Program, device, learning model, and method

Abstract

A direction indicator of a motorcycle is controlled with good precision. A program causes a computer to execute: a process for acquiring, from at least one sensor, travel data of a two-wheeled motor vehicle, said travel data being continuously measured over a predetermined period of time; and determining a corresponding driving operation, which is a driving operation corresponding to the travel data, on the basis of the travel data and a plurality of reference travel data that are continuous for a predetermined period of time and that are associated with each of a plurality of driving operations of the two-wheeled motor vehicle.

Description

Technical Field

[0001] The present invention relates to a program, a device, a learning model and a method. Background Art

[0002] The driver of a two-wheeled motor vehicle uses a direction indicator (turn indicator) when turning right or left (referred to as a right turn, etc.) or changing lanes on the road while driving. The direction indicator is used, for example, to show the driver's driving intention to surrounding drivers or pedestrians, thereby achieving safe driving. Regarding the direction indicator of the two-wheeled motor vehicle, it is necessary that the driver first turns the direction indicator to an on state (for example, flashing), and then turns the direction indicator to an off state (for example, extinguishing) by the driver himself after the right turn, etc. At this time, sometimes the driver forgets to operate, so that the direction indicator is not set to off. Patent document 1 shows a direction indicator canceling device for avoiding forgetting to turn off the direction indicator.

[0003] Prior art literature

[0004] Patent Literature

[0005] Patent Document 1: Japanese Patent Application Laid-Open No. 62-53256 Summary of the invention

[0006] Problem that the invention aims to solve

[0007] The turn signal canceling device described in Patent Document 1 performs the following control: the turn signal is turned off based on the speed difference between the rotation speed of the front wheel and the rotation speed of the rear wheel generated when turning right and left. More specifically, the following control is performed: each time the speed difference at a certain point in time is obtained, the turn signal is turned off based on its change. In the case of using the value (instantaneous value) at a certain point in time like this, it is determined whether the value exceeds a threshold. However, in the case of a two-wheeled vehicle, there is the following problem: since the vehicle body tilts according to the speed when turning right or left, it is necessary to change the threshold according to the speed, making it difficult to make a determination. In addition, with respect to a two-wheeled vehicle, since the degree of freedom of the position that the vehicle body can occupy is large relative to the width of the road, there is a problem of determining unintentional activities as right turns or left turns.

[0008] The present invention has been made in view of such circumstances, and an object of the present invention is to provide a program or the like for controlling a direction indicator of a motorcycle with high accuracy.

[0009] Solutions for solving problems

[0010] A program involved in one aspect of the present invention causes a computer to perform the following processing: obtaining driving data of a motor vehicle that is continuously measured within a specified period from at least one sensor; and determining a driving action corresponding to the driving data, i.e., a corresponding driving action, based on the driving data and a plurality of reference driving data that are continuous within the specified period and respectively associated with a plurality of driving actions of the motor vehicle.

[0011] The device involved in another aspect of the present invention comprises: an acquisition unit, which acquires driving data of a motor vehicle measured continuously within a specified period from at least one sensor; and a determination unit, which determines a corresponding driving action corresponding to the driving data based on the driving data and a plurality of reference driving data that are continuous within the specified period and respectively associated with a plurality of driving actions of the motor vehicle.

[0012] Another aspect of the present invention involves a learning model that is learned using training data that takes as input a plurality of reference driving data of a motor two-wheeled vehicle continuously within a specified period and takes as output driving actions respectively associated with the plurality of reference driving data, and the learning model enables a computer to function in the following manner: for input of driving data of a motor two-wheeled vehicle continuously measured within a specified period, the learning model outputs a driving action corresponding to the driving data, i.e., a corresponding driving action.

[0013] A method according to another aspect of the present invention includes a computer performing the following processing: obtaining driving data of a motor vehicle measured continuously within a prescribed period from at least one sensor; and determining a corresponding driving action corresponding to the driving data based on the driving data and a plurality of reference driving data that are continuous within the prescribed period and respectively associated with a plurality of driving actions of the motor vehicle.

[0014] Effects of the Invention

[0015] According to the present invention, it is possible to provide a program or the like for accurately controlling a direction indicator of a motorcycle. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is a block diagram of the motorcycle according to the first embodiment.

[0017] Figure 2 It is a diagram for explaining the outline of control in the motorcycle according to the first embodiment.

[0018] Figure 3 It is a diagram for explaining the outline of control in the motorcycle according to the first embodiment.

[0019] Figure 4 This is a flowchart for explaining the determination of the driving action according to the first embodiment.

[0020] Figure 5 It is a diagram for explaining the learning model involved in the first embodiment.

[0021] Figure 6 It is a diagram for explaining the determination of the driving action according to the first embodiment.

[0022] Figure 7 This is a flowchart for explaining the process of determining the end of the driving action according to the first embodiment.

[0023] Figure 8 It is a diagram for explaining the determination of the driving action and the determination of the end of the driving action according to the first embodiment.

[0024] Fig. 9 It is a block diagram of a motorcycle according to the second embodiment.

[0025] Fig.10 It is a diagram for explaining the outline of control in the motorcycle according to the second embodiment.

[0026] Fig.11 This is a diagram for explaining an example of classification probability in the motorcycle according to the second embodiment.

[0027] Fig.12 This is a flowchart for explaining determination of a driving action and determination of the end of the driving action according to the second embodiment.

[0028] Fig.13 This is a block diagram of a motorcycle control system according to a third embodiment.

[0029] Fig.14 This is another block diagram of the motorcycle control system according to the third embodiment.

[0030] Fig.15 This is a graph showing an example of sensor values ​​in the motorcycle control system according to the first to third embodiments.

[0031] Fig.16 This is a graph showing an example of classification probability in the motorcycle control system according to the first to third embodiments.

[0032] Fig.17 This is a graph showing another example of sensor values ​​in the motorcycle control system according to the first to third embodiments.

[0033] Fig.18 This is a graph showing an example of a case where a sensor value is converted into a binary value in the motorcycle control system according to the first to third embodiments. DETAILED DESCRIPTION

[0034] First, a reference example for comparison with the embodiment of the present invention will be described.

[0035] Regarding the determination of right turns and left turns, generally speaking, in the case of a car (four-wheeled vehicle), right turns and left turns are performed by turning the steering wheel, so the angle of the steering wheel can be detected, and the right turn and left turn action can be determined based on the threshold value set for the angle of the steering wheel. On the other hand, in the case of a motorized two-wheeled vehicle (two-wheeled vehicle), especially when traveling at high speed, the body of the vehicle is tilted to rotate in order to avoid the body of the vehicle from falling due to centrifugal force. At this time, the change in the handlebar angle becomes smaller accordingly with the inclination of the body. Therefore, it is difficult to determine right turns and left turns using only the handlebar angle. Therefore, it is necessary to determine right turns and left turns based on multiple parameters including speed, inclination of the body, etc. in addition to the handlebar angle. On the other hand, when the parameters used in the determination increase, there is a problem that it is difficult to set the parameters and the threshold values ​​used for determination.

[0036] In this case, as a method of determining whether a right turn or a left turn has been made by considering the correlation between multiple parameters, it is possible to consider using an ensemble tree in the field of machine learning. By using an ensemble tree, the probability of a right turn or a left turn can be set based on the instantaneous values ​​of multiple parameters. On the other hand, the instantaneous values ​​are used after all, and the noise of the instantaneous values ​​will affect the determination result.

[0037] Furthermore, in the case of a two-wheeled vehicle, the width of the vehicle body is small relative to the width of the road, so there is a degree of freedom in the position of the vehicle body on the road. Therefore, it is possible that a right turn or a left turn is determined to have been made due to irregular activities such as moving left and right on the road. In order to prevent this phenomenon, the following method can be considered, which is different from the above-mentioned method of using instantaneous values: judging right turns and left turns based on continuous data using multiple sensors. For example, the self-position estimation technology can be used to estimate the movement trajectory of the vehicle body based on the accumulated value of the sensor value. In this way, the accumulated movement distance and angle that cannot be known if the instantaneous value is used can be estimated, so it is not easily affected by the noise of the instantaneous value. However, in this case, the steady-state error of the sensor value will also be accumulated, so as a result, it is easy to cause problems in the determination accuracy.

[0038] In view of the above, the present invention proposes the following embodiments to solve the problems.

[0039] A first embodiment will be described. Figure 1 Detailed Description of the Invention A block diagram of a motorcycle 101 according to the first embodiment is shown in FIG.

[0040] The motorcycle 101 includes a body portion 201 , an operation portion 202 , a direction indicator 203 , a drive portion 204 , a sensor portion 205 , and a control portion 206 .

[0041] The vehicle body 201 is a vehicle body of the motorcycle 101 , and the motorcycle 101 is configured by mounting an operation unit 202 , a direction indicator 203 , a drive unit 204 , a sensor unit 205 , and a control unit 206 on the vehicle body 201 .

[0042] The operation unit 202 is a device for the driver to operate the motorcycle 101. The operation unit 202 includes, for example, a handlebar that the driver holds when driving, an accelerator for acceleration, a brake for deceleration, and an operation unit for switching the direction indicator 203 on and off.

[0043] The direction indicator 203 is a device for indicating the driving action to the drivers of other vehicles or pedestrians when the motorcycle 101 performs an action (driving action) such as turning right, turning left or changing lanes. For example, the direction indicator 203 is a light installed on the motorcycle 101 that flashes.

[0044] The driving unit 204 is an engine, a motor, wheels, etc. The driving unit 204 drives the motorcycle 101 based on the operation of the driver through the operation unit 202 and the control from the control unit 206 .

[0045] The sensor unit 205 is a variety of devices for detecting the driving data of the motorcycle 101. Driving data include, for example, the speed, acceleration, tilt, wheel rotation speed, throttle opening, brake input, gear position, etc. of the motorcycle 101. The sensor unit 205 has a speed sensor, an acceleration sensor, a gravity sensor (tilt sensor), a tachometer (rotation speed sensor), etc., to detect various driving data. The sensor data is usually one-dimensional data, but sometimes a camera or a radar is used to obtain two-dimensional data. The sensor unit 205 does not need to obtain all of the above-mentioned driving data, but only needs to obtain at least one driving data. In addition, the types of driving data are not limited to the above-mentioned examples.

[0046] The control unit 206 performs signal processing and information processing in the motorcycle 101. The control unit 206 includes a communication unit 2061, a storage unit 2062, a travel data acquisition unit 2063, a conversion unit 2064, a determination unit 2065, and an action control unit 2066. The control unit 206 is a computer that performs predetermined processing by executing a program stored in the storage unit 2062 through a processor.

[0047] The communication unit 2061 controls information communication of the motorcycle 101. For example, the communication unit 2061 transmits and receives data with the sensor unit 205 by performing wired communication or wireless communication with the sensor unit 205. Alternatively, the communication unit 2061 may control communication between the motorcycle 101 and an external update terminal. The update terminal transmits information for updating a program for causing the control unit 206 to operate to the control unit 206, so as to update the program of the control unit 206.

[0048] The storage unit 2062 is, for example, a semiconductor memory. The storage unit 2062 stores a learning model 20621. The learning model 20621 takes input data based on driving data as input, and outputs a driving action (corresponding driving action) corresponding to the input data. In addition, the learning model 20621 also outputs a classification probability, which is the probability that the input data is classified as the corresponding driving action. The details of the learning model 20621 are described later.

[0049] The driving data acquisition unit 2063 acquires driving data from the sensor unit 205. The driving data acquisition unit 2063 acquires driving data continuously measured within a prescribed period. Here, the prescribed period refers to, for example, a period prescribed based on the moving distance of the two-wheeled motor vehicle 101. Specifically, it is a period required for the two-wheeled motor vehicle 101 to move a certain distance. When the two-wheeled motor vehicle 101 moves a certain distance, the driving data acquisition unit 2063 performs sampling at intervals corresponding to the rotation of the wheels, for example, at intervals of equal distance, and acquires driving data within the range from the start to the end of the movement as one set of driving data.

[0050] The conversion unit 2064 converts the driving data into input data input to the determination unit 2065. For example, the conversion unit 2064 may also set the horizontal axis to be, for example, the moving distance and relatively plot the information of the driving data to set it as two-dimensional data (input two-dimensional data). At this time, the conversion unit 2064 generates the input data input to the determination unit 2065. The conversion unit 2064 generates a chart with the horizontal axis set to the moving distance based on the driving data acquired by the driving data acquisition unit 2063, for example, and the chart is generated by converting based on the rotation speed of the wheel and the moving distance. The input data is generated, for example, as data obtained by plotting multiple variables such as speed, tilt, and handlebar angle in one chart. In addition, when the driving data is one-dimensional, the conversion unit 2064 may also regard the chart as a two-dimensional image, for example, when there are features in the frequency, the driving data may also be converted into two-dimensional data. As a method for converting one-dimensional data into two-dimensional data, for example, methods such as Fourier transform and wavelet transform are known, but the conversion method may also be other methods that can classify the final action.

[0051] The determination unit 2065 determines the driving action corresponding to the driving data based on the driving data. Specifically, the determination unit 2065 inputs the input data transformed by the transformation unit 2064 into the learning model 20621 and obtains the corresponding driving action and classification probability from the learning model 20621.

[0052] The action control unit 2066 controls the action of the motorcycle 101 based on the classification probability of the corresponding driving action. Specifically, the action control unit 2066 determines the end of the driving action of turning right or left based on the change in the classification probability. When the action control unit 2066 determines that the right turn or left turn of the motorcycle 101 has been completed, it switches the direction indicator 203 from the on state to the off state.

[0053] Reference Figure 2 The outline of the processing in the motorcycle 101 will be described.

[0054] exist Figure 2 , the movement of the motorcycle 101 when looking down at the intersection is schematically shown. When the traffic light S indicates that the motorcycle 101 can turn right and left, the motorcycle 101 turns left along the path P1 and goes straight along the path P2. Alternatively, the motorcycle 101 turns right along the path P3 and goes straight along the path P4.

[0055] When turning right or left, the driver of the motorcycle 101 operates the operation unit 202 to turn on the direction indicator 203 .

[0056] Here, the case where the motorcycle 101 turns left is described as an example. In step (a), the travel data acquisition unit 2063 acquires the travel data of the motorcycle 101. In step (b), the conversion unit 2064 generates an input image based on the travel data.

[0057] The processing of steps (a) and (b) is repeatedly performed during the travel of the motorcycle 101. For each certain travel distance, a plurality of input images such as the input images I11, I12, and I13 are generated at different times.

[0058] In step (c), the determination unit 2065 inputs a certain input image (for example, input image I11) to the learning model. In step (d), the learning model 20612 calculates the classification probability of the driving action and determines the driving action of the motorcycle (corresponding driving action) based on the classification probability. The processing up to step (d) is performed for each of the plurality of input images.

[0059] In step (e), the end timing of the left turn is estimated based on the change in the classification probability. In step (f), the movement of the two-wheeled motor vehicle 101 is controlled based on the end timing, and the direction indicator 203 is turned off. The case of the left turn has been described so far, and the same process is performed in the case of the right turn.

[0060] In the two-wheeled motor vehicle 101, the corresponding driving action is determined based on the input image generated for each moving distance. Thus, for example, in the two-wheeled motor vehicle 101, the influence of the waiting time caused by the result of waiting for a right turn or waiting for a left turn according to the state of the traffic light S, or the need to temporarily stop at an intersection due to the presence of an oncoming vehicle B3, can be suppressed. This is because the moving distance does not change during the temporary stop. Thus, even in the case of a temporary stop, etc., it is not affected by the waiting time, and the determination of the right turn or left turn can be made based on the overall driving data of the paths P1 and P3 when turning right or left. Therefore, the determination accuracy of the corresponding driving state can be made high.

[0061] In addition, when the motorcycle is traveling at high speed, the motorcycle turns right by tilting the vehicle body, and when traveling at low speed, the motorcycle turns right by operating the handlebars to change the direction of the wheels, so it is necessary to select variables according to the traveling state and determine the right turn based on the selected variables. In this case as well, in the motorcycle 101, the input image generated based on a plurality of variables is used to determine the right turn, etc., thereby improving the determination accuracy of the corresponding driving action.

[0062] In addition, the learning model 20621 may learn to distinguish between right turns and left turns, and also learn to distinguish between right lane changes and left lane changes. Figure 3 As shown, there is a case where the two-wheeled motor vehicle 101 in the lane on the central side of the road flashes the direction indicator 203 indicating the left direction and changes lanes to the left. In this case, by distinguishing between a left turn and a left lane change, it is possible to avoid determining the left lane change as a left turn, and the determination accuracy can be improved. The same is true when distinguishing between a right turn and a right lane change.

[0063] Alternatively, the learning model 20621 may not distinguish between right turns and left turns, but may learn to determine a driving action such as "the two-wheeled motor vehicle 101 is turning." Alternatively, when the corresponding driving action is determined to be "the two-wheeled motor vehicle 101 is turning" and it is determined that the two-wheeled motor vehicle 101 has completed turning, the action control unit 2066 may turn off the direction indicator 203.

[0064] Reference Figures 4 to 6, the details of the processing and the learning model in the motorcycle 101 are described. In the motorcycle 101, when the direction indicator 203 is operated to turn on, the determination using the learning model 20621 is performed.

[0065] Figure 4 2 is a flowchart showing the processing in the control unit 206.

[0066] In step S401, the travel data acquisition unit 2063 acquires the travel data of the motorcycle 101 from the sensor unit 205. In step S402, the conversion unit 2064 generates an input image based on the travel data.

[0067] In step S403 , the determination unit 2065 inputs the input image to the learning model 20621 .

[0068] Learning model 20621 for reference Figure 5 The learning model 20621 is a learning model obtained by learning using training data that takes a plurality of reference driving data as input and outputs driving actions respectively associated with the plurality of reference driving data. The learning model 20621 is learned using training data TD, which is obtained in the following manner: driving actions are respectively corresponding to a group of a plurality of training images generated based on the reference driving data, namely, training image groups TI1, TI2, TI3, TI4, and TI5. The learning model 20621 is learned by taking a plurality of training image groups TI1, TI2, TI3, TI4, and TI5 as input and outputting respectively corresponding correct driving actions.

[0069] The reference driving data is, for example, data measured in a manner including a prescribed period from when the motorcycle turns right with the direction indicator 203 turned on to when the motorcycle turns right and turns the direction indicator 203 turned off after the right turn. The reference driving data is not data based on the control of the direction indicator 203 by the action control unit 2066, but data based on, for example, manual operation of the direction indicator 203. The same is true for left turns and other driving actions. In addition, in the case of not distinguishing between right turns and left turns, the reference driving data is data measured in a manner including a prescribed period from when the motorcycle turns right or left with the direction indicator 203 turned on to when the motorcycle turns right or left and turns the direction indicator 203 turned off after the right turn or left turn.

[0070] The training image group TI1 corresponds to the driving action of turning right. The training image group TI2 corresponds to the driving action of turning left. The training image group TI3 corresponds to the driving action of changing lanes to the right. The training image group TI4 corresponds to the driving action of changing lanes to the left. The training image group TI5 corresponds to other driving actions (for example, going straight). In addition, in the case of not distinguishing between turning right and turning left, the training image group TI1 and the training image group TI2 correspond to the driving action of "the two-wheeled motor vehicle 101 turns".

[0071] The learning model 20621 is learned by using a general machine learning method such as a neural network method. More specifically, the learning model 20621 is a learning model using a two-dimensional convolutional neural network. In addition, the learning model 20621 may also be a learning model using a one-dimensional convolutional neural network. By setting it as a one-dimensional convolution, it is possible to perform calculations at a high speed. In addition, the learning model 20621 may be configured as a single learning model or as a learning model composed of a plurality of learning models.

[0072] When an input image is input to the learning model 20621, in step S501, the feature quantity of the input image is extracted by convolving the input image with a filter (kernel).

[0073] In step S502, the learning model 20621 classifies the input image based on the feature quantity and generates a classification result. The classification result is, for example, Figure 6 As shown, it is expressed as a classification probability, which is the probability that the input image is classified into each driving action.

[0074] In step S503, the learning model 20621 determines the corresponding driving action based on the classification result and outputs the corresponding driving action. The corresponding driving action is, for example, the driving action with the highest classification probability. Figure 6 In the example of , the corresponding driving action is a right turn. In step S504, the determination unit 2065 obtains the corresponding driving action and the classification probability of the corresponding driving action.

[0075] Through the processing so far, the corresponding driving action can be determined based on a certain driving data. The end of the corresponding driving action can be determined based on the plurality of corresponding driving actions that have been determined. Figure 7 , Figure 8 The end determination of the corresponding driving action will be described.

[0076] Figure 7 7 is a flowchart of the processing in the control unit 206. In step S701, the determination unit 2065 acquires a corresponding driving action and a first classification probability based on first travel data at a first timing.

[0077] Here, refer to Figure 8 An example of the first timing and the second timing and the third timing used in the following description will be described.

[0078] Figure 8 (a) is a schematic diagram of the motorcycle 101 turning right along the path P3. The motorcycle 101 turns right by moving the length (L3) of the path P3 based on the moving distance at the start time of the right turn (MD=0). The time when the right turn starts is set as T=T0.

[0079] The conversion unit 2064 generates an input image based on the driving data detected in a certain moving distance. Here, it is assumed that the conversion unit 2064 generates an input image based on the driving data detected when the motorcycle 101 moves a distance a. More specifically, Figure 8 As shown in (b), the input image is generated by extracting the driving data using a window of distance a. The driving data is added according to the driving, and the old data is removed or deleted from the window of distance a.

[0080] exist Figure 8 In the example (a) of FIG. 1 , the following case is used as an example to illustrate: the input image is generated at the timing when the moving distance MD becomes L1, L2, L3, that is, at the timing T=T1, T2, T3. Specifically, at T=T1 (first timing), based on the driving data corresponding to the moving distance MD=0 to L1 (first driving data), the input image is generated as follows: Figure 8 When the input image I21 is generated, the determination unit 2065 inputs the input image I21 to the learning model 20621, and obtains the following information from the learning model 20621: Figure 8 The corresponding driving action and the classification probability (first classification probability) are shown in (c) of FIG. 1 . In the case of the input image I21 , the corresponding driving action is “right turn” and the classification probability is “0.40”. The process of step S701 is performed in this way.

[0081] After that, the two-wheeled motor vehicle 101 moves, and when it moves to the time T=T2 (second timing) of the distance L2, the input image I22 is generated based on the driving data (second driving data) corresponding to the moving distance up to the distance L2. The corresponding driving action and the classification probability (second classification probability) are also acquired from the learning model 20621 based on the input image I22. In the case of the input image I22, the corresponding driving action is "right turn" and the classification probability is "0.60". In this way, in step S702, the determination unit 2065 acquires the corresponding driving action and the second classification probability based on the second driving data of the second timing.

[0082] Furthermore, the two-wheeled motor vehicle 101 moves, and when it moves to the time T=T3 (third timing) of the distance L3, the input image I23 is generated based on the driving data (third driving data) corresponding to the moving distance up to the distance L3. The corresponding driving action and the classification probability (third classification probability) are also acquired from the learning model 20621 based on the input image I23. In the case of the input image I23, the corresponding driving action is "right turn" and the classification probability is "0.40". In this way, in step S703, the determination unit 2065 acquires the corresponding driving action and the third classification probability based on the third driving data of the third timing.

[0083] In step S704, the operation control unit 2066 determines whether the second classification probability, which is the classification probability at the middle timing among the three timings, is greater than the first classification probability and the third classification probability. If the determination in step S704 is negative, the processing of steps S701 to S704 is repeated.

[0084] If the determination in step S704 is affirmative, then in step S705 , the action control unit 2066 determines that the corresponding driving action has been completed.

[0085] The reason why the above determination can be performed in step S705 is explained. Figure 7 In the case of a right turn shown in (a), the driving data near the start of the right turn (e.g., MD = 0 to L1) is a slightly similar action to the reference driving data used in the learning of the right turn. The degree of consistency (correlation) between the driving data and the reference driving data is slightly small, so although the corresponding driving action is a right turn, its classification probability is relatively small in the classification result.

[0086] Next, the driving data of the right turn (eg, MD = L1 to L2) is closer and more consistent with the reference driving data used in the learning of the right turn. Therefore, in this case, the corresponding driving action is the right turn, and its classification probability is relatively large in the classification result.

[0087] Finally, the driving data near the end of the right turn (e.g., MD = L2 to L3) is slightly similar to the reference driving data used in the learning of the right turn. The driving data is slightly less consistent with the reference driving data, and although the corresponding driving action is a right turn, its classification probability is relatively small in the classification result.

[0088] That is, when the classification probability of right turn changes at different moving distances to reach a peak, it can be determined that the two-wheeled motor vehicle 101 has performed the action closest to the right turn. The fact that the two-wheeled motor vehicle 101 has performed the action closest to the right turn means that the right turn has been completed. Therefore, the action control unit 2066 can determine that the corresponding driving action (here, right turn) has been completed.

[0089] In step S706, the motion control unit 2066 turns off the direction indicator 203. Figure 8 In the example shown in FIG. 2 , the direction indicator 203 is turned off at the time point T=T3.

[0090] In this way, the direction indicator 203 is turned off based on the change in the classification probability in the determination using the driving data of the motorcycle 101 continuously measured within a predetermined period, so that the direction indicator 203 can be turned off at a natural timing after the motorcycle 101 turns right or left. Thus, by processing the data extracted with a width of a predetermined period instead of processing the instantaneous value, the noise within the period is averaged, so that the possibility of the direction indicator being turned off in the middle of a right turn or a left turn can be reduced.

[0091] The second embodiment is described. After the second embodiment, the description of the matters common to the first embodiment is omitted, and only the different aspects are described. In particular, the same effects produced by the same structure are not mentioned one by one in each embodiment.

[0092] Fig. 9 , a block diagram of a motorcycle 101A according to the second embodiment is shown. The motorcycle 101A according to the second embodiment is different from the motorcycle 101 according to the first embodiment in the following aspects: the learning model 20621A includes a learning model 20622 (first learning model) for determining a right turn and a learning model 20623 (second learning model) for determining a left turn. The learning model 20622 for determining a right turn is, for example, a model dedicated to determining a right turn, a lane change to the right, and other aspects of the driving action of the motorcycle 101A. The learning model 20623 for determining a left turn is similarly learned to determine a left turn, a lane change to the left, and other aspects of the driving action of the motorcycle 101A.

[0093] First, the case where the motorcycle 101A turns left is used as an example for explanation. The processing of step (a1) and step (b1) is the same as the processing of step (a) and (b) in the first embodiment. In step (c1), the determination unit 2065 inputs a certain input image (for example, input image I11) to the learning model 20623 for left turn determination. In step (d1), the learning model 20623 for left turn determination calculates the classification probability of the driving action and determines the driving action (left turn) of the motorcycle based on the classification probability. Fig.11An example of the classification result in the case of a left turn is shown in FIG. In step (e1), the end timing of the left turn is estimated based on the change in the classification probability. In step (f1), the movement of the two-wheeled motor vehicle 101A is controlled based on the end timing, and the direction indicator 203 is turned off. Similarly, when the two-wheeled motor vehicle 101A turns right, the processing of steps (a2) to (f2) is repeated, and based on the input images I31, I32, and I33, the direction indicator 203 is finally turned off.

[0094] Reference Fig.12 The processing of the control unit 206A in the second embodiment is described below. In the motorcycle 101A, when the direction indicator 203 is turned on, the judgment using the learning model 20621A is performed. In the control unit 206A, when the direction indicator 203 indicates a right turn, the processing of steps S12011 to S12014 is performed. At this time, the control unit 206A obtains an input indicating a right turn from the operation unit 202 or the direction indicator 203 in step S12011.

[0095] In step S12012, the determination unit 2065 and the action control unit 2066 use the learning model 20622 for right turn determination to determine the end of the right turn. The determination unit 2065 and the action control unit 2066 perform determination of the corresponding driving action (here, right turn) and the end of the right turn in the same manner as the method described in the first embodiment. In step S12013, the determination unit 2065 and the action control unit 2066 determine whether the right turn has ended. When the determination is negative in step S12013, the process returns to step S12012. When the determination is positive in step S12013, in step S12014, the action control unit 2066 turns off the direction indicator 203 that is indicating a right turn.

[0096] Similarly, in the control unit 206A, the processing of steps S12021 to S12024 is performed when the direction indicator 203 indicates a left turn. At this time, the control unit 206A obtains an input indicating a left turn from the operating unit 202 or the direction indicator 203 in step S12021. In step S12022, the determination unit 2065 and the action control unit 2066 use the learning model 20623 for left turn determination to determine the end of the left turn. The determination unit 2065 and the action control unit 2066 perform the determination of the corresponding driving action (left turn) and the determination of the end of the left turn in the same way as the method described in the first embodiment. In step S12023, the determination unit 2065 and the action control unit 2066 determine whether the left turn has ended. When it is determined to be negative in step S12024, the process returns to step S12023. When the determination in step S12023 is affirmative, in step S12024, the operation control unit 2066 turns off the direction indicator 203 indicating a left turn.

[0097] For example, as shown in the paths P1 and P3, the moving distance when turning left is different from the moving distance when turning right. When passing on the left side, the moving distance of turning left is often shorter than the moving distance of turning right, and when passing on the right side, the moving distance of turning right is often shorter than the moving distance of turning left. In the case where the moving distance required to complete the action is different between turning right and turning left, the determination accuracy can be improved by using different learning models to determine right and left turns as in the two-wheeled motor vehicle 101A. In addition, unnecessary activities in the opposite direction (right turn action when turning left) can be excluded from the determination.

[0098] Reference Fig.13 A third embodiment will be described. Fig.13 2 is a block diagram of a motorcycle control system 1300 according to the third embodiment. The motorcycle control system 1300 includes a motorcycle 101B and a determination device 1301 .

[0099] In the motorcycle control system 1300, the conversion unit 2064, the determination unit 2065, and the learning model 20621 of the motorcycle 101 are implemented in the determination device 1301 outside the motorcycle 101B. The determination device 1301 includes a communication unit 13011 for communicating with the outside, a storage unit 13012 for storing various information, a learning model 13013, a sensor unit 13014, a conversion unit 13015, and a determination unit 13016. The functions of the learning model 13013, the conversion unit 13015, and the determination unit 13016 are the same as the functions of the learning model 20621, the conversion unit 2064, and the determination unit 2065 described in the first embodiment. The determination device 1301 is, for example, an information processing device such as a smart phone that can communicate with the motorcycle 101B. The storage unit 13012 may store a dedicated application program for realizing the functions of the learning model 13013 , the sensor unit 13014 , the conversion unit 13015 , and the determination unit 13016 .

[0100] In this case, the action control unit 2066 sends the driving data acquired by the driving data acquisition unit 2063 to the determination device 1301, and acquires information indicating the instruction to turn off the direction indicator from the determination device 1301. At this time, the determination unit 13016 may also use the data acquired by the sensor unit 13014. Thereafter, the action control unit 2066 controls the direction indicator 203 in the same manner as in the first embodiment. By using the determination device 1301 to perform the determination corresponding to the driving action and sending the information indicating the instruction to turn off the direction indicator to the motorcycle 101, it is not necessary for the motorcycle 101 to store the learning model 20621. Thus, the determination corresponding to the driving action and the function of automatically turning off the direction indicator 203 as described in the first embodiment can be easily provided. In addition, the sensor of the determination device can also be used to supplement the insufficient sensor.

[0101] Alternatively, it can be, Fig.14As shown in the motorcycle control system 1400, the driving data acquisition unit 2063 acquires sensor information from an external sensor 1401 outside the motorcycle 101B in addition to the information from the sensor unit 205. For example, when the sensor unit 205 does not have a function of detecting the tilt of the motorcycle 101B, the motorcycle 101B may acquire information indicating the tilt of the motorcycle 101B from the external sensor 1401 connected to the motorcycle 101B by wire. Alternatively, the external sensor 1401 may be a camera or a radar device. By acquiring the information required for determination from the outside of the motorcycle 101B, even when the structure of the sensor unit 205 is insufficient to provide the functions of determining the corresponding driving action and automatically turning off the direction indicator 203, these functions can be provided. In addition, the external sensor 1401 may directly send the sensor information or driving data to the determination device 1301 through wireless communication.

[0102] The motorcycles 101, 101A, and 101B involved in the first to third embodiments are supplemented. In the following description, the motorcycle 101 involved in the first embodiment is used as an example for description, but the following description can also be applied to the motorcycle 101A involved in the second embodiment and 101B involved in the third embodiment. In the motorcycle 101, it is assumed that the following description is made: the reference driving data is, for example, data measured during a prescribed period from when the motorcycle turns right and sets the direction indicator 203 to the open state and the closed state after the right turn. In the motorcycle 101, the straight-ahead movement is performed at a certain time after the direction indicator 203 becomes the open state, and the data related to the straight-ahead movement is not necessary. The prescribed period may be, for example, a certain period from the time when the direction indicator 203 is set to the closed state to a certain time point later than the time when the direction indicator 203 is turned to the open state, including the period during which the motorcycle 101 performs right and left turns.

[0103] In addition, in the first embodiment, the training image groups TI1, TI2, TI3, TI4, and TI5 are described as corresponding to the driving actions. Here, when the predetermined period is set based on the operation of the direction indicator 203 and the data is acquired, since the straight driving is usually not accompanied by the operation of the direction indicator 203, the data cannot be acquired by actual measurement. Therefore, since the actual measured values ​​of various sensors take approximately constant values ​​when driving straight, the values ​​can also be simulated by the information processing device, and the produced data can be used as a training image group as a substitute.

[0104] In addition, the learning model 20621 may be a learning model using an image transformer (Vision Transformer).

[0105] In the first embodiment, the following case is used as an example for explanation: when the classification probability of a right turn changes to a peak value, it is determined that the corresponding driving action (here, a right turn) has been completed. Here, a left turn is used as an example for supplementary explanation. Fig.15 The figure shows the sensor values ​​and speed of each axis (x-axis: moving direction, y-axis: right hand side of the moving direction, z-axis: facing the ground and perpendicular to the x-axis and y-axis) obtained by the gyro sensor when the vehicle makes two left turns. Fig.15 In FIG. 1 , the gyro sensor value G1 of the x-axis, the gyro sensor value G2 of the y-axis, the gyro sensor value G3 of the z-axis, and the speed V1 are shown respectively. Fig.15 In FIG. 1 , the left turn action for tilting the vehicle body is shown as a change in the value of the gyro sensor value G3. Fig.16 It is shown in Fig.15 In the case of movement, the classification probability calculated by the determination unit 2065.

[0106] Here, the determination unit 2065 can either Fig.15 The image data shown in is used as one image data to calculate the classification probability, or a plurality of images drawn based on the values ​​of the respective axes of the gyro sensor and an image drawn based on the speed may be used as inputs to calculate the classification probability.

[0107] Fig.16 The classification probabilities PS of driving actions including straight driving except right turn and left turn, the classification probability PL of left turn, and the classification probability PR of right turn are shown in FIG. Fig.15 The classification probability PL of a left turn reaches a peak at the position where the correlation between the input image and the previously learned image is the largest, that is, the position where the direction indicator 203 is set to the off state, after the gyro sensor value G3 decreases. Fig.16 In the figure, the peak occurs at a distance of about 40 m, then decreases and occurs again at a distance of about 70 m. In other words, it is shown that two left turns have occurred.

[0108] The determination unit 2065 determines driving actions such as going straight, turning right or left based on the change in the classification probability. The operation of turning the direction indicator 203 off is controlled based on the fact that the classification probability PL of the left turn has reached a peak. In addition, the operation of turning the direction indicator 203 off may not be the maximum point of the classification probability of turning left or right, but the point where the classification probability of going straight becomes the maximum in the classification probability after the classification probability of turning left or right becomes the maximum. In this way, the maximum value that is easily affected by road noise can be suppressed from changing due to road noise, thereby suppressing the change in the timing of the extinction determination. In this determination, the prescribed period as the acquisition range of the reference driving data used in learning can be set not to be traced back from the time point when the direction indicator 203 becomes the off state, but to be a range traced back based on a position before this time point.

[0109] In the above example, the value of the gyro sensor and the speed are used as parameters. By considering the speed, for example, it is possible to distinguish between a normal lane change without deceleration and a low-speed movement at an intersection. Therefore, it is possible to turn off the direction indicator 203 only at high speed and when changing lanes without distinguishing between right turns, left turns, and lane changes.

[0110] In the first embodiment, the control unit 206 including the driving data acquisition unit 2063 and the conversion unit 2064 may be provided in the handlebar portion (cockpit portion) provided with an electronic substrate for controlling the instrument and the like. The processing of the control unit 206 is not easily affected by the vibration of the handlebar portion, so the sensor unit including the gyro sensor and the like can be provided in the handlebar portion.

[0111] In the first embodiment, the conversion unit 2064 may generate an image by binarizing information based on the sensor value, and use the image for determination by the determination unit 2065. Fig.17 In, with Fig.15 Similarly, the sensor values ​​and speeds of each axis (x-axis: travel direction, y-axis: right hand side of the travel direction, z-axis: toward the ground and perpendicular to the x-axis and y-axis) obtained by the gyro sensor when a certain driving action is performed are respectively expressed as the gyro sensor value G1 of the x-axis, the gyro sensor value G2 of the y-axis, the gyro sensor value G3 of the z-axis and the speed V1.

[0112] For example, as an example of converting the x-axis gyro sensor value G1 Fig.18 As shown in FIG. 1 , the conversion unit 2064 may convert the sensor value into a binary value. Fig.18In the image, the dark-colored part is set as 1, and the other part is set as 0. It is also possible to generate the information of the chart as an image with a specified number of pixels in this way. Alternatively, the conversion unit 2064 may generate a plurality of images based on each gyro sensor value and speed, and the determination unit 2065 may use the generated images to determine driving actions such as going straight, turning right or left. In addition, at this time, the reference driving data is also similarly binarized and learned. Furthermore, it is also possible that the image of the state in which the sensor value is binarized is made into an image of cumulative values ​​by using a cumulative function, and the influence of road noise or the like on the changes is offset by the cumulative function, thereby suppressing the generation of noise in the image and the reduction in determination accuracy. In addition, instead of using the binarized image for determining driving actions such as going straight, turning right or left, the conversion unit 2064 may use the image based on the binarized state as shown in the figure. Fig.17 The sensor values ​​shown and the image of the integrated value of the sensor values ​​obtained using the integration function are used to determine driving actions such as going straight, turning right or left.

[0113] The above is an explanation of an exemplary embodiment of the present invention. A program is stored in the motorcycle 101, and the program causes the motorcycle 101 to execute the following processing: acquiring driving data of the motorcycle measured continuously within a predetermined period from at least one sensor; and determining a driving action corresponding to the driving data, i.e., a corresponding driving action, based on the driving data and a plurality of reference driving data that are continuous within a predetermined period and respectively associated with a plurality of driving actions of the motorcycle.

[0114] By using the driving data of the motor vehicle continuously measured within a predetermined period to determine the driving action of the motor vehicle, even in the case of a temporary stop, the influence of the waiting time can be avoided, and the influence of temporary noise can be alleviated. Therefore, the determination accuracy of the corresponding driving state can be increased.

[0115] In the motorcycle 101, the plurality of driving actions include right and left turns of the motorcycle, and the action control unit 2066 controls the direction indicator of the motorcycle to be switched from the on state to the off state based on the corresponding driving action of either right turn or left turn.

[0116] The motorcycle 101 calculates the probability that the driving data is classified into each of at least one driving action, that is, the classification probability, based on the driving data and the plurality of reference driving data, and determines the corresponding driving action based on the classification probability.

[0117] Thus, for example, by setting a condition for the classification probability of at least one driving action, a corresponding driving action more suitable for the driving data can be determined. Here, the condition means that the driving action associated with the largest classification probability among multiple classification probabilities is used as the corresponding driving action, and the driving action associated with the classification probability exceeding a predetermined threshold is used as the corresponding driving action.

[0118] In the motorcycle 101, the determination unit 2065 determines the corresponding driving action by inputting an input image as driving data to a learning model and acquiring the corresponding driving action as output from the learning model, wherein the learning model is learned using training data that takes a plurality of reference driving data as input and takes driving actions respectively associated with the plurality of reference driving data as output. Thus, the determination accuracy of the corresponding driving state can be increased.

[0119] In the motor two-wheeled vehicle 101A, the learning model 20621A includes a right turn determination learning model 20622 for determining a right turn and a left turn determination learning model 20623 for determining a left turn. The driving data acquisition unit 2063 acquires information indicating an operation for making the motor two-wheeled vehicle turn right or left. The determination unit 2065 uses the first learning model to determine the corresponding driving action when the operation corresponds to a right turn, and uses the second learning model to determine the corresponding driving action when the operation corresponds to a left turn.

[0120] Thus, even when the moving distance required to complete the action is different between a right turn and a left turn, it is possible to determine the right turn and the left turn using different learning models, thereby improving the determination accuracy.

[0121] In the motorcycle 101, the plurality of driving actions also include right lane change and left lane change of the motorcycle, and the learning model 20621 is learned to be able to distinguish right turn, left turn, right lane change, and left lane change. A right lane change in which the direction indicator 203 is operated in the same manner as a right turn can be distinguished from a left lane change in which the direction indicator 203 is operated in the same manner as a left turn. Therefore, the determination accuracy of the corresponding driving action is improved.

[0122] In the motor two-wheeled vehicle 101, the driving data includes: first driving data obtained within a prescribed period from a first timing; second driving data obtained within a prescribed period from a second timing later than the first timing; and third driving data obtained within a prescribed period from a third timing later than the second timing, and the classification probabilities corresponding to the driving actions include: a first classification probability based on the first driving data; a second classification probability based on the second driving data; and a third classification probability based on the third driving data.

[0123] In the motorcycle 101, the action control unit 2066 obtains the classification probability, and controls the action of the motorcycle based on the corresponding driving action at the third timing or later when the second classification probability is greater than the first and third classification probabilities.

[0124] Thus, the end of the corresponding driving action can be determined based on the fact that the second classification probability, which is located in the middle in terms of timing, reaches a peak. For example, the classification probability of a right turn at the beginning of the right turn (first timing) (first classification probability) is less than the classification probability of a right turn in the middle of the right turn (second timing) (second classification probability). In addition, the classification probability of a right turn at the end of the right turn (third timing) (third classification probability) is also less than the second classification probability. In this way, since the learning data is data for a certain period of time from the time of extinguishing, the end of the corresponding driving action is determined by the change in the classification probability based on its similarity, for example, the direction indicator 203 can be extinguished at a timing that is more natural for the driver.

[0125] In the motorcycle 101, the conversion unit 2064 generates input two-dimensional data based on the driving data, and the determination unit 2065 uses the input two-dimensional data as the driving data, and determines the driving action corresponding to the driving data, that is, the corresponding driving action, based on the input two-dimensional data and a plurality of reference driving data. Thus, even when the driving data includes a plurality of variables, the correlation between the variables can be included by setting the driving data as, for example, one image. By using a plurality of variables for determination, the determination accuracy of the corresponding driving action is improved.

[0126] In the motorcycle 101, the conversion unit 2064 converts the driving data into two-dimensional data when the driving data is one-dimensional data, and the input image is the converted input data based on the driving data after being converted into two-dimensional data. Thus, even when the sensor can only obtain one variable, the corresponding driving action can be determined using the two-dimensional data.

[0127] In the motorcycle 101, the predetermined period is the range required for the motorcycle to move a predetermined distance. By basing the period on distance rather than time, the influence of driving actions that affect time and do not affect the movement of the motorcycle 101, such as temporary stops of the motorcycle 101, can be suppressed.

[0128] The motorcycle 101 further includes a communication unit 2061 that acquires information for updating the learning model 20621 from a terminal outside the motorcycle 101 and updates the learning model 20621. This makes it possible to easily update the learning model.

[0129] In addition, in the motor two-wheeled vehicle 101, a learning model is stored in the storage unit 2062, and the learning model is learned using training data that uses a plurality of reference driving data of the motor two-wheeled vehicle continuously within a specified period as input and uses driving actions respectively associated with the plurality of reference driving data as output. The learning model enables the control unit 206 to function in the following manner: for the input of driving data of the motor two-wheeled vehicle continuously measured within a specified period, the control unit 206 outputs a driving action corresponding to the driving data, i.e., a corresponding driving action.

[0130] In the motorcycle 101, the plurality of reference travel data may include data acquired in a period going back from the time when the direction indicator of the motorcycle was switched from the on state to the off state.

[0131] The plurality of reference driving data may also include data acquired during a period of time traced back from a time point before the time point when the direction indicator of the motor vehicle was switched from the on state to the off state. At this time, the action control unit 2066 may acquire a classification probability, wherein the first classification probability indicates a right turn or a left turn, and the second classification probability indicates a straight line, and when the second classification probability is greater than the first classification probability, the action of the motor vehicle is controlled based on the corresponding driving action at a moment after the second timing. In the case where the plurality of reference driving data is acquired during a period of time traced back from a time point before the time point when the direction indicator was switched from the on state to the off state in this way, the maximum value of the classification probability that is easily affected by road noise can be suppressed from changing due to the road noise, thereby suppressing the change in the timing of the extinction determination.

[0132] In the motorcycle 101, the conversion unit 2064 may convert the travel data into cumulative values ​​using a cumulative function to generate input two-dimensional data. Thus, the influence of road noise on the change is offset by the cumulative function, and the generation of noise in the image and the reduction of determination accuracy can be suppressed.

[0133] In the motorcycle 101, the reference travel data may include data generated by the information processing device, different from the data obtained by actual travel. Thus, even when data cannot be obtained by actual measurement, data can be generated for use in machine learning.

[0134] In the motorcycle 101, the running data acquisition unit 2063 and the determination unit 2065 may be provided at the handlebar of the motorcycle 101. In addition, the running data and the reference running data may be data acquired from a gyro sensor and a speed sensor provided at the handlebar.

[0135] In addition, the above-described embodiments are used to facilitate the understanding of the present invention, and are not used to limit the interpretation of the present invention. The present invention can be changed / improved without departing from its purpose, and the present invention also includes its equivalents. That is, the mode obtained by appropriately applying design changes to each embodiment by a person skilled in the art as long as it has the characteristics of the present invention can also be included in the scope of the present invention. For example, each element and its configuration, condition, shape, size, etc. possessed by each embodiment are not limited to the illustrated elements and their configuration, condition, shape, size, etc., and can be appropriately changed. In addition, the program described in the above embodiment can be stored in a storage medium. The storage medium storing the program can also be a non-transitory computer readable medium. The non-transitory storage medium is not particularly limited, for example, it can also be a storage medium such as a USB memory or a CD-ROM. In addition, each embodiment is an example, and it is self-evident that partial replacement and combination of the structures shown in different embodiments can be performed, and they can also be included in the scope of the present invention as long as they include the characteristics of the present invention.

[0136] <1>

[0137] A program that causes a computer to:

[0138] Acquiring driving data of the motorcycle continuously measured over a predetermined period of time from at least one sensor; and

[0139] A driving action corresponding to the driving data, that is, a corresponding driving action is determined based on the driving data and a plurality of reference driving data that are continuous within the predetermined period and are respectively associated with a plurality of driving actions of the motorcycle.

[0140] <2>

[0141] according to <1> The program, wherein

[0142] The plurality of driving actions include right turn and left turn of the motor two-wheeled vehicle,

[0143] The program causes the computer to further execute the following processing:

[0144] Based on the corresponding driving action of either the right turn or the left turn, control is performed to switch the direction indicator of the motorcycle from an on state to an off state.

[0145] <3>

[0146] according to <1> or <2> The program, wherein

[0147] The determination includes the following processing:

[0148] Based on the driving data and the plurality of reference driving data, a probability that the driving data is classified into each of the at least one driving action, ie, a classification probability is calculated, and the corresponding driving action is determined based on the classification probability.

[0149] <4>

[0150] according to <1> to <3> A program as described in any of the items, wherein

[0151] The determination includes the following processing:

[0152] The corresponding driving action is determined by inputting the driving data into a learning model and obtaining the corresponding driving action as an output from the learning model, wherein the learning model is a model that is learned using training data that takes the multiple reference driving data as input and takes the driving actions respectively associated with the multiple reference driving data as output to be able to distinguish between the right turn and the left turn.

[0153] <5>

[0154] according to <4> The program, wherein

[0155] The learning model includes a first learning model for determining the right turn and a second learning model for determining the left turn,

[0156] The program causes the computer to further execute the following processing:

[0157] acquiring information indicating an operation for causing the motorcycle to make the right turn or the left turn,

[0158] The determination is performed as follows:

[0159] In a case where the operation corresponds to the right turn, the first learning model is used to determine the corresponding driving action,

[0160] When the operation corresponds to the left turn, the corresponding driving action is determined using the second learning model.

[0161] <6>

[0162] according to <4> or <5> The program, wherein

[0163] The plurality of driving actions further include a lane change to the right and a lane change to the left of the motor vehicle,

[0164] The learning model is learned to be able to distinguish the right turn, the left turn, the right lane change, and the left lane change.

[0165] <7>

[0166] according to <3> The program, wherein

[0167] The driving data includes:

[0168] first driving data acquired from a first timing within the prescribed period; second driving data acquired from a second timing later than the first timing within the prescribed period; and third driving data acquired from a third timing later than the second timing within the prescribed period,

[0169] The classification probability of the corresponding driving action includes:

[0170] a first classification probability based on the first driving data; a second classification probability based on the second driving data; and a third classification probability based on the third driving data,

[0171] The control further executes the following processing: acquiring the classification probability, and when the second classification probability is greater than the first classification probability and the third classification probability, controlling the action of the motorcycle based on the corresponding driving action at a time after the third timing.

[0172] <8>

[0173] according to <1> to <7> The program of any one of claims , wherein

[0174] The program causes the computer to further execute the following processing:

[0175] generating input two-dimensional data based on the driving data,

[0176] The determination is performed as follows:

[0177] The input two-dimensional data is used as the driving data, and the driving action corresponding to the driving data, i.e., the corresponding driving action, is determined based on the input two-dimensional data and the multiple reference driving data; the input two-dimensional data is used as the driving data, and the driving action corresponding to the driving data, i.e., the corresponding driving action, is determined based on the input two-dimensional data and the multiple reference driving data.

[0178] <9>

[0179] according to <8> The program, wherein

[0180] The program causes the computer to further execute the following processing:

[0181] When the driving data is one-dimensional data, the driving data is converted into two-dimensional data.

[0182] The input two-dimensional data is data based on the travel data converted into the two-dimensional data.

[0183] <10>

[0184] according to <1> to <9> A program as described in any of the items, wherein

[0185] The predetermined period is a range required for the two-wheeled motor vehicle to move a predetermined distance.

[0186] <11>

[0187] A device comprising:

[0188] an acquisition unit that acquires, from at least one sensor, running data of the motorcycle that is continuously measured within a predetermined period; and

[0189] The determination unit determines a corresponding driving action corresponding to the driving data based on the driving data and a plurality of reference driving data that are continuous within the predetermined period and are respectively associated with a plurality of driving actions of the motorcycle.

[0190] <12>

[0191] according to <11> The device, wherein

[0192] The determination unit determines the corresponding driving action by inputting the driving data into a learning model and obtaining the corresponding driving action as an output from the learning model, wherein the learning model is learned using training data that takes the plurality of reference driving data as input and takes the driving actions respectively associated with the plurality of reference driving data as output,

[0193] The device further includes a communication unit that acquires information for updating the learning model from a terminal outside the device and updates the learning model.

[0194] <13>

[0195] according to <11> or <12> The device, wherein

[0196] The acquisition unit acquires at least a part of the travel data from a terminal capable of communicating with the acquisition unit.

[0197] <14>

[0198] according to <11> to <13> The device of any of the items , wherein

[0199] The acquisition unit acquires at least a part of the travel data from a radar or a camera.

[0200] <15>

[0201] according to <11> to <14> The device of any one of claims , wherein

[0202] The plurality of driving actions include right turn and left turn of the motor two-wheeled vehicle,

[0203] The device further includes an operation control unit configured to switch a direction indicator of the motorcycle from an on state to an off state based on the corresponding driving action of either the right turn or the left turn.

[0204] <16>

[0205] according to <15> The device, wherein

[0206] The determination unit calculates a probability that the driving data is classified into each of the plurality of driving actions, ie, a classification probability, based on the driving data and the plurality of reference driving data, and determines the corresponding driving action based on the classification probability.

[0207] <17>

[0208] according to <15> or <16> The device, wherein

[0209] The determination unit determines the corresponding driving action by inputting the driving data into a learning model and obtaining the corresponding driving action as an output from the learning model, wherein the learning model is a model that is learned using training data that takes the multiple reference driving data as input and takes the driving actions respectively associated with the multiple reference driving data as output to be able to distinguish between a right turn of the motor two-wheeled vehicle and a left turn of the motor two-wheeled vehicle.

[0210] <18>

[0211] according to <17> The device, wherein

[0212] The learning model includes a first learning model for determining the right turn and a second learning model for determining the left turn,

[0213] The device further includes an acquisition unit that acquires information indicating an operation when the motorcycle is caused to make the right turn or the left turn.

[0214] The determination unit determines the corresponding driving action using the first learning model when the operation corresponds to the right turn.

[0215] The determination unit determines the corresponding driving action using the second learning model when the operation corresponds to the left turn.

[0216] <19>

[0217] according to <17> The device, wherein

[0218] The plurality of driving actions further include a lane change to the right and a lane change to the left of the motor vehicle,

[0219] The learning model is learned to be able to distinguish the right turn, the left turn, the right lane change, and the left lane change.

[0220] <20>

[0221] according to <15> to <19> The device of any one of claims , wherein

[0222] The plurality of reference travel data includes data acquired within a period going back from a time point when the direction indicator of the motorcycle was switched from the on state to the off state.

[0223] <21>

[0224] according to <20> The device, wherein

[0225] The driving data includes: first driving data acquired from a first timing within the prescribed period; second driving data acquired from a second timing later than the first timing within the prescribed period; and third driving data acquired from a third timing later than the second timing within the prescribed period,

[0226] The classification probability corresponding to the driving action includes: a first classification probability based on the first driving data; a second classification probability based on the second driving data; and a third classification probability based on the third driving data,

[0227] The action control unit that switches the direction indicator of the motor two-wheeled vehicle from an open state to a closed state based on the corresponding driving action of either the right turn of the motor two-wheeled vehicle or the left turn of the motor two-wheeled vehicle performs the following processing: obtaining the classification probability, and when the second classification probability is greater than the first classification probability and the third classification probability, controlling the action of the motor two-wheeled vehicle based on the corresponding driving action at a time after the third opportunity.

[0228] <22>

[0229] according to <15> to <19> The device of any one of claims , wherein

[0230] The plurality of reference travel data includes data acquired in a period going back from a time point before a time point when the direction indicator of the motorcycle was switched from an on state to a closed state.

[0231] <23>

[0232] according to <22> The device, wherein

[0233] The driving data includes: first driving data acquired from a first timing within the prescribed period; and second driving data acquired from a second timing later than the first timing within the prescribed period,

[0234] The classification probability corresponding to the driving action includes: a first classification probability based on the first driving data; and a second classification probability based on the second driving data,

[0235] The action control unit that switches the direction indicator of the motor two-wheeled vehicle from an open state to a closed state based on the corresponding driving action of either the right turn of the motor two-wheeled vehicle or the left turn of the motor two-wheeled vehicle also performs the following processing: obtaining the classification probability, the first classification probability represents the right turn or the left turn, and the second classification probability represents straight driving, and when the second classification probability is greater than the first classification probability, at a time after the second timing, controlling the action of the motor two-wheeled vehicle based on the corresponding driving action.

[0236] <24>

[0237] according to <11> to <23> The device of any one of claims , wherein

[0238] further comprising a conversion unit that generates input two-dimensional data based on the driving data,

[0239] The determination unit uses the input two-dimensional data as the driving data, and determines a driving action corresponding to the driving data, that is, a corresponding driving action, based on the input two-dimensional data and the plurality of reference driving data.

[0240] <25>

[0241] according to <24> The device, wherein

[0242] When the travel data is one-dimensional data, the conversion unit converts the travel data into two-dimensional data.

[0243] The input two-dimensional data is data based on the travel data converted into the two-dimensional data.

[0244] <26>

[0245] according to <24> or <25> The device, wherein

[0246] The conversion unit converts the travel data into a cumulative value using a cumulative function to generate the input two-dimensional data.

[0247] <27>

[0248] according to <11> to <26> The device of any one of claims , wherein

[0249] The predetermined period is a time range required for the two-wheeled motor vehicle to move a predetermined distance.

[0250] <28>

[0251] according to <11> to <27> The device of any one of claims , wherein

[0252] The reference travel data is different from data obtained by actual travel, and includes data generated by an information processing device.

[0253] <29>

[0254] according to <11> to <28> The device of any one of claims , wherein

[0255] The acquisition unit and the determination unit are provided on a handlebar portion of the motorcycle.

[0256] <30>

[0257] according to <29> The device, wherein

[0258] The running data and the reference running data are data acquired from a gyro sensor and a speed sensor provided on the handlebar portion.

[0259] <31>

[0260] A learning model is learned using training data that takes as input a plurality of reference driving data of a motorcycle continuously within a predetermined period and takes as output driving actions respectively associated with the plurality of reference driving data,

[0261] The learning model enables the computer to function in the following ways:

[0262] In response to the input of the driving data of the motorcycle continuously measured during the predetermined period, a corresponding driving action which is a driving action corresponding to the driving data is output.

[0263] <32>

[0264] according to <31> The learning model, wherein

[0265] The predetermined period is a time range required for the two-wheeled motor vehicle to move a predetermined distance.

[0266] <33>

[0267] A method comprising performing the following processing by a computer:

[0268] Acquiring driving data of the motorcycle continuously measured over a predetermined period of time from at least one sensor; and

[0269] A corresponding driving action corresponding to the driving data is determined based on the driving data and a plurality of reference driving data that are continuous within the predetermined period and are respectively associated with a plurality of driving actions of the motorcycle.

[0270] <34>

[0271] according to <33> The method, wherein

[0272] The predetermined period is a time range required for the two-wheeled motor vehicle to move a predetermined distance.

[0273] Description of Reference Numerals

[0274] 101, 101A, 101B: two-wheeled motor vehicle; 203: direction indicator; 205: sensor unit; 206, 206A: control unit; 1100: two-wheeled motor vehicle control system; 1101: determination device; 1102: sensing terminal; 2063: driving data acquisition unit; 2064: conversion unit; 2065: determination unit; 2066: action control unit; 20612: learning model.

Claims

1. A program that causes a computer to perform the following processing: Acquiring driving data of the motorcycle continuously measured over a predetermined period of time from at least one sensor; and A driving action corresponding to the driving data, that is, a corresponding driving action is determined based on the driving data and a plurality of reference driving data that are continuous within the predetermined period and are respectively associated with a plurality of driving actions of the motorcycle.

2. The program according to claim 1, wherein: The plurality of driving actions include right turn and left turn of the motor two-wheeled vehicle, The program causes the computer to further execute the following processing: Based on the corresponding driving action of either the right turn or the left turn, control is performed to switch the direction indicator of the motorcycle from an on state to an off state.

3. The program according to claim 1 or 2, wherein: The determination includes the following processing: Based on the driving data and the plurality of reference driving data, a probability that the driving data is classified into each of the plurality of driving actions, namely, a classification probability is calculated, and the corresponding driving action is determined based on the classification probability.

4. The program according to any one of claims 1 to 3, wherein: The determination includes the following processing: The corresponding driving action is determined by inputting the driving data into a learning model and obtaining the corresponding driving action as an output from the learning model, wherein the learning model is a model that is learned using training data that takes the multiple reference driving data as input and takes driving actions respectively associated with the multiple reference driving data as output and is capable of distinguishing between a right turn of the motor two-wheeled vehicle and a left turn of the motor two-wheeled vehicle.

5. The program according to claim 4, wherein: The learning model includes a first learning model for determining the right turn and a second learning model for determining the left turn, The program causes the computer to further execute the following processing: acquiring information indicating an operation for causing the motorcycle to make the right turn or the left turn, The determination is performed as follows: In a case where the operation corresponds to the right turn, the first learning model is used to determine the corresponding driving action, When the operation corresponds to the left turn, the corresponding driving action is determined using the second learning model.

6. The program according to claim 4 or 5, wherein: The plurality of driving actions further include a lane change to the right and a lane change to the left of the motor vehicle, The learning model is learned to be able to distinguish the right turn, the left turn, the right lane change, and the left lane change.

7. The program according to claim 3, wherein: The driving data includes: first driving data acquired from a first timing within the prescribed period; second driving data acquired from a second timing later than the first timing within the prescribed period; and third driving data acquired from a third timing later than the second timing within the prescribed period, The classification probability of the corresponding driving action includes: a first classification probability based on the first driving data; a second classification probability based on the second driving data; and a third classification probability based on the third driving data, The program causes the computer to also perform the following control: switching the direction indicator of the motor vehicle from an on state to a off state based on the corresponding driving action of either the right turn of the motor vehicle or the left turn of the motor vehicle; and obtaining the classification probability, and when the second classification probability is greater than the first classification probability and the third classification probability, controlling the movement of the motor vehicle based on the corresponding driving action at a time after the third opportunity.

8. The program according to any one of claims 1 to 7, wherein: The program causes the computer to further execute the following processing: generating input two-dimensional data based on the driving data, The determination is performed as follows: The input two-dimensional data is used as the driving data, and a driving action corresponding to the driving data, that is, a corresponding driving action, is determined based on the input two-dimensional data and the plurality of reference driving data.

9. The program according to claim 8, wherein: The program causes the computer to further execute the following processing: When the driving data is one-dimensional data, the driving data is converted into two-dimensional data. The input two-dimensional data is data based on the travel data converted into the two-dimensional data.

10. The program according to any one of claims 1 to 9, wherein: The predetermined period is a time range required for the two-wheeled motor vehicle to move a predetermined distance.

11. A device comprising: an acquisition unit that acquires, from at least one sensor, running data of the motorcycle that is continuously measured within a predetermined period; and The determination unit determines a corresponding driving action corresponding to the driving data based on the driving data and a plurality of reference driving data that are continuous within the predetermined period and are respectively associated with a plurality of driving actions of the motorcycle.

12. The device according to claim 11, wherein The determination unit determines the corresponding driving action by inputting the driving data into a learning model and obtaining the corresponding driving action as an output from the learning model, wherein the learning model is learned using training data that takes the plurality of reference driving data as input and takes the driving actions respectively associated with the plurality of reference driving data as output, The device further includes a communication unit that acquires information for updating the learning model from a terminal outside the device and updates the learning model.

13. The device according to claim 11 or 12, wherein: The acquisition unit acquires at least a part of the travel data from a terminal capable of communicating with the acquisition unit.

14. The device according to any one of claims 11 to 13, wherein: The acquisition unit acquires at least a part of the travel data from a radar or a camera.

15. The device according to any one of claims 11 to 14, wherein: The plurality of driving actions include right turn and left turn of the motor two-wheeled vehicle, The device further includes an operation control unit configured to switch a direction indicator of the motorcycle from an on state to an off state based on the corresponding driving action of either the right turn or the left turn.

16. The device according to claim 15, wherein: The determination unit calculates a probability that the driving data is classified into each of the plurality of driving actions, ie, a classification probability, based on the driving data and the plurality of reference driving data, and determines the corresponding driving action based on the classification probability.

17. The device according to claim 15 or 16, wherein: The determination unit determines the corresponding driving action by inputting the driving data into a learning model and obtaining the corresponding driving action as an output from the learning model, wherein the learning model is a model that is learned using training data that takes the multiple reference driving data as input and takes the driving actions respectively associated with the multiple reference driving data as output to be able to distinguish between a right turn of the motor two-wheeled vehicle and a left turn of the motor two-wheeled vehicle.

18. The device according to claim 17, wherein: The learning model includes a first learning model for determining the right turn and a second learning model for determining the left turn, The device further includes an acquisition unit that acquires information indicating an operation when the motorcycle is caused to make the right turn or the left turn. The determination unit determines the corresponding driving action using the first learning model when the operation corresponds to the right turn. The determination unit determines the corresponding driving action using the second learning model when the operation corresponds to the left turn.

19. The device according to claim 17, wherein: The plurality of driving actions further include a lane change to the right and a lane change to the left of the motor vehicle, The learning model is learned to be able to distinguish the right turn, the left turn, the right lane change, and the left lane change.

20. The device according to any one of claims 15 to 19, wherein: The plurality of reference travel data includes data acquired within a period going back from a time point when the direction indicator of the motorcycle was switched from the on state to the off state.

21. The device according to claim 20, wherein: The driving data includes: first driving data acquired from a first timing within the prescribed period; second driving data acquired from a second timing later than the first timing within the prescribed period; and third driving data acquired from a third timing later than the second timing within the prescribed period, The classification probability corresponding to the driving action includes: a first classification probability based on the first driving data; a second classification probability based on the second driving data; and a third classification probability based on the third driving data, The action control unit that switches the direction indicator of the motor two-wheeled vehicle from an open state to a closed state based on the corresponding driving action of either the right turn of the motor two-wheeled vehicle or the left turn of the motor two-wheeled vehicle performs the following processing: obtaining the classification probability, and when the second classification probability is greater than the first classification probability and the third classification probability, controlling the action of the motor two-wheeled vehicle based on the corresponding driving action at a time after the third opportunity.

22. The device according to any one of claims 15 to 19, wherein: The plurality of reference travel data includes data acquired in a period going back from a time point before a time point when the direction indicator of the motorcycle was switched from an on state to a closed state.

23. The device according to claim 22, wherein: The driving data includes: first driving data acquired from a first timing within the prescribed period; and second driving data acquired from a second timing later than the first timing within the prescribed period, The classification probability corresponding to the driving action includes: a first classification probability based on the first driving data; and a second classification probability based on the second driving data, The action control unit that switches the direction indicator of the motor two-wheeled vehicle from an open state to a closed state based on the corresponding driving action of either the right turn of the motor two-wheeled vehicle or the left turn of the motor two-wheeled vehicle also performs the following processing: obtaining the classification probability, the first classification probability represents the right turn or the left turn, and the second classification probability represents straight driving, and when the second classification probability is greater than the first classification probability, at a time after the second timing, controlling the action of the motor two-wheeled vehicle based on the corresponding driving action.

24. The device according to any one of claims 11 to 23, wherein: further comprising a conversion unit that generates input two-dimensional data based on the driving data, The determination unit uses the input two-dimensional data as the driving data, and determines a driving action corresponding to the driving data, that is, a corresponding driving action, based on the input two-dimensional data and the plurality of reference driving data.

25. The device according to claim 24, wherein: When the travel data is one-dimensional data, the conversion unit converts the travel data into two-dimensional data. The input two-dimensional data is data based on the travel data converted into the two-dimensional data.

26. The device according to claim 24 or 25, wherein: The conversion unit converts the travel data into a cumulative value using a cumulative function to generate the input two-dimensional data.

27. The device according to any one of claims 11 to 26, wherein: The predetermined period is a time range required for the two-wheeled motor vehicle to move a predetermined distance.

28. The device according to any one of claims 11 to 27, wherein The reference travel data is different from data obtained by actual travel, and includes data generated by an information processing device.

29. The device according to any one of claims 11 to 28, wherein The acquisition unit and the determination unit are provided on a handlebar portion of the motorcycle.

30. The device according to claim 29, wherein: The running data and the reference running data are data acquired from a gyro sensor and a speed sensor provided on the handlebar portion.

31. A learning model obtained by learning using training data having as input a plurality of reference driving data of a motorcycle continuously within a predetermined period and having as output driving actions respectively associated with the plurality of reference driving data, The learning model enables the computer to function in the following ways: In response to the input of the driving data of the motorcycle continuously measured during the predetermined period, a corresponding driving action which is a driving action corresponding to the driving data is output.

32. The learning model according to claim 31, wherein: The predetermined period is a time range required for the two-wheeled motor vehicle to move a predetermined distance.

33. A method comprising performing the following processing by a computer: Acquiring driving data of the motorcycle continuously measured over a predetermined period of time from at least one sensor; and A corresponding driving action corresponding to the driving data is determined based on the driving data and a plurality of reference driving data that are continuous within the predetermined period and are respectively associated with a plurality of driving actions of the motorcycle.

34. The method of claim 33, wherein: The predetermined period is a time range required for the two-wheeled motor vehicle to move a predetermined distance.

Images