount, the influence of noise or defects contained in the input information can be suppressed, and a target control amount with less vibration can be calculated. Furthermore, in the present invention, among a total of M control amount prediction units, the i-th control amount prediction unit (i is an integer between 0 and M-1) calculates a predicted control amount using a control amount prediction model that associates input information from the moment i hours before the current moment (i.e., counting from the current moment to the i-th moment in the past, assuming the current moment is the 0th moment) with the control amount at the current moment. That is, each control amount prediction unit calculates the same control amount at the current moment as the predicted control amount based on input information at different moments. Thus, according to the present invention, the effects of noise or defects contained in the input information can be suppressed while achieving control with less delay, further contributing to the development of continuous transportation systems.

[0026] (2) In the present invention, each control variable prediction unit calculates a predicted control variable based on input information using a control variable prediction model. The control variable prediction model is constructed using machine learning, which uses input sample data (time series data of the input information) and ideal output data (time series data of the ideal control variable for the input sample data) as training data. Thus, according to the present invention, the effects of noise or defects contained in the input information can be suppressed, while achieving ideal control with minimal delay.

[0027] (3) In the present invention, the i-th control variable prediction unit calculates the predicted control variable at the current time based on input information i hours prior to the current time using the i-th control variable prediction model. The i-th control variable prediction model is constructed using training data consisting of input sample data and ideal output data that is i hours ahead of the input sample data. Thus, according to the present invention, the effects of noise or defects contained in the input information can be suppressed while achieving ideal control with minimal delay.

[0028] (4) In the present invention, the weighted average calculation unit sets the i-th weight corresponding to the predicted control quantity calculated by the i-th control quantity prediction unit to a value greater than the j-th weight corresponding to the predicted control quantity calculated by the j-th control quantity prediction unit (j is an integer greater than i). In other words, the weighted average calculation unit sets the weight corresponding to the i-th predicted control quantity calculated based on input information i hours before the current moment to a value greater than the weight corresponding to the j-th predicted control quantity calculated based on input information j hours before the current moment. Generally speaking, the closer the time of input information is to the current moment, the higher the prediction accuracy of the predicted control quantity tends to be. By setting the weight to a larger value according to the improvement in prediction accuracy, a high-precision target control quantity can be calculated.

[0029] (5) In the present invention, the weighted average calculation unit sets the value of the kth weight (k is an integer between 0 and M-1) to decrease exponentially with respect to the value k. This allows the target control amount to be calculated with high precision using simple calculations.

[0030] (6) In the present invention, the credibility acquisition unit acquires the credibility of the input information i hours prior to the current moment, and the weighted average calculation unit sets the i-th weight value based on the i-th credibility acquired. This allows the credibility of the input information at each moment to be reflected, allowing the target control amount to be calculated with high precision.

[0031] (7) In the present invention, the target control amount calculation unit calculates the target control amount of the steering angle to be generated by the steering mechanism based on the time series data of input information including external information about the vehicle's surroundings, according to the above-described process. The automatic operation unit operates the steering mechanism based on the calculated target control amount. Thus, according to the present invention, the influence of noise or defects contained in the external information can be suppressed, while achieving steering control with less delay.

[0032] (8) In the present invention, the target control amount calculation unit calculates the target control amount of the travel drive force generated by the travel drive device according to the above process based on the time series data of the input information including external information about the vehicle's surroundings, and the automatic operation unit operates the travel drive device based on the calculated target control amount. Thus, according to the present invention, it is possible to suppress the influence of noise or defects contained in the external information while achieving travel drive force control with less delay.

[0033] (9) In the present invention, the target control amount calculation unit calculates the target control amount of the braking force generated by the braking device according to the above process based on the time series data of the input information including external information about the vehicle's surroundings, and the automatic operation unit operates the braking device based on the calculated target control amount. Thus, according to the present invention, it is possible to suppress the influence of noise or defects contained in the external information while achieving braking control with less delay.

[0034] (10) In the present invention, the learning unit learns the control amount prediction model based on time series data of control amounts and input information during manual driving by the vehicle driver as the control subject. Thus, according to the present invention, the input and output characteristics of the control amount prediction model can be changed according to temporal changes in the characteristics of the vehicle driver or the characteristics of the external sensor used to obtain external information.

[0035] (11) The control method according to the present invention has the same effect as the above-mentioned control device.

[0036] (12) According to the storage medium of the present invention, the same effects as those of the above-mentioned control device are achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is a diagram schematically showing the configuration of a vehicle control device and a vehicle equipped with the vehicle control device according to the first embodiment of the present invention.

[0038] Figure 2 This is a diagram showing the configuration of an automatic steering control module in a vehicle control device.

[0039] Figure 3This is a diagram comparing the target controlled amounts calculated by the target controlled amount calculation unit in Comparative Example 1 (upper section), Comparative Example 2 (middle section), and the present embodiment (lower section).

[0040] Figure 4 It is a diagram schematically showing the structure of training data.

[0041] Figure 5 FIG2 is a diagram schematically showing the structure of training data when constructing M control variable prediction models using machine learning.

[0042] Figure 6 This is a diagram showing the configuration of an automatic steering control module in a vehicle control device according to a second embodiment of the present invention.

[0043] Figure 7 This is a diagram showing the configuration of an automatic steering control module in a vehicle control device according to a third embodiment of the present invention. DETAILED DESCRIPTION

[0044] <First embodiment>

[0045] Hereinafter, a vehicle control device according to a first embodiment of the present invention will be described with reference to the drawings.

[0046] Figure 1 Schematically shows the configuration of the vehicle control device 1 and the vehicle V equipped with the vehicle control device 1 according to the present embodiment. Figure 1 The upper part of the diagram shows a top view of the vehicle V. Figure 1 The lower section shows a side view. The following description will be based on a case where the vehicle V is a so-called right-hand drive four-wheel vehicle, in which the driver's seat is located on the right side in the vehicle width direction as viewed in the direction of travel. However, the present invention is not limited to this. The vehicle V may also be a so-called left-hand drive four-wheel vehicle, in which the driver's seat is located on the left side in the vehicle width direction as viewed in the direction of travel.

[0047] The vehicle V includes: an electric power steering device 9 as a steering mechanism, which steers the left and right front wheels Wf; a power device 8 as a driving drive device, which generates a driving force for rotating the front wheels Wf serving as driving wheels in the vehicle V; a braking device 7, which generates a braking force for stopping the rotation of the front wheels Wf and the rear wheels Wr; a sensor unit 6, which is arranged on the vehicle body; and a vehicle control device 1, which controls the electric power steering device 9, the power device 8 and the braking device 7 based on the detection signal of the sensor unit 6 or the driving operation performed by the driver (for example, steering operation, acceleration and deceleration operation and braking operation, etc.).

[0048] The electric power steering device 9 includes a gear box 93 connecting a pinion shaft 92 extending from a steering wheel 91 that receives a steering operation by the driver to the left and right front wheels Wf; a motor 94 provided on the gear box 93; and a steering sensor 95 that detects the steering angle of the steering wheel 91.

[0049] The gearbox 93 includes a rack shaft extending in the vehicle width direction and meshing with the pinion shaft 92, or tie rods connecting the rack shaft's ends to the left and right front wheels Wf. It converts the rotational motion of the steering wheel 91, caused by the driver's steering operation, into motion in the vehicle width direction, thereby steering the left and right front wheels Wf in the direction of travel. The electric motor 94 rotates in response to a control signal output from the vehicle control unit 1, generating a driving force to assist the driver's steering operation or to automatically steer the front wheels Wf independently of the driver's steering operation. The steering sensor 95 detects the steering angle of the steering wheel 91 and transmits a signal corresponding to the detected value to the vehicle control unit 1.

[0050] The power unit 8 is a driving force generator that generates a driving force for rotating the front wheels Wf in response to acceleration / deceleration operations of the accelerator pedal (not shown) by the driver or control signals output from the vehicle control unit 1, thereby moving the vehicle V forward or backward in the direction of travel. The following description uses a drive motor as the power unit 8. This drive motor generates the driving force by consuming electricity supplied from a high-voltage battery or fuel cell stack (not shown), but the present invention is not limited to this embodiment. Alternatively, the power unit 8 may be an engine that generates the driving force by consuming fuel stored in a fuel tank (not shown), or a transmission that changes the speed of the engine's output and transmits it to the front wheels Wf.

[0051] The braking device 7 includes a disc brake device or a parking brake, etc. Based on the braking operation of the brake pedal (not shown) performed by the driver or the control signal output from the vehicle control device 1, the disc brake device mainly tightens the disc set on the axle of each wheel Wf, Wr when driving to generate a braking force for slowing down or stopping the rotation of each wheel Wf, Wr. The parking brake mainly generates a braking force for maintaining the rotation of each wheel Wr, Wf in a state of stopping when parking.

[0052] The sensor unit 6 includes a camera unit 61, multiple (for example, five) lidar units 62a, 62b, 62c, 62d, 62e, multiple (for example, five) radar units 63a, 63b, 63c, 63d, 63e, a vehicle body sensor 64 and an external recognition device 65.

[0053] The camera unit 61 is a camera that captures images of the front of the vehicle V. The camera unit 61 is installed, for example, on the roof of the vehicle V, near the front window on the interior side of the vehicle cabin. Images captured by the camera unit 61 are sent to the external recognition device 65 .

[0054] The LiDAR units 62a through 62e are each LiDAR (Light Detection and Ranging) units that detect objects around the vehicle V by measuring scattered light from objects in response to pulsed laser light. The first LiDAR unit 62a is located at the right corner of the front of the vehicle V when viewed in the direction of travel, and detects objects slightly to the right and forward of the vehicle V. The second LiDAR unit 62b is located at the left corner of the front of the vehicle V when viewed in the direction of travel, and detects objects slightly to the left and forward of the vehicle V. The third LiDAR unit 62c is located at the center of the rear of the vehicle V in the vehicle width direction, and detects objects behind the vehicle V. The fourth LiDAR unit 62d is located at the rear of the right side of the vehicle V, and detects objects slightly to the right and rear of the vehicle V. The fifth LiDAR unit 62e is located at the rear of the left side of the vehicle V, and detects objects slightly to the left and rear of the vehicle V. The detection signals from these LiDAR units 62a through 62e are transmitted to the external recognition device 65.

[0055] Radar units 63a through 63e are millimeter-wave radars that detect objects around the vehicle V by measuring reflected waves from objects in response to millimeter-wave irradiation. The first radar unit 63a is located at the right corner of the front of the vehicle V when viewed in the direction of travel, and detects objects slightly to the right and forward of the vehicle V. The second radar unit 63b is located at the left corner of the front of the vehicle V when viewed in the direction of travel, and detects objects slightly to the left and forward of the vehicle V. The third radar unit 63c is located at the center of the front of the vehicle V in the vehicle width direction, and detects objects slightly to the front of the vehicle V. The fourth radar unit 63d is located at the right corner of the rear of the vehicle V when viewed in the direction of travel, and detects objects slightly to the right and rear of the vehicle V. The fifth radar unit 63e is located at the left corner of the rear of the vehicle V when viewed in the direction of travel, and detects objects slightly to the left and rear of the vehicle V. The detection signals from these radar units 63a through 63e are transmitted to the external recognition device 65.

[0056] The vehicle body sensor 64 transmits a signal corresponding to the motion state of the vehicle body, such as the vehicle speed or acceleration of the vehicle V, to the external recognition device 65 .

[0057] The external recognition device 65 is a computer that performs sensor fusion processing on the detection results of the camera unit 61, the lidar units 62a-62e, the radar units 63a-63e, and the body sensor 64. It identifies external information related to the vehicle V's surroundings (such as the position, distance, and relative speed of obstacles or other vehicles, and the type and position of marking lines), as well as vehicle state information related to the vehicle V's motion state (such as vehicle speed, direction of travel, and acceleration), and further evaluates the reliability of the recognition results. The external recognition device 65 transmits information related to the recognition results and information related to the reliability of the recognition results to the vehicle control device 1.

[0058] The vehicle control device 1 is a computer comprised of hardware components, including a central processing unit (CPU), a secondary storage unit (HDD) or solid-state drive (SSD) for storing various programs, and a primary storage unit (RAM), such as random access memory (RAM), for storing data temporarily required by the processing unit while executing programs. This hardware configuration includes an automatic steering control module that controls the electric power steering system 9, an automatic driving control module that controls the power plant 8, and an automatic braking control module that controls the braking system 7.

[0059] The automatic steering control module of the vehicle control device 1 calculates a target control amount relative to the control amount (for example, steering angle) specified for the electric power steering device 9 based on the input information obtained from the external identification device 65, and operates the electric power steering device 9 so that the control amount of the electric power steering device 9 is the target control amount.

[0060] The automatic driving control module of the vehicle control device 1 calculates the target control amount relative to the control amount specified for the power device 8 (for example, driving driving force) based on the input information obtained from the external identification device 65, and operates the power device 8 so that the control amount of the power device 8 is the target control amount.

[0061] The automatic braking control module of the vehicle control device 1 calculates the target control amount relative to the control amount (for example, braking force) specified for the braking device 7 based on the input information obtained from the external identification device 65, and operates the braking device 7 so that the control amount of the braking device 7 is the target control amount.

[0062] Figure 21 is a diagram showing the structure of the automatic steering control module related to the control of the electric power steering device 9 in the vehicle control device 1. In addition, since the structure of the automatic driving control module and the automatic braking control module are similar to Figure 2 The configuration of the automatic steering control modules shown is substantially the same, and therefore detailed description thereof will be omitted.

[0063] The vehicle control device 1 includes a target control amount calculation unit 11 that calculates a target control amount (hereinafter referred to as "u") relative to a control amount (hereinafter referred to as "u") output from a steering sensor 95 of an electric power steering device 9 based on input information (hereinafter referred to as "x") transmitted from an external recognition device 65. cmd ”); and, the motor drive unit 12, based on the target control amount u calculated by the target control amount calculation unit 11 cmd To operate the electric motor 94 of the electric power steering device 9.

[0064] In addition, the variables at time t are expressed in brackets below. That is, the input information, control amount, and target control amount at time t are expressed as x(t), u(t), and u(t), respectively. cmd (t).

[0065] The motor driving unit 12 operates the motor 94 (i.e., adjusts the drive current to the motor 94) based on a known feedback control algorithm so that the target control amount u calculated by the target control amount calculating unit 11 using the following process becomes cmd It is consistent with the control amount u output from the steering sensor 95.

[0066] The target control amount calculation unit 11 includes an input information acquisition unit 2, a plurality of control amount predictors 30, 31, ..., 3 M-1 , and weighted average calculator 4.

[0067] The input information acquisition unit 2 acquires information required for controlling the control object, that is, time series data of input information x, based on the recognition result of the external recognition device 65. In addition, as shown in the following formula (1), the input information x acquired by the input information acquisition unit 2 is set as an N-dimensional vector, which consists of N (N is an integer greater than or equal to 2) input variables (x1, x2, ..., x N )constitute.

[0068] [Mathematical formula 1]

[0069] x(t)=[x1(t),x2(t),…,x N (t)] (1)

[0070] The input information acquisition unit 2 acquires input information at M different times (M is an integer greater than or equal to 2, hereinafter also referred to as "sample number") between the current time and a specific time before from the external recognition device 65. That is, when the current time is set to "t" and the sampling time is recorded as "Δt", the input information acquisition unit 2 acquires M different times [t-τ0, t-τ1, t-τ2, ..., t-τ M-1 ]’s input information [x(t-τ0),x(t-τ1),x(t-τ2),…,x(t-τ M-1 )]. In the following, when any integer between 0 and M-1 is set to "i", the difference between the current time t and the time t-Δt×i i points before the current time t is recorded as "τ i ”, also referred to as i hours.

[0071] like Figure 2 As shown, the input information acquisition unit 2 is connected to the control amount predictors 30, 31, 32, ..., 3 in the input information acquisition unit 2, which are the same number as the number of samples M. M-1 Each control quantity predictor 30,31,32,…,3 M-1 Based on the different time points [t-τ0, t-τ1, t-τ2, ..., t-τ M-1 ]’s input information [x(t-τ0),x(t-τ1),x(t-τ2),…,x(t-τ M-1 )], calculate the predicted control quantity [u τ0 (t),u τ1 (t),u τ2 (t),…,u τM-1 (t)].

[0072] In addition, the following is based on the input information x(t-τ i ) Calculate the predicted control quantity u τi (t) Control variable predictor 3 i It is called the i-th control quantity predictor, which will be used by the i-th control quantity predictor 3 i The calculated predicted control quantity u τi (t) is called the i-th predicted control variable. In addition, the i-th control variable predictor 3 is also referred to below. i The input x(t-τ i ) and output u τi (t) time difference τ i It is called prediction time.

[0073] That is, the 0th control amount predictor 30 calculates the predicted value of the control amount at the current time t as the 0th predicted control amount u based on the input information x(t) at the current time t. τ0 The first control variable predictor 31 calculates the predicted value of the control variable at the current time t as the first predicted control variable u based on the input information x(t-τ1) before the prediction time τ1 from the current time t. τ1 The second control variable predictor 32 calculates the predicted value of the control variable at the current time t as the second predicted control variable u based on the input information x(t-τ2) before the prediction time τ2 from the current time t. τ2 (t). In addition, the M-1th control quantity predictor 3 M-1 Predicted time τ from current time t M-1 Input information x(t-τ M-1 ), calculate the predicted value of the control amount at the current time t as the M-1th predicted control amount u τM-1 (t).

[0074] In addition, these M control variable predictors 30, 31, 32, ..., 3 M-1 Each includes a control quantity prediction model that associates the input information before the prediction time from the current moment with the control quantity at the current moment, and calculates the predicted control quantity by using the control quantity prediction model. i Including predicting the time τ from the current time t i The input information before x(t-τ i ) establishes the i-th control quantity prediction model associated with the control quantity u(t) at the current time t, and uses the i-th control quantity prediction model to calculate the control quantity according to the input information x(t-τ i ) Calculate the i-th predicted control quantity u τi (t).

[0075] Hereinafter, the input-output characteristics of the i-th control amount prediction model are represented by function f, as shown in the following formula (2). The function f is based on the input information x(t-τ i ) is the explanatory variable, and the control quantity u is predicted τi (t) is the target variable. In the following formula (2), “A τi " is a parameter that characterizes the input-output characteristics of the i-th control quantity prediction model. That is, as described below, when the control quantity prediction model is constructed using a neural network, "A τi " is equivalent to multiple weighted coefficients that characterize the input and output characteristics of the neural network.

[0076] [Mathematical formula 2]

[0077] u τi (t) = f(x(t-τ)i ), A τi ) (2)

[0078] As shown in the following formula (3), the weighted average calculator 4 calculates the M control amount predictors 30, 31, 32, ..., 3 M-1 The calculated M prediction control quantities [u τ0 (t),u τ1 (t),u τ2 (t),…,u τM-1 The weighted average value of (t)] is taken as the target control quantity u cmd (t) and input to the motor driving unit 12.

[0079] [Mathematical formula 3]

[0080]

[0081] Here, in the above formula (3), “w k " is related to the kth (k is an integer between 0 and M-1) predicted control quantity u τk The weight corresponding to (t) is also referred to as the kth weight. In addition, there is a prediction time τ k The shorter the predicted control quantity u τk Therefore, the weighted average calculator 4 preferably sets the i-th weight w i Set to be greater than the jth (j is an integer greater than i) weight w j In other words, the weighted average calculator 4 preferably takes the predicted control amount u τk (t) The corresponding weight w k Set as the prediction time τ k More specifically, for example, as shown in the following equations (4-1) to (4-3), the weighted average calculator 4 preferably sets the kth weight w k The value of is set to decrease exponentially with respect to the value k. In the following formula (4-1), "β" is a positive time constant.

[0082] [Formula 4]

[0083] w′ k =exp(-k / β) (4-1)

[0084]

[0085] w k =w′ k / α (4-3)

[0086] Next, the process of controlling the control amount specified for the control object by the vehicle control device 1 as described above will be described. First, the input information acquisition unit 2 acquires the time from the current time t to the specific time τ M-1 The input information at M different moments before [x(t-τ0),x(t-τ1),x(t-τ2),…,x(t-τ M-1 )]. Then, M control variable predictors 30, 31, 32, ..., 3 M-1 Based on the input information acquisition unit 2, M pieces of input information at different times [x(t-τ0), x(t-τ1), x(t-τ2), ..., x(t-τ M-1 )], calculate M predicted control quantities [uτ0(t),uτ1(t),uτ2(t),…,uτ M-1 (t)]. More specifically, the i-th control variable predictor 3 i By using the predicted time τ from the current time t i The input information before x(t-τ i ) establishes the i-th control quantity prediction model associated with the control quantity u(t) at the current time t, and calculates the i-th predicted control quantity u τi (t).

[0087] Next, the weighted average calculator 4 calculates the weights of the control amount predictors 30, 31, 32, ..., 3 M-1 The calculated M prediction control quantities [u τ0 (t),u τ1 (t),u τ2 (t),…,u τM-1 The weighted average value of (t)] is taken as the target control quantity u cmd (t) In addition, the motor drive unit 12 operates the control object based on a known feedback control algorithm so that the target control amount u calculated according to the above process is cmd (t) is consistent with the control variable u(t) output from the steering sensor 95. The vehicle control device 1 controls the control object by repeatedly executing the above steps at each specific time (for example, sampling time Δt).

[0088] Next, the target control amount u calculated by the target control amount calculation unit 11 is compared with the comparative examples 1 and 2. cmd Here, the target control amount calculation unit of Example 1 calculates the target control amount u based only on the input variable x(t). cmd (t). That is, Figure 2 In the target control amount calculation unit 11 shown in FIG, the target control amount u calculated by the target control amount calculation unit of the comparative example 1 is cmd(t) is the same as the 0th predicted controlled variable u calculated by the 0th controlled variable predictor 30. τ0 (t) is equal. In addition, the target control amount calculation unit of the comparative example 2 is the amount obtained by applying an FIR filter with a sample number M to the output of the target control amount calculation unit of the comparative example 1 as the target control amount u cmd (t). That is, after the time τ from the current time t i The output of the comparative example 1 is set to y (i is an integer between 0 and M-1). τi (t), the target control amount u of Example 2 cmd (t) is represented by the following formula (5).

[0089] [Formula 5]

[0090]

Number 5

[0091]

[0092] Figure 3 The u calculated by the target control amount calculation unit in Comparative Example 1 (upper section), Comparative Example 2 (middle section) and this embodiment (lower section) is cmd In addition, Figure 3 For reference, the following ideal control amount is represented by a dotted line.

[0093] First, in order to calculate the target control quantity u cmd The time series data of the input information x contains noise or defects. Therefore, in the comparative example 1 that only takes the input information of a single moment as input, the target control amount u is calculated. cmd In the case of , the influence of noise or defects contained in the time series data of the input information x appears directly, such as Figure 3 As shown in the previous paragraph, it sometimes vibrates.

[0094] In addition, if Figure 3 As shown in the middle section, Figure 3 The vibration shown in the upper part of can be eliminated by applying FIR filter. However, if only FIR filter is applied, the past information will be dragged along, such as Figure 3 As shown in the middle section, the delay of the ideal control amount becomes significant.

[0095] In this regard, according to this embodiment, by combining the above-mentioned input information acquisition unit 2, M control amount predictors 30, 31, 32, ..., 3 M-1 , and weighted average calculator 4 to calculate the target control amount u cmd (t), which can eliminate the influence of noise or defects contained in the time series data of the input information x and reduce the delay of the ideal control amount.

[0096] Next, M control quantity predictors 30, 31, 32, ..., 3 are constructed using a neural network. M-1 The process of using the M control quantity prediction model is described below.

[0097] First, if Figure 4 As shown in , the operator prepares the training data required to construct M control quantity prediction models using machine learning. Figure 4 As shown, the training data consists of time series data [x(t0), x(t1), x(t2), ..., x(t L-3 ),x(t L-2 ),x(t L-1 )] (hereinafter, the time series data of the input information of the sample number L is also referred to as "input sample data"), and the ideal control amount u of the sample number L corresponding to the input sample data of the sample number L ideal Time series data[u ideal (t0),u ideal (t1),u ideal (t2),…,u ideal (t L-3 ),uideal(t L-2 ),u ideal (t L-1 )] (hereinafter, the time series data of the ideal control amount of these sample numbers L are also referred to as "ideal output data"). Here, the ideal control amount u ideal This corresponds to the control amount to be achieved in the target control amount calculation unit 11 .

[0098] For example, the input sample data and ideal output data of the sample number L can utilize actual driving data obtained when a model driver who can manually achieve ideal driving maneuvers manually drives the vehicle V during a specific period. More specifically, the input sample data of the sample number L can utilize time-series data of input information x acquired by the input information acquisition unit 2 when the model driver manually drives the vehicle V during the specific period. Furthermore, the ideal output data of this input sample data can utilize time-series data of the control variable u manually achieved by the model driver during the same period as the acquisition of the input sample data.

[0099] Next, the operator uses machine learning using the input sample data and ideal output data prepared by the above process as training data to sequentially construct M control variable prediction models. Figure 5 As shown, for example, the i-th control amount prediction model is constructed using Li group training data, wherein the Li group training data is the input sample data of sample number L and the forward prediction time τ relative to the input sample data.i (=t i -t0) as the ideal output data set.

[0100] That is, Figure 5 As shown in the upper part, the prediction model of the 0th control quantity with prediction time 0 is constructed using L sets of training data, which are the input sample data [x(t0), x(t1),…, x(t L-2 ),x(t L-1 )] and the ideal output data at the same time [u ideal (t0),u ideal (t1),…,u ideal (t L-2 ),u ideal (t L-1 )] as a data set. Figure 5 As shown in the middle section, the first control quantity prediction model with prediction time τ1 (= t1-t0) is constructed using L-1 sets of training data, which are input sample data [x(t0), x(t1),…, x(t L-3 ),x(t L-2 )] and the ideal output data [u ideal (t1),u ideal (t2),…,u ideal (t L-2 ),u ideal (t L-1 )] as a data set. In addition, Figure 5 As shown in the lower part of i (=t i -t0) is constructed using Li group training data, wherein the Li group training data is the input sample data [x(t0), x(t1),…, x(t L-2-i ),x(t L-1-i )] and advance the time prediction time τ relative to the input sample data i The ideal output data [u ideal (t i ),u ideal (t i+1 ),…,u ideal (t L-2 ),u ideal (t L-1 )] as the dataset.

[0101] In addition, as shown in the following formulas (6-1) and (6-2), the weight coefficient A of the neural network that represents the input-output characteristics of the i-th control variable prediction model isτi Determine according to a known algorithm that the error function E(A) defined using the above Li group training data τi )minimum.

[0102] [Formula 6]

[0103]

[0104] According to the vehicle control device 1 of this embodiment, the following effects are achieved.

[0105] (1) The target controlled variable calculation unit 11 calculates the target controlled variable u relative to the controlled variable u specified for the controlled object based on the time series data of the input information x. cmd The motor drive unit 12 calculates the target control amount u cmd , operating the control object. In addition, the target control amount calculation unit 11 includes: an input information acquisition unit 2, which obtains the current time t to a specific time τ M The input information at M different moments before [x(t-τ0),x(t-τ1),x(t-τ2),…,x(t-τ M-1 )]; M control quantity predictors 30,31,32,…,3 M-1 , based on the input information of each moment in M time, calculate the predicted control amount [u τ0 (t),u τ1 (t),u τ2 (t),…,u τM-1 (t)]; and, a weighted average calculator 4, which calculates the M control amount predictors 30, 31, 32, ..., 3 M-1 The calculated M prediction control quantities [u τ0 (t),u τ1 (t),u τ2 (t),…,u τM-1 The weighted average value of (t)] is taken as the target control quantity u cmd (t) According to the vehicle control device 1, by calculating M predicted control quantities [u τ0 (t),u τ1 (t),u τ2 (t),…,u τM-1 The weighted average value of (t)] is taken as the target control quantity u cmd (t), can suppress the influence of noise or defects contained in the time series data of the input information, and calculate the target control amount u with less vibration cmd (t). In addition, in the present invention, there are a total of M control amount predictors 30, 31, 32, ..., 3 M-1 In the example, the i-th control quantity predictor 3 iBy using the control variable prediction model to calculate the predicted control variable u τi (t), the control quantity prediction model is to predict the time τ from the current time t. i The input information x(t-τ i ) is associated with the control amount u(t) at the current time t. That is, each control amount predictor 30, 31, 32, ..., 3 M-1 Based on the input information at different times, the control quantity at the same current moment is calculated as the predicted control quantity [u τ0 (t),u τ1 (t),u τ2 (t),…,u τM-1 (t)]. Thus, according to the vehicle control device 1, it is possible to suppress the influence of noise or defects included in the time series data of input information while performing control with little delay, and further contribute to the development of a sustainable transportation system.

[0106] (2) In the vehicle control device 1, each control amount predictor 30, 31, 32, ..., 3 M-1 A predicted controlled variable is calculated based on input information using a controlled variable prediction model. This controlled variable prediction model is constructed using machine learning, using input sample data (time series data of the input information) and ideal output data (time series data of the ideal controlled variable for the input sample data) as training data. Consequently, the vehicle control device 1 can suppress the effects of noise and defects contained in the time series data of the input information while achieving ideal control with minimal delay.

[0107] (3) In the vehicle control device 1, the i-th control variable predictor 3 i Using the i-th control quantity prediction model, the prediction time τ is calculated based on the current time t. i The input information before x(t-τ i ), calculate the predicted control quantity u at the current time t τi (t), the i-th control quantity prediction model is constructed using training data, and the training data is the input sample data and the prediction time τ relative to the input sample data. i Thus, according to the vehicle control device 1, it is possible to suppress the influence of noise or defects included in the time series data of the input information and realize ideal control with less delay.

[0108] (4) In the vehicle control device 1, the weighted average calculator 4 and the i-th control amount predictor 3 i The calculated predicted control quantity u τi (t) corresponds to the i-th weight w i, set to be compared with the jth (j is an integer greater than i) control quantity predictor 3 j The calculated predicted control quantity u τj (t) corresponds to the jth weight w j In other words, the weighted average calculator 4 will be based on the predicted time τ from the current time t i The input information before x(t-τ i ) The i-th predicted control quantity u calculated τi The weight corresponding to (t) is set to the prediction time τ earlier than j The input information before x(t-τ j ) The j-th predicted control quantity u calculated τj (t) The corresponding weight w j Generally speaking, the closer the input information is to the current time, the higher the prediction accuracy of the predicted control amount tends to be. By setting the weight to a larger value according to the improvement of the prediction accuracy, a high-precision target control amount u can be calculated. cmd .

[0109] (5) In the vehicle control device 1, the weighted average calculator 4 calculates the kth (k is an integer between 0 and M-1) weight w k The value of is set to decrease exponentially with respect to the value k. Thus, the target control amount u can be calculated with high precision by simple calculation. cmd .

[0110] (6) In the vehicle control device 1, the target control amount calculation unit 11 calculates the target control amount u of the steering angle generated by the electric power steering device 9 according to the above process based on the time series data of the input information including the external information around the vehicle. cmd The motor drive unit 12 calculates the target control amount u cmd , and operates the electric power steering device 9. Thus, according to the vehicle control device 1, it is possible to suppress the influence of noise or defects included in the external information and realize steering control with less delay.

[0111] (7) In the vehicle control device 1, the target control amount calculation unit included in the automatic driving control module calculates the target control amount of the driving force generated by the power unit 8 according to the above process based on the time series data of the input information including the external information around the vehicle. The automatic operation unit operates the power unit 8 based on the calculated target control amount. Thus, according to the vehicle control device 1, it is possible to suppress the influence of noise or defects contained in the external information while achieving driving force control with less delay.

[0112] (8) In the vehicle control device 1, the target control amount calculation unit included in the automatic brake control module calculates the target control amount of the braking force generated by the brake device 7 according to the above process based on the time series data of the input information including the external information around the vehicle, and the automatic operation unit operates the brake device 7 based on the calculated target control amount. Thus, according to the vehicle control device 1, it is possible to suppress the influence of noise or defects contained in the external information while achieving brake control with less delay.

[0113] <Second embodiment>

[0114] Next, a vehicle control device according to a second embodiment of the present invention will be described with reference to the drawings. In the following description, the same components as those of the vehicle control device 1 according to the first embodiment are denoted by the same reference numerals, and detailed description thereof will be omitted.

[0115] Figure 6 1A is a diagram showing the structure of the automatic steering control module related to the control of the electric power steering device 9 in the vehicle control device 1A of this embodiment. Figure 6 The configurations of the automatic steering control modules shown are substantially the same, and therefore detailed descriptions thereof will be omitted.

[0116] like Figure 6 As shown, the target control amount calculation unit 11A includes an input information acquisition unit 2, a credibility acquisition unit 2A, M control amount predictors 30, 31, ..., 3 M-1 , and weighted average calculator 4A.

[0117] The credibility acquisition unit 2A acquires the time series data of the credibility c of the time series data of the input information x acquired by the input information acquisition unit 2 from the external recognition device 65. More specifically, the credibility acquisition unit 2A acquires the M input information [x(t-τ0), x(t-τ1), x(t-τ2), ..., x(t-τ M-1 )] their respective credibility [c(t-τ0),c(t-τ1),c(t-τ2),…,c(t-τ M-1 )]. That is, the credibility c(t-τ0) is the credibility of the input information x(t-τ0) at the same time, the credibility c(t-τ1) is the credibility of the input information x(t-τ1) at the same time, and the credibility c(t-τ M-1 ) is the input information x(t-τ M-1 )’s credibility.

[0118] As shown in the following equations (7-1) and (7-2), the weighted average calculator 4A calculates the M control amount predictors 30, 31, 32, ..., 3M-1 The calculated M prediction control quantities [u τ0 (t),u τ1 (t),u τ2 (t),…,u τM-1 (t)] and the credibility [c(t-τ0), c(t-τ1), c(t-τ2), ..., c(t-τ M-1 )] is taken as the target control quantity u cmd (t) and input to the motor driving unit 12.

[0119] [Formula 7]

[0120]

[0121] In the above formulas (7-1) and (7-2), “w k (c(t-τ k ))” is related to the kth predicted control quantity u τk (t) corresponds to the kth weight, and is the kth input information x(t-τ k ) corresponds to the kth credibility c(t-τ k More specifically, the weighted average calculator 4A preferably sets the weight w k (c(t-τ k )) is set as the prediction time τ k In addition, the weighted average calculator 4A preferably sets the weight w k (c(t-τ k )) is set as the confidence c(t-τ k ) decreases, in other words, as the k-th input information x(t-τ k ) decreases as the credibility of the .

[0122] According to the vehicle control device 1A of the present embodiment, in addition to the effects described in (1) to (8) above, the following effects are achieved.

[0123] (9) In the vehicle control device 1A, the reliability acquisition unit 2A acquires the predicted time τ from the current time t. i The input information before x(t-τ i )’s credibility c(t-τ i ), the weighted average calculator 4A calculates the value of the obtained i-th credibility c(t-τ i ), set the i-th weight w i (c(t-τ i )) value. Thus, the credibility c of the input information x at each moment can be reflected to calculate the target control amount u with high precision cmd .

[0124] <Third embodiment>

[0125] Next, a vehicle control device according to a third embodiment of the present invention will be described with reference to the accompanying drawings. In the following description, the same components as those of the vehicle control device 1 according to the first embodiment are denoted by the same reference numerals, and detailed descriptions thereof will be omitted.

[0126] Figure 7 1B is a diagram showing the structure of the automatic steering control module related to the control of the electric power steering device 9 in the vehicle control device 1B of this embodiment. Figure 7 The configuration of the automatic steering control modules shown is substantially the same, and therefore detailed description thereof will be omitted.

[0127] like Figure 7 As shown, the target control amount calculation unit 11B includes an input information acquisition unit 2, a learning device 2B, M control amount predictors 30, 31, ..., 3 M-1 , and weighted average calculator 4.

[0128] The learning device 2B learns the control amount predictors 30, 31, ..., 3 based on the time series data of the control amount u and input information x during manual driving with the driver of the vehicle V as the operator of the electric power steering device 9 as the control object. M-1 The control quantity prediction model.

[0129] More specifically, the learning device 2B calculates the error function e(A) defined by the following equation (8-1) based on the time series data of the control amount u and input information x acquired during the manual operation of the electric power steering device 9 by the driver. τi ). In addition, as shown in the following formula (8-2), the learning device 2B updates the update amount defined by the gradient descent method (the second term on the right side of formula (8-2)) and the parameter A that represents the input-output characteristics of the i-th control amount prediction model. τi Add them in sequence to update the i-th control quantity prediction model.

[0130] [Formula 8]

[0131] e(A τi )=(u ideal (t)-f(x(t-τ i ),A τi )) 2 (8-1)

[0132]

[0133] Here, in the above formula (8-1), “u ideal" is the ideal control amount, and the time series data of the control amount u manually implemented by the driver can be used. In addition, in the above formula (8-2), "η" represents the learning rate and is set to a predetermined value.

[0134] According to the vehicle control device 1B of the present embodiment, in addition to the effects described in (1) to (8) above, the following effects are achieved.

[0135] (10) In the vehicle control device 1B, the learning device 2B learns M control amount prediction models based on the time series of the control amount u and input information x during manual driving by the operator whose vehicle V is the control target. Thus, according to the vehicle control device 1B, the input-output characteristics of the M control amount prediction models can be changed according to the characteristics of the driver of the vehicle V or the characteristics of the sensor unit 6 for acquiring external information over time.

[0136] While one embodiment of the present invention has been described above, the present invention is not limited thereto and the detailed configuration can be modified as appropriate within the scope of the present invention.

[0137] Reference numerals

[0138] V Vehicle

[0139] 1,1A,1B Vehicle control unit (control unit)

[0140] 11,11A,11B Target control amount calculation unit (target control amount calculation unit)

[0141] 2 Input information acquisition unit (input information acquisition unit)

[0142] 2A Credibility Acquisition Unit (Credibility Acquisition Unit)

[0143] 2B Learning Device (Learning Unit)

[0144] 30,31,…,3 M-1 Controlled variable predictor (controlled variable prediction unit)

[0145] 4,4A Weighted average calculator (additive average calculation unit)

[0146] 12 Motor drive unit (automatic operation unit)

[0147] 6 sensor units

[0148] 7 Braking device (controlled object)

[0149] 8 Power equipment (travel drive device, control object)

[0150] 9 Electric power steering device (steering mechanism, control object)

Claims

1. A control device that calculates a target control variable relative to a control variable specified for a control object based on input information, and operates the control object based on the target control variable, the control device comprising: a target control amount calculation unit, which calculates the target control amount based on the time series data of the input information; and, An automatic operation unit operates the control object based on the target control amount; and, The aforementioned target control amount calculation unit includes: an input information acquisition unit, which acquires the input information at M different times (M is an integer greater than or equal to 2) between the current time and a specific time; M control amount prediction units calculate the predicted control amount based on the input information acquired by the input information acquisition unit; and a weighted average calculation unit that calculates a weighted average of the M predicted control quantities calculated by the M control quantity prediction units as the target control quantity; and The i-th (i is an integer between 0 and M-1) control quantity prediction unit calculates the predicted control quantity by using a control quantity prediction model, and the control quantity prediction model establishes an association between the input information i hours before the current moment and the control quantity at the current moment.

2. The control device according to claim 1, wherein: The aforementioned control quantity prediction model is constructed using machine learning, which uses input sample data and ideal output data as training data. The input sample data is the time series data of the aforementioned input information, and the ideal output data is the time series data of the ideal control quantity for the input sample data.

3. The control device according to claim 2, wherein: The i-th control quantity prediction model is constructed using the aforementioned training data, and the aforementioned training data is a data set consisting of the aforementioned input sample data and the aforementioned ideal output data advanced i hours relative to the input sample data.

4. The control device according to any one of claims 1 to 3, wherein: The weighted average calculation unit sets the i-th weight corresponding to the predicted control quantity calculated by the i-th control quantity prediction unit to a value greater than the j-th weight corresponding to the predicted control quantity calculated by the j-th (j is an integer greater than i) control quantity prediction unit.

5. The control device according to claim 4, wherein: The weighted average calculation unit sets the value of the k-th (k is an integer between 0 and M-1) weight to decrease exponentially with respect to the value k.

6. The control device according to claim 5, wherein: The target control amount calculation unit further includes a credibility acquisition unit, which acquires the credibility of the input information i hours before the current time. The weighted average calculation unit sets a value of the i-th weight based on the i-th reliability acquired by the reliability acquisition unit.

7. The control device according to claim 1, wherein: The aforementioned control object is the steering mechanism of the vehicle, The aforementioned control amount is the steering angle generated by the aforementioned steering mechanism, The input information includes external information related to the surroundings of the vehicle.

8. The control device according to claim 1, wherein: The aforementioned control object is the vehicle's driving device. The control amount is the driving force generated by the driving device. The input information includes external information related to the surroundings of the vehicle.

9. The control device according to claim 1, wherein: The aforementioned control object is the vehicle's braking device, The aforementioned control amount is the braking force generated by the aforementioned braking device, The input information includes external information related to the surroundings of the vehicle.

10. The control device according to any one of claims 7 to 9, wherein: The control device further includes a learning unit configured to learn the control amount prediction model based on time series data of the control amount and the input information during manual driving by an operator having a driver of the vehicle as the control target.

11. A control method for controlling a control variable specified for a control object using a computer, the control method comprising the following steps: Get input information at M different times (M is an integer greater than or equal to 2) between the current time and a specific time; Based on the aforementioned input information at M different moments, M predicted control variables are calculated; Calculating a weighted average of the M predicted control quantities as a target control quantity relative to the control quantity; and Based on the aforementioned target control amount, operating the aforementioned control object; and, In the step of calculating the aforementioned predicted control quantity, the i-th aforementioned predicted control quantity is calculated by using a control quantity prediction model, and the control quantity prediction model establishes an association between the aforementioned input information i hours before the aforementioned current moment (i is an integer between 0 and M-1) and the aforementioned control quantity at the aforementioned current moment.

12. A storage medium storing a computer program for causing a computer to control a predetermined control variable of a control object, wherein: The computer program causes the computer to execute the following steps: Get input information at M different times (M is an integer greater than or equal to 2) between the current time and a specific time; Based on the aforementioned input information at M different moments, M predicted control variables are calculated; Calculating a weighted average of the M predicted control quantities as a target control quantity relative to the control quantity; and Based on the aforementioned target control amount, operating the aforementioned control object; and, In the step of calculating the aforementioned predicted control quantity, the i-th aforementioned predicted control quantity is calculated by using a control quantity prediction model, and the aforementioned control quantity prediction model establishes an association between the aforementioned input information i hours before the aforementioned current moment (i is an integer between 0 and M-1) and the aforementioned control quantity at the aforementioned current moment.

Images