Control device for human-powered vehicle, learning method, control method for human-powered vehicle, and computer program

Abstract

The present application provides a kind of shortening the time required for learning model for automatic control and also can realize automatic control in unlearned scene Human-powered vehicle control device, learning method, the control method of human-powered vehicle and computer program.Person-powered vehicle control device has: obtaining unit, obtains the input information related to the travel of human-powered vehicle;Storage unit, store first learning model, which is learned based on the input information obtained, to output the output information related to the control of the device mounted on human-powered vehicle;Control unit, using the control data determined by the output information obtained by inputting input information into first learning model controls device;And supplementary processing unit, perform the processing using second learning model to supplement first learning model in storage unit, which is learned by input information in at least one of human-powered vehicle and human-powered vehicle rider different human-powered vehicle.

Description

Technical Field

[0001] This invention relates to a control device, learning method, control method for a human-powered vehicle, and computer program.

[0002] With the development of electrification in human-powered vehicles, automatic control of mounted devices, including transmissions, braking systems, and auxiliary devices, has been achieved. A scheme using a model learned through deep learning (Patent Document 1, etc.) is proposed to output information related to the control of the mounted devices, given input information obtained from at least one of a speed sensor, cadence sensor, torque sensor, and camera installed in the human-powered vehicle. Background Technology

[0003] Existing technical documents

[0004] Patent documents

[0005] Patent Document 1: Japanese Patent No. 6985217. Summary of the Invention

[0006] The problem that the invention aims to solve

[0007] By utilizing deep learning to achieve automatic control of human-powered vehicles, the automatic control can be optimized for each rider based on their physical characteristics, preferences, and other factors.

[0008] The purpose of this invention is to provide a control device, learning method, control method for a human-powered vehicle, and computer program that shortens the learning time required for an automatic control model and enables automatic control even in scenarios where the model has not been learned.

[0009] means for solving problems

[0010] (1) A control device for a human-powered vehicle according to a first aspect of the present invention includes: an acquisition unit that acquires input information related to the driving of a human-powered vehicle; a storage unit that stores a first learning model that learns based on the acquired input information to output output information related to the control of a device mounted on the human-powered vehicle; a control unit that controls the device using control data determined based on output information obtained by inputting the input information to the first learning model; and a supplementary processing unit that performs processing to supplement the first learning model in the storage unit with a second learning model that has been learned using input information from at least one of the human-powered vehicle and the rider of the human-powered vehicle.

[0011] According to the human-powered vehicle control device of the first aspect described above, the first learning model for controlling the device is supplemented by other second learning models. In the case of learning about at least one of the human-powered vehicle and the rider separately, learning about scenarios where the vehicle is not in motion is not performed. By supplementing the first learning model with other second learning models, the time required to learn input information from unlearned driving scenarios can be shortened, and automatic control itself can also be achieved.

[0012] (2) In the human-powered vehicle control device of the second aspect of the present invention, the supplementary processing unit updates at least a portion of the first learning model in the storage unit by means of the second learning model.

[0013] According to the second aspect of the above-mentioned human-driven vehicle control device, by using other second learning models as supplements, the time for learning input information of unlearned driving scenarios can be shortened, and automatic control itself can also be achieved.

[0014] (3) In the human-powered vehicle control device of the third aspect of the present invention, the supplementary processing unit uses the input information acquired by the acquisition unit and the output information output when the input information is input to the second learning model as learning data to learn the first learning model.

[0015] According to the third aspect of the above-mentioned human-driven vehicle control device, by supplementing the learning with the output information output from other second learning models, the learning time for input information of unlearned driving scenarios can be shortened, and automatic control itself can also be realized.

[0016] (4) In the human-powered vehicle control device according to the fourth aspect of the present invention, in any of the first to third aspects, the first learning model learns using input information acquired by the acquisition unit in multiple different driving scenarios. For a driving scenario that is not learned but is different from the driving scenario that the first learning model has already learned, and a driving scenario that the second learning model has already learned, the supplementary processing unit uses the second learning model to supplement the first learning model.

[0017] Based on the fourth aspect of the above-mentioned human-driven vehicle control device, supplements are made according to various driving scenario categories. This can shorten the learning time for input information of driving scenarios that have not yet been learned, and can also realize automatic control itself.

[0018] (5) In the human-powered vehicle control device according to the fifth aspect of the present invention, in any of the first to third aspects, the first learning model includes a plurality of learning models stored according to each driving scenario, and the supplementary processing unit uses a part of a second learning model that has been learned for the unlearned driving scenario as a learning model corresponding to an unlearned driving scenario that is different from the driving scenario that has been learned by the first learning model.

[0019] According to the fifth aspect of the above-mentioned human-driven vehicle control device, by storing the learning model of the unlearned scenario as the learning model of the first learning model, the time for learning input information can be shortened, and automatic control itself can also be realized.

[0020] (6) In the human-powered vehicle control device according to the sixth aspect of the present invention, the driving scenario is divided into at least one of highway, off-road, and urban.

[0021] According to the human-driven vehicle control device in the sixth aspect mentioned above, by using other second learning models as supplements, the learning time for input information of driving scenarios that have not been learned in highways, off-road, and urban areas can be shortened, and automatic control itself can also be achieved.

[0022] (7) The human-powered vehicle control device according to the seventh aspect of the present invention, in any of the above-mentioned fourth to sixth aspects, the driving scenario is divided into at least one of uphill, flat and downhill.

[0023] According to the seventh aspect of the above-mentioned human-driven vehicle control device, by using other second learning models as supplements, the learning time for input information of driving scenarios that have not been learned in uphill, flat, and downhill driving can be shortened, and automatic control itself can also be achieved.

[0024] (8) In the human-powered vehicle control device according to the eighth aspect of the present invention, in any of the first to seventh aspects of the above-mentioned human-powered vehicle control device, the supplementary processing unit uses a second learning model as follows: the second learning model is the output information output when the same input information is input, and the model is similar to the output information output when the same input information is input to the first learning model.

[0025] According to the human-driven vehicle control device in the eighth aspect above, by using other similar second learning models as supplements, the time for learning input information of unlearned driving scenarios can be shortened, and automatic control itself can also be achieved.

[0026] (9) In the human-powered vehicle control device of the ninth aspect of the present invention, the first learning model learns using input information acquired by the acquisition unit in multiple different driving scenarios, and the supplementary processing unit uses, among multiple second learning models, output information output when input information of a driving scenario already learned by the first learning model is input, and a second learning model similar to the output information output when the input information is input to the first learning model.

[0027] According to the human-driven vehicle control device of the ninth aspect above, by using other similar second learning models as supplements, the learning time for input information of driving scenarios that have not been learned in highways, off-road, and urban areas can be shortened, and automatic control itself can also be achieved.

[0028] (10) In the human-powered vehicle control device of the tenth aspect of the present invention, the first learning model and the second learning model include a plurality of learning models stored according to each driving scenario, and the supplementary processing unit is configured to, among the plurality of second learning models, use the output information output when input information of a driving scenario that has been learned by the first learning model is input, and a second learning model similar to the output information output when the input information is input to the first learning model, to obtain a learning model corresponding to an unlearned driving scenario that is different from the driving scenario that has been learned.

[0029] According to the human-powered vehicle control device of the tenth aspect above, by supplementing the learning model included in the other second learning model as the learning model of the first learning model, the time for learning input information of the unlearned driving scenario can be shortened, and automatic control itself can also be realized.

[0030] (11) In any of the first to seventh aspects of the human-powered vehicle control device according to the eleventh aspect of the present invention, the supplementary processing unit uses a second learning model, which is a model used in other similar human-powered vehicle control devices based on control data determined by the same input information.

[0031] According to the human-driven vehicle control device of the eleventh aspect above, by supplementing it with other second learning models that output similar control data, the time for learning input information of unlearned driving scenarios can be shortened, and automatic control itself can also be achieved.

[0032] (12) In the human-powered vehicle control device according to the twelfth aspect of the present invention, in the human-powered vehicle control device of the eleventh aspect described above, the first learning model learns using input information acquired by the acquisition unit in multiple different driving scenarios, and the supplementary processing unit uses, among multiple second learning models, control data determined based on output information when input information of a driving scenario already learned by the first learning model is input, and a second learning model similar to control data determined based on output information when the input information is input to the first learning model.

[0033] According to the human-driven vehicle control device of the twelfth aspect above, by supplementing it with other second learning models that output similar control data, the time for learning input information of unlearned driving scenarios can be shortened, and automatic control itself can also be achieved.

[0034] (13) In the eleventh aspect of the human-powered vehicle control device according to the thirteenth aspect of the present invention, the first learning model and the second learning model include a plurality of learning models stored according to each driving scenario, and the supplementary processing unit is configured to, among the plurality of second learning models, use control data determined based on the output information when input information of a driving scenario already learned by the first learning model is input, and a second learning model similar to the control data determined based on the output information when the input information is input to the first learning model, to obtain a learning model among the plurality of learning models included in the second learning model that is not learned for a driving scenario different from the driving scenario already learned.

[0035] According to the human-driven vehicle control device of the thirteenth aspect above, by supplementing the learning model of the first learning model with the learning model of other second learning models that output similar control data, the time for learning input information of driving scenarios that have not been learned can be shortened, and automatic control itself can also be realized.

[0036] (14) In the human-powered vehicle control device according to the fourteenth aspect of the present invention, in any of the first to seventh aspects, the supplementary processing unit uses a second learning model in which input information from other human-powered vehicles that are the same as or similar to the human-powered vehicle in terms of at least one of type and size to be learned in a plurality of second learning models.

[0037] According to the human-powered vehicle control device of the fourteenth aspect above, by supplementing it with other second learning models of similar type and size of human-powered vehicles, the time for learning input information of unlearned driving scenarios can be shortened, and automatic control itself can also be achieved.

[0038] (15) In the human-powered vehicle control device according to the fifteenth aspect of the present invention, in any of the first to seventh aspects, the supplementary processing unit uses, in a plurality of second learning models, a second learning model that has been learned using input information from other human-powered vehicles equipped with devices of the same or similar category and manufacturer as the device.

[0039] According to the human-powered vehicle control device of the fifteenth aspect above, by supplementing it with a second learning model of a human-powered vehicle with the same or similar mounted devices, the time for learning input information of driving scenarios that have not been learned can be shortened, and automatic control itself can also be achieved.

[0040] (16) The human-powered vehicle control device according to the sixteenth aspect of the present invention, in the human-powered vehicle control device of the fifteenth aspect above, the device is identified as belonging to at least one of a transmission device, suspension, seat post, braking device, and auxiliary device.

[0041] According to the human-powered vehicle control device of the sixteenth aspect above, by supplementing it with a second learning model of a human-powered vehicle that is the same as or similar to the transmission, suspension, seat post, braking device and auxiliary device, the learning time can be shortened and automatic control itself can be achieved.

[0042] (17) In the human-powered vehicle control device according to the seventeenth aspect of the present invention, in any of the first to seventh aspects, the supplementary processing unit uses, among a plurality of second learning models, a second learning model that has been learned using input information from a human-powered vehicle of a rider of the same or similar type as the rider of the human-powered vehicle.

[0043] According to the human-powered vehicle control device of the seventeenth aspect above, by supplementing it with other second learning models that have been learned in human-powered vehicles of the same or similar rider type, the time for learning input information of unlearned driving scenarios can be shortened, and automatic control itself can also be achieved.

[0044] (18) The human-powered vehicle control device according to the eighteenth aspect of the present invention transmits the first learning model stored in the storage unit to other human-powered vehicle control devices in any of the first to seventeenth aspects of the present invention.

[0045] According to the human-powered vehicle control device of the eighteenth aspect above, it is possible to send a first learning model that has been learned in each human-powered vehicle to supplement the first learning model in other human-powered vehicles.

[0046] (19) In the human-powered vehicle control device according to the nineteenth aspect of the present invention, in any of the first to eighteenth aspects, the second learning model is a model obtained by statistically processing parameters including one of the weights and biases of multiple models learned in multiple other human-powered vehicles.

[0047] According to the human-powered vehicle control device of the nineteenth aspect above, the model obtained by statistically processing the parameters of at least one of the weights and biases of the first learning model learned in each human-powered vehicle can be used to supplement the first learning model in other human-powered vehicles.

[0048] (20) According to the twentieth aspect of the invention, a learning method wherein a computer mounted on a human-powered vehicle performs: learning a first learning model based on input information related to the driving of the human-powered vehicle to output output information related to the control of a device mounted on the human-powered vehicle; selecting from the outside a second learning model that has been learned using input information from at least one of the human-powered vehicle and the rider of the human-powered vehicle; and performing processing to supplement the first learning model with the selected second learning model.

[0049] According to the learning method in the twentieth aspect mentioned above, by using other second learning models to supplement the first learning model, the time required to learn the input information of the unlearned driving scenario can be shortened, and automatic control itself can also be achieved.

[0050] (21) A control method for a human-powered vehicle according to a twenty-first aspect of the present invention, wherein a computer mounted on the human-powered vehicle performs: learning a first learning model based on input information related to the driving of the human-powered vehicle to output output information related to the control of a device mounted on the human-powered vehicle; selecting a second learning model from an external source that has been learned using input information from at least one of the human-powered vehicle and the rider of the human-powered vehicle; performing processing to supplement the first learning model with the selected second learning model; determining control data based on output information obtained by inputting the input information to the supplemented first learning model; and controlling the device according to the determined control data.

[0051] According to the control method of the human-driven vehicle in aspect 21 above, by using other second learning models to supplement the first learning model, the time for learning input information of the unlearned driving scenario can be shortened, and automatic control itself can also be achieved.

[0052] (22) According to the twenty-second aspect of the present invention, a computer program causes a computer equipped with a human-powered vehicle to perform the following processing: learning a first learning model based on input information related to the driving of the human-powered vehicle to output output information related to the control of a device mounted on the human-powered vehicle; selecting from the outside a second learning model that has been learned using input information from at least one of the human-powered vehicle and the rider of the human-powered vehicle; and performing processing to supplement the first learning model with the selected second learning model.

[0053] According to the computer program in aspect 22 above, by using other second learning models to supplement the first learning model, the time required to learn input information of unlearned driving scenarios can be shortened, and automatic control itself can also be achieved.

[0054] Invention Effects

[0055] According to the control device, learning method, control method and computer program for human-powered vehicles involved in the invention, the learning time required for models used for automatic control can be shortened, and automatic control can be achieved even in scenarios where no model has been learned. Attached Figure Description

[0056] Figure 1 This is a side view of a manually driven vehicle that uses the control device described in the first embodiment;

[0057] Figure 2 This is a block diagram illustrating the structure of the control device;

[0058] Figure 3 This is a diagram illustrating an example of the first learning model;

[0059] Figure 4 This is a diagram illustrating another example of the first learning model;

[0060] Figure 5 This is a diagram showing the control device and information processing device of the first embodiment;

[0061] Figure 6 It is a block diagram illustrating the structure of an information processing device;

[0062] Figure 7 This is a flowchart illustrating an example of a learning method for the first learning model in the first embodiment;

[0063] Figure 8 This is a flowchart illustrating an example of control processing using the first learning model of the first embodiment;

[0064] Figure 9This is a block diagram illustrating the structure of the control device according to the second embodiment;

[0065] Figure 10 This is a flowchart illustrating an example of a learning method for the first learning model in the third implementation method;

[0066] Figure 11 This is a schematic diagram of the first learning model in the fourth implementation method;

[0067] Figure 12 This is a flowchart illustrating an example of a learning method for the first learning model in the fourth embodiment;

[0068] Figure 13 This is a flowchart illustrating an example of a learning method for the first learning model in the fourth embodiment;

[0069] Figure 14 This is a flowchart illustrating an example of control processing using the first learning model of the fourth implementation;

[0070] Figure 15 This is a diagram illustrating the first learning model of the fifth implementation method;

[0071] Figure 16 This is a flowchart illustrating an example of a learning method for the first learning model in the fifth embodiment;

[0072] Figure 17 This is a flowchart illustrating an example of a learning method for the first learning model in the fifth embodiment;

[0073] Figure 18 This diagram illustrates the processing performed by the supplementary processing unit in the fifth embodiment;

[0074] Figure 19 This is a flowchart illustrating an example of control processing using the first learning model of the fifth embodiment;

[0075] Figure 20 This is a diagram illustrating the control device and information processing device according to the sixth embodiment;

[0076] Figure 21 This is a flowchart illustrating an example of the processing sequence of the information processing apparatus according to the sixth embodiment;

[0077] Figure 22 This is a schematic diagram of the second learning model in the sixth implementation method. Detailed Implementation

[0078] The following descriptions of the embodiments are examples of the possible ways in which the human-powered vehicle control device, learning method, human-powered vehicle control method, and computer program of the present invention can be adopted, and are not intended to limit the methods. The human-powered vehicle control device, learning method, human-powered vehicle control method, and computer program of the present invention can be adopted in ways different from the respective embodiments, such as variations of the embodiments and combinations of at least two non-contradictory variations.

[0079] In the following descriptions of various embodiments, terms indicating direction such as front, back, front, rear, left, right, horizontal, up, and down are used based on the orientation of the user sitting in the seat of a manually driven vehicle.

[0080] In the following embodiments, the human-powered vehicle control device of the present invention will be referred to as a control device.

[0081] (First Implementation)

[0082] Figure 1 A side view of a human-powered vehicle 1 using the control device 100 of the first embodiment. The human-powered vehicle 1 is a vehicle that uses human power at least partially for its propulsion. Vehicles that use only an internal combustion engine or an electric motor as their propulsion are excluded from the human-powered vehicle 1 of this embodiment. The human-powered vehicle 1 includes, for example, bicycles such as mountain bikes, road bikes, cross-country bikes, city bikes, and electric-assisted bicycles (e-bikes).

[0083] The human-powered vehicle 1 has a vehicle body 10, handlebars 12, front wheel 14, rear wheel 16, and seat 18. The human-powered vehicle 1 also has a drive mechanism 20, components 30, operating device 40, battery 50, and sensor 60.

[0084] The vehicle body 10 includes a frame 10A and a front fork 10B. The front wheel 14 is rotatably supported on the front fork 10B. The rear wheel 16 is rotatably supported on the frame 10A. The handlebars 12 are supported on the frame 10A in a manner that allows the direction of travel of the front wheel 14 to be changed.

[0085] The drive mechanism 20 includes a crank 21, a first sprocket assembly 23, a second sprocket assembly 25, a chain 27, and a pair of pedals 29.

[0086] Crank 21 includes crankshaft 21A, right crank 21B, and left crank 21C. Crankshaft 21A is rotatably supported on frame 10A. Right crank 21B and left crank 21C are respectively connected to crankshaft 21A. One of a pair of pedals 29 is rotatably supported on right crank 21B. The other of a pair of pedals 29 is rotatably supported on left crank 21C.

[0087] The first sprocket assembly 23 is integrally rotatably connected to the crankshaft 21A. The first sprocket assembly 23 includes one or more sprockets 23A. In one example, the first sprocket assembly 23 includes a plurality of sprockets 23A with different outer diameters.

[0088] The second sprocket assembly 25 is rotatably supported on the rear hub of the rear wheel 16. The second sprocket assembly 25 includes one or more sprockets 25A. In one example, the second sprocket assembly 25 includes a plurality of sprockets 25A with different outer diameters.

[0089] Chain 27 is wound around either sprocket 23A in the first sprocket assembly 23 and either sprocket 25A in the second sprocket assembly 25. When crank 21 rotates forward by a human-driven force applied to pedal 29, sprocket 23A rotates forward together with crank 21, and the rotation of sprocket 23A is transmitted via chain 27 to sprocket 25A in the second sprocket assembly 25. The rotation of sprocket 25A causes rear wheel 16 to rotate. A belt or driveshaft can be used instead of chain 27.

[0090] In one example, the control device 100 is mounted on the battery 50, speedometer, transmission unit, etc. of the manually driven vehicle 1. The control device 100 is connected to the device 30, the operating device 40, and the battery 50. The connection method and details of the control device 100 will be described later.

[0091] The human-powered vehicle 1 includes a device 30 that operates by electricity supplied by a battery 50 and is controlled by a control device 100. The device 30 includes a transmission 31, a suspension 33, a seat post 35, a braking device 37, and an auxiliary device 39. The device 30 operates primarily through control by the control device 100 based on operations performed in the operating device 40. The control object of the control device 100 is at least one of the transmission 31, suspension 33, seat post 35, braking device 37, and auxiliary device 39 within the device 30.

[0092] The gearbox 31 changes the ratio of the rear wheel 16's rotational speed to the crank 21's rotational speed, which is the gear ratio of the manually driven vehicle 1. The gear ratio is expressed as the ratio of the output rotational speed of the gearbox 31 to the input rotational speed. The gear ratio can be expressed as "gear ratio = output rotational speed / input rotational speed". In the first example, the gearbox 31 is an external derailleur (rear derailleur) that changes the connection between the second sprocket assembly 25 and the chain 27. In the second example, the gearbox 31 is an external derailleur (front derailleur) that changes the connection between the first sprocket assembly 23 and the chain 27. In the third example, it is an internal derailleur located on the hub of the rear wheel 16. The gearbox 31 can be a continuously variable transmission (CVT).

[0093] In one example, suspension 33 is a front suspension mounted on the front fork 10B, damping the impact applied to the front wheel 14. In another example, suspension 33 may be a rear suspension mounted on the frame 10A, damping the impact applied to the rear wheel 16. Suspension 33 includes a motor capable of being controlled to rotate or lock based on control data including damping rate, travel amount, and whether it is in a locked state. Suspension 33 may include either a valve or a solenoid valve for controlling the internal oil flow path, which can be controlled based on control data including damping rate, travel amount, and whether it is in a locked state.

[0094] The seat post 35 is mounted on the frame 10A. The seat post 35 includes a motor. The seat post 35 includes a motor to raise or lower the seat 18 relative to the frame 10A. The seat post 35 can be controlled to rotate the motor according to control data including the support position.

[0095] The braking system 37 includes a front braking device 371 configured to brake the front wheels 14 and a rear braking device 372 configured to brake the rear wheels 16. The front braking device 371 and the rear braking device 372 each include, for example, caliper brakes or disc brakes. The front braking device 371 and the rear braking device 372 include motors or the like that actuate the caliper brakes or disc brakes, and are capable of varying the braking force.

[0096] The auxiliary device 39 is a device that assists in the human-powered driving force of the human-powered vehicle 1. In one example, the auxiliary device 39 is disposed within the transmission unit. In another example, the auxiliary device 39 is disposed within the battery 50. The auxiliary device 39 includes a motor. In one example, the auxiliary device 39 is interposed between the crankshaft 21A and the frame 10A, transmitting torque to the first sprocket assembly 23 and assisting in the human-powered driving force of the human-powered vehicle 1. In another example, the auxiliary device 39 drives the chain 27 that transmits driving force to the rear wheel 16 of the human-powered vehicle 1, thereby assisting in the human-powered driving force of the human-powered vehicle 1.

[0097] An operating device 40 is provided on the handlebars 12. The operating device 40 includes an operating section 40A operated by the rider. The operating section 40A includes one or more buttons. The operating section 40A includes a brake lever. The operating section 40A can be operated by tilting the brake levers located on the left and right handlebars to the left and right. An information terminal device 7 held by the rider can be used as the operating section 40A.

[0098] The operating device 40 includes a gear indicator 40B. In one example, the gear indicator 40B is a plurality of buttons included in the operating unit 40A. In other examples, the gear indicator 40B is a device mounted on the brake lever. Whenever the rider performs an operation such as tilting the brake lever relative to the gear indicator 40B or pressing any of the plurality of buttons, manual operation of the gear transmission 31 can be performed, either by increasing or decreasing the gear ratio.

[0099] The operating device 40 includes a suspension indicator 40C. The suspension indicator 40C is, for example, a button included in the operating unit 40A. By pressing the button corresponding to the suspension indicator 40C, control data such as suspension attenuation rate and travel can be set.

[0100] The operating device 40 includes a seat post indicator 40D. The seat post indicator 40D is, for example, a button included in the operating unit 40A. By pressing the button corresponding to the seat post indicator 40D, the seat 351 can be raised and lowered.

[0101] The operating device 40 includes a brake indicator 40E, which is a brake lever. Operation of the brake lever activates either the caliper brake or the disc brake of the braking device 37.

[0102] The operating device 40 includes an auxiliary indicator 40F. The auxiliary indicator 40F is, for example, a button included in the operating unit 40A. By pressing the button corresponding to the auxiliary indicator 40F, the auxiliary mode can be set to any of the multiple levels (high / medium / low).

[0103] The operating device 40 is communicatively connected to the control device 100 in a manner that enables it to send operation-related signals to the control device 100. The operating device 40 can also be communicatively connected in a manner that enables it to directly output operation-related signals to the transmission 31, suspension 33, seat post 35, braking device 37, and auxiliary device 39. In the first example, the operating device 40 communicates with the control device 100 via a communication line or a power line capable of PLC (Power Line Communication). The operating device 40 can communicate with the transmission 31, suspension 33, seat post 35, braking device 37, auxiliary device 39, and control device 100 via a communication line or a power line capable of PLC. In the second example, the operating device 40 communicates with the control device 100 wirelessly. The operating device 40 can communicate with the transmission 31, suspension 33, seat post 35, braking device 37, auxiliary device 39, and control device 100 wirelessly.

[0104] The operating device 40 may include a notification unit that informs the rider of the operating status. The operating device 40 may inform the rider of the control status of the transmission 31, suspension 33, seat post 35, braking device 37, and auxiliary device 39 through lights, displays, speakers, etc.

[0105] The battery 50 includes a battery body 51 and a battery bracket 53. The battery body 51 is a rechargeable battery comprising one or more battery elements. The battery bracket 53 is fixed to the frame 10A of the human-powered vehicle 1. The battery body 51 is detachable from the battery bracket 53. The battery 50 is electrically connected to the device 30, the operating device 40, and the control device 100, and supplies power as needed. The battery 50 preferably includes a control unit for communicating with the control device 100. The control unit preferably includes a processor using a CPU.

[0106] The human-powered vehicle 1 is equipped with sensors 60 at various locations to acquire riding-related information, including the rider's status and the riding environment. The sensors 60 include a speed sensor 61, an acceleration sensor 62, a torque sensor 63, a cadence sensor 64, a gyroscope sensor 65, a seating sensor 66, a camera 67, and a position information sensor 68.

[0107] For example, a speed sensor 61 is located on the front wheel 14 and sends a signal to the control device 100 corresponding to the number of rotations of the front wheel 14 per unit time. The control device 100 can calculate the speed and distance traveled of the manually driven vehicle 1 based on the output of the speed sensor 61.

[0108] For example, an acceleration sensor 62 is fixed to the frame 10A. The acceleration sensor 62 is a sensor that outputs the vibration of the manually driven vehicle 1 in three axes (forward and backward, left and right, and up and down) with the frame 10A as the reference, and is set up to detect the movement and vibration of the manually driven vehicle 1. The acceleration sensor 62 sends signals corresponding to the magnitude of the movement and vibration to the control device 100.

[0109] For example, torque sensor 63 is configured to measure the torque applied to the right crank 21B and the left crank 21C respectively. Torque sensor 63 sends a signal corresponding to the torque measured in at least one of the right crank 21B and the left crank 21C to control device 100.

[0110] For example, the cadence sensor 64 is configured to measure the cadence of either the right crank 21B or the left crank 21C. The cadence sensor 64 sends a signal corresponding to the measured cadence to the control device 100.

[0111] For example, a gyroscope sensor 65 is fixed to the frame 10A. The gyroscope sensor 65 is configured to detect the yaw, roll, and pitch rotations of the manually driven vehicle 1. The gyroscope sensor 65 sends signals corresponding to the rotation amounts of each of the three axes to the control device 100. Yaw is rotation about the vertical axis. Roll is rotation about the horizontal axis. Pitch is rotation about the horizontal axis.

[0112] A seating sensor 66 is disposed on the inner surface of the saddle 351 to determine whether the rider is seated on the saddle 351. The seating sensor 66 uses, for example, a piezoelectric sensor to send a signal corresponding to the weight applied to the saddle 351 to the control device 100.

[0113] Camera 67 is mounted on the front fork 10B in a forward-facing manner. In the first example, it is mounted on the front fork 10B together with the headlights in a forward-facing manner. In the second example, it is mounted on the handlebars 12. Camera 67 uses a camera module to output an image corresponding to the user's field of vision. Camera 67 outputs an image signal showing an object present in the direction of travel.

[0114] For example, a position information sensor 68 is fixed to the frame 10A. The position information sensor 68 is configured to detect information related to the position of the manually driven vehicle 1. For example, the position information sensor 68 is configured to detect information related to the longitude and latitude of the manually driven vehicle 1 on Earth. For example, the position information sensor 68 is a GPS sensor. The position information sensor 68 sends signals corresponding to the position information of the manually driven vehicle 1 to the control device 100.

[0115] Sensor 60 may include, but is not limited to, a speed sensor 61, an acceleration sensor 62, a torque sensor 63, a cadence sensor 64, a gyroscope sensor 65, a seating sensor 66, a camera 67, and a position information sensor 68.

[0116] Figure 2 This is a block diagram illustrating the structure of the control device 100. The control device 100 includes a control unit 110 and a storage unit 112.

[0117] The control unit 110 is a processor that uses a CPU. The control unit 110 uses built-in memory such as ROM (Read Only Memory) and RAM (Random Access Memory). The control unit 110 performs processing in a functionally differentiated manner through the device control unit 116 and the supplementary processing unit 118.

[0118] The device control unit 116 acquires input information related to the movement of the manually driven vehicle from the sensor 60. Following the device control program 10P, the device control unit 116 controls the device 30 using control data determined based on output information obtained by inputting the acquired input information into the first learning model 11M (described later). Based on the determined control data and following the device control program 10P, the device control unit 116 controls the movement of the controlled object mounted on the manually driven vehicle 1, supplies power to the controlled object, and communicates with the controlled object.

[0119] The supplementary processing unit 118 performs a supplementary processing procedure 12P to supplement the first learning model 11M stored in the storage unit 112 using a second learning model. The second learning model is learned using input information from at least one of the human-powered vehicle 1 and the rider of the human-powered vehicle 1.

[0120] The control unit 110 performs processing in a control state different from the mode in which the device 30 is automatically controlled using the device control unit 116 via the first learning model 11M (described later) and the learning mode of the first learning model 11M. The control unit 110 essentially performs processing based on the learning mode of the operation of the operating device 40 until the accuracy of the first learning model 11M reaches a certain level. If the accuracy of the first learning model 11M reaches a certain level, the control unit 110 essentially performs processing using the automatic control mode of the first learning model 11M while simultaneously supplementing the first learning model 11M via the supplementary processing unit 118.

[0121] The storage unit 112 includes non-volatile memory such as flash memory. The storage unit 112 stores a device control program 10P and a supplementary processing program 12P. The device control program 10P and the supplementary processing program 12P can each be a program in which the control unit 110 reads the device control program 90P and the supplementary processing program 92P stored in the non-temporary storage medium 900 and copies them into the storage unit 112.

[0122] Storage unit 112 stores a first learning model 11M. Details about the first learning model 11M will be described later. The first learning model 11M may also be a model that the control unit 110 reads from the first learning model 91M stored in the non-temporary storage medium 900 and copies to the storage unit 112.

[0123] The control unit 110 (device control unit 116 and supplementary processing unit 118) communicates with the controlled object. In this case, the control unit 110 itself may have a communication unit (not shown) oriented towards the controlled object, and the control unit 110 may be connected to a communication unit oriented towards the controlled object provided inside the control device 100. Preferably, the control unit 110 has a connection unit for communicating with the controlled object or the communication unit.

[0124] The control unit 110 preferably communicates with the controlled object via at least one of PLC and CAN communication. The communication between the control unit 110 and the controlled object is not limited to wired communication, but can also be wireless communication such as ANT, ANT+, Bluetooth, WiFi, ZigBee, etc.

[0125] The control unit 110 is connected to the sensor 60 via a signal line. The control unit 110 obtains input information related to the driving of the manually driven vehicle 1 from the signal output by the sensor 60 via the signal line.

[0126] The control unit 110 and the information processing device 8 (described later) can communicate via a wireless communication device 114 with an antenna. The wireless communication device 114 can be built into the control device 100. The wireless communication device 114 is a device capable of communication via the so-called Internet. The wireless communication device 114 can be a wireless communication device such as ANT, ANT+, Bluetooth, WiFi, ZigBee, or LTE (Long Term Evolution). The wireless communication device 114 can be based on communication networks such as 3G, 4G, 5G, LTE (Long Term Evolution), WAN (Wide Area Network), LAN (Local Area Network), Internet lines, dedicated lines, and satellite lines.

[0127] The control functions of the control device 100 configured in this way will be explained. The control unit 110 of the control device 100, through the function of the device control unit 116, automatically controls the device 30 according to the device control program 10P, using control data determined based on output information obtained by inputting acquired input information into the first learning model 11M (described later). In the first embodiment, the control unit 110 automatically controls the transmission device 31 based on information obtained by inputting input information into the first learning model 11M through the device control unit 116.

[0128] Figure 3This is a diagram illustrating an example of the first learning model 11M. The first learning model 11M is a learning model that learns using deep learning trained with a neural network (hereinafter referred to as NN). The first learning model 11M can be a model that learns using a recurrent neural network. Figure 3 The first learning model 11M shown learns to reproduce the gear ratio indicated by the transmission device 31 when input information related to the driving of the human-powered vehicle 1 is obtained by the sensor 60.

[0129] The first learning model 11M includes an input layer 111 that receives input information related to the driving of the manually driven vehicle 1 obtained by the sensor 60. The first learning model 11M also includes an output layer 115 that outputs output information related to the control of the transmission device 31 of the manually driven vehicle 1. The first learning model 11M includes an intermediate layer 113, which comprises one or more node groups located between the input layer 111 and the output layer 115. Learning is performed on the output information based on learning data including the operation content received by the operating device 40. The nodes of the intermediate layer 113 each have parameters including at least one of weight and bias in their relationship with nodes in the preceding layers. The control unit 110, based on a portion of the learning function of the supplementary processing unit 118, marks the actual indicated gear ratio of the transmission device 31 in the corresponding input information, thereby creating learning data. The control unit 110 inputs the created learning data to the input layer 111 and learns the parameters in the intermediate layer 113 to reduce the error between the gear ratio output from the output layer 115 and the actual gear ratio indicated by the rider. Thus, a first learning model 11M is learned, which, based on the input information obtained from the sensor 60, reproduces the gear ratio indicated by the rider on the gearbox 31 according to scenarios such as the speed, acceleration, and road type of the human-powered vehicle 1.

[0130] Figure 4 This is a diagram illustrating another example of the first learning model 11M. Figure 4 The first learning model 11M shown is... Figure 3 The example shown is the same learning model learned through deep learning trained using a neural network. The first learning model 11M learns from input information related to the driving of the human-powered vehicle 1 obtained by the sensor 60, and outputs the probability of instructing the transmission device 31.

[0131] The first learning model 11M includes an input layer 111, an intermediate layer 113, and an output layer 115. Another example of the first learning model 11M is to create learning data by marking whether a change in gear ratio has been indicated to the gearshift device 31 in its input information. The control unit 110 inputs the created learning data to the input layer 111 and learns the parameters in the intermediate layer 113 to reduce the error between the probability output from the output layer 115 and the result of whether the rider actually made the indication. Thus, the first learning model 11M is learned to output the probability that the rider will indicate the gearshift device 31 based on scenarios such as the speed, acceleration, and road type of the human-powered vehicle 1, according to the input information obtained from the sensor 60.

[0132] In the manually driven vehicle 1, the device control unit 116 uses output information from the first learning model 11M to control the transmission device 31. Figure 3 In the case of the first learning model 11M shown, the control unit 110 can perform speed change via the speed change device 31 based on the speed ratio output from the first learning model 11M. When using... Figure 4 In the case of the first learning model 11M shown, the control unit 110 can change the gear ratio of the transmission device 31 based on the probability output from the first learning model 11M.

[0133] The first learning model 11M, in the stage before the human-powered vehicle 1 leaves the factory, uses general learning data to learn from the input information to output, such as... Figure 3 or Figure 4 The output information is as shown. In the manufactured manual-drive vehicle 1, the control device 100 operates in learning mode, and the control unit 110 advances the learning of the first learning model 11M by combining the rider's preferences and characteristics. However, the control data for the transmission 31 to be suitable for the rider's learning in an environment where the manual-drive vehicle 1 is not moving is still insufficient. Therefore, in the first embodiment, the control device 100 supplements the first learning model 11M by using the supplementary processing unit 118 to learn a second learning model that is different from the input information of the manual-drive vehicle 1 and at least one of the riders of the manual-drive vehicle 1 from other manual-drive vehicles.

[0134] The second learning model used for supplementation is a learning model that has learned from other human-powered vehicles 1 or other riders. It is a model collected from the control device 100 corresponding to each human-powered vehicle 1 into the information processing device 8.

[0135] Figure 5 This diagram illustrates the control device 100 and the information processing device 8 according to the first embodiment. Figure 5As shown, the control device 100 and the information processing device 8 of the first embodiment can communicate with each other via a communication network N. The communication network N is composed of communication lines such as 3G, 4G, 5G, LTE, WAN, LAN, Internet lines, dedicated lines, satellite lines, or communication equipment such as base stations. The control device 100 can use an information terminal device 7 configured to communicate with the information processing device 8 via the communication network N. The information terminal device 7 is, for example, a smartphone or cycling computer used by the rider of the human-powered vehicle 1, and can also function as a user interface for inputting the rider's instructions or outputting information to the rider.

[0136] The information processing device 8 has a storage unit 802 that stores a plurality of second learning models 82M. The control device 100 can utilize any one of the plurality of second learning models 82M through communication with the information processing device 8.

[0137] Figure 6 This is a block diagram illustrating the structure of the information processing device 8. The information processing device 8 includes a control unit 800, a storage unit 802, and a communication unit 804.

[0138] The control unit 800 is a processor that uses a CPU. The control unit 800 can also use a GPU (Graphics Processing Unit). The control unit 800 can also use both a CPU and a GPU. The control unit 800 uses built-in memory such as ROM and RAM to receive and send data with the control device 100.

[0139] The control unit 800 can be one or more processing circuits equipped with FPGA (Field Programmable Gate Array), DSP (Digital Signal Processor), quantum processor, volatile or non-volatile memory, etc.

[0140] Storage unit 802 is a high-capacity non-volatile memory such as a hard disk or SSD (Solid State Drive). Storage unit 802 stores the multiple second learning models 82M collected from multiple human-powered vehicles 1 and the model recognition data that recognizes the second learning models 82M respectively.

[0141] The storage unit 802 stores the model recognition data of the second learning model 82M in correspondence with the learned data for recognizing at least one of the human-powered vehicle 1 and the rider of the human-powered vehicle 1. The storage unit 802 includes a model database 822 storing the specifications of the human-powered vehicle 1 and the type of the rider, in a manner corresponding to the data used for recognizing at least one of the human-powered vehicle 1 and the rider. Therefore, the control unit 800 can determine which type of input information the second learning model 82M has learned.

[0142] The communication unit 804 is a communication device that communicates with the control device 100 via network N. The communication unit 804 is based on communication networks such as 3G, 4G, 5G, LTE, WAN, LAN, Internet lines, dedicated lines, and satellite lines. The control unit 800 sends and receives data with the control device 100 via the communication unit 804.

[0143] With the control device 100 and information processing device 8 configured in this way, the control device 100 uses a first learning model 11M supplemented by a second learning model 82M to control the speed change device 31.

[0144] Figure 7 This is a flowchart illustrating an example of the learning method of the first learning model 11M in the first embodiment. The control device 100 performs the following processing in learning mode.

[0145] The control unit 110 of the control device 100 acquires input information related to the driving of the manually driven vehicle 1 from the sensor 60 and stores it (step S101). The control unit 110 executes the processing of step S101 at multiple time points during driving.

[0146] In step S101, the control unit 110 acquires data from at least one of the speed sensor 61, acceleration sensor 62, torque sensor 63, cadence sensor 64, gyroscope sensor 65, seating sensor 66, and camera 67.

[0147] The control unit 110 inputs the acquired input information related to the driving of the human-powered vehicle 1 to the first learning model 11M (step S103) and obtains the output information output from the first learning model 11M (step S105).

[0148] The control unit 110 determines whether it is an unlearned scenario by comparing the output information obtained in step S105 with the actual operation performed by the rider on the gear indicator device 40B corresponding to the input information obtained in step S101 (step S107).

[0149] If the scenario is determined to be unlearned ("Yes" in S107), the control unit 110 sends a request for the second learning model 82M to the information processing device 8 via the wireless communication device 114 (step S109).

[0150] If the information processing device 8 receives a request from the second learning model 82M (step S801), the control unit 800 determines the identification data of the human-powered vehicle 1 in the control device 100, which is the source of the request (step S803).

[0151] The control unit 800 extracts from the multiple second learning models 82M stored in the information processing device 8 the second learning model 82M corresponding to at least one of the human-powered vehicle 1 and the rider in the control device 100 which is the request source (step S805).

[0152] In step S805, as a first example, the control unit 800 extracts from multiple second learning models 82M a second learning model that has been learned from other human-powered vehicles 1 that are the same as or similar to at least one of the types and sizes of the human-powered vehicle 1 in the control device 100 that is the request source. For example, if the human-powered vehicle 1 equipped with the control device 100 that is the request source is a mountain bike, the control unit 800 refers to the model database 822 and extracts the second learning model 82 corresponding to the identification data of other human-powered vehicles 1 that are mountain bikes.

[0153] In step S805, as a second example, the control unit 800 extracts from multiple second learning models 82M the second learning model 82 learned from other human-powered vehicles 1 equipped with devices 30 of the same or similar category as those on the human-powered vehicle 1 equipped with the control device 100 as the request source, and devices 30 of the manufacturer. Device 30 is not limited to the transmission device 31 as the control object, but may also be at least one of the suspension 33, seat post 35, braking device 37, and auxiliary device 30. The control unit 800 refers to the model database 822 to extract the second learning model 82M learned from human-powered vehicles 1 equipped with devices that are the same or similar to at least one of the transmission device 31, suspension 33, seat post 35, braking device 37, and auxiliary device 39.

[0154] In step S805, as a third example, the control unit 800 extracts from multiple second learning models 82M a second learning model 82M that has been learned from human-powered vehicles 1 ridden by riders of the same or similar type as the rider of the human-powered vehicle 1 in the control device 100, which is the request source. Referring to the model database 822, the control unit 800 extracts the second learning model 82M that has been learned from human-powered vehicles 1 ridden by riders of similar types, based on the identification data of riders classified as high-frequency high-torque, high-frequency low-torque, low-frequency high-torque, and low-frequency low-torque types.

[0155] The control unit 800 can extract multiple second learning models 82M by one of the methods in the first to third examples, or by combining two or three methods to extract multiple second learning models 82M.

[0156] In step S805, the control unit 800 can sort the extracted multiple second learning models 82M according to their similarity and narrow them down to a predetermined number of second learning models 82M. For example, similarity can be calculated in a way that the more identical items there are, the higher the similarity.

[0157] The control unit 800 sends the extracted complements of multiple second learning models 82M from the communication unit 804 to the control device 100 (step S807).

[0158] The control unit 110 receives from the information processing device 8 multiple complements of a second learning model 82M that has been learned using input information from at least one of the human-powered vehicles 1 and the rider (step S111). The control unit 110 selects the second learning model 82M from the multiple complements (step S113).

[0159] In step S113, in the first example, the control unit 110 selects from multiple alternatives a second learning model 82 whose output information when the input information obtained in step S101 is input is closest to the actual operation performed by the rider on the gear shift indicator 40B.

[0160] In step S113, in the second example, the control unit 110 selects a second learning model 82M from multiple alternatives. This second learning model 82M is a model similar to the control data determined by the device control unit 116 based on the output information obtained when the same input information is input. More specifically, the control unit 110 inputs the input information obtained in step S101 to each of the alternatives of the second learning model 82M. The control unit 110 determines the control data most similar to the actual operation performed by the rider on the gear indicator device 40B from the control data determined by the device control unit 116 based on the output information output from each of the alternatives of the second learning model 82M. The control unit 110 selects the second learning model 82M that outputs the output information as the source of the determined control data.

[0161] In step S113, in the third example, the control unit 110 can arbitrarily select one from the multiple received backups.

[0162] The control unit 110 performs supplementary processing (step S115) by updating at least a portion of the first learning model 11M using the second learning model 82M selected in step S113. In step S115, the control unit 110 can replace the entire first learning model 11M with the selected second learning model 82M. In step S115, the control unit 110 can update a portion of the parameters of the first learning model 11M using the parameters of the selected second learning model 82M.

[0163] The control unit 110 uses the learning data obtained by marking the actual operation of the rider on the gear indicator device 40B in the input information obtained in step S101 to relearn the supplemented first learning model 11M (step S117), and then ends the processing.

[0164] If the control unit 110 determines in step S107 that it is not a scenario that has not been learned (the value is "No" in S107), the control unit 110 sends the first learning model 11M that has been learned from the wireless communication device 114 to the information processing device 8 (step S119).

[0165] The information processing device 8 receives the first learning model 11M after learning (step S809), stores it in the storage unit 802 as the second learning model 82M (step S811), and then ends the processing.

[0166] Figure 7 The processing sequence shown in the flowchart is not limited to the driving of the manually driven vehicle 1, but can also be executed at the end of each journey based on the input information stored in step S101. Figure 7 The flowchart shows the processing sequence, where the first learning model 11M, which is determined to be unlearned, is supplemented by the second learning model 82M. Therefore, the likelihood of a scenario being determined to be unlearned in step S107 is reduced, and even in unlearned scenarios, the first learning model 11M can be used for automatic control.

[0167] Figure 8 This is a flowchart illustrating an example of control processing using the first learning model 11M of the first embodiment. In control mode, the control device 100 repeatedly executes the following processing.

[0168] The control unit 110 of the control device 100 acquires input information related to the driving of the manually driven vehicle 1 from the sensor 60 (step S201).

[0169] The control unit 110 inputs the acquired input information related to the driving of the human-powered vehicle 1 to the first learning model 11M (step S203), and then acquires the output information output from the first learning model 11M (step S205).

[0170] Based on the output information obtained in step S205, the control unit 110 determines the control data of the transmission device 31 through the device control unit 116 (step S207). In step S207, the control unit 110 can determine the transmission ratio itself or whether to change speed.

[0171] The control unit 110 uses the determined control data to control the transmission device 31 (step S209), and then ends the process. The control unit 110 repeatedly executes the processes of steps S201 to S209.

[0172] Through the processing of the control device 100 in the first embodiment, even for driving scenarios that have not been learned, the second learning model 82M can be used to promote the learning of the first learning model 11M, shortening the learning time required. Through the processing of the control device 100, even for driving scenarios that have not been learned, the second learning model 82M can be used to supplement the first learning model 11M, enabling automatic control even in scenarios where no learning has been performed.

[0173] (Second Implementation)

[0174] In the second embodiment, the control objects include the suspension 33, seat post 35, braking device 37, and auxiliary device 39, excluding the transmission device 31. In the second embodiment, the control object based on the first learning model 11M only needs to be at least one of the transmission device 31, suspension 33, seat post 35, braking device 37, and auxiliary device 39.

[0175] In the second embodiment, a first learning model 11M is additionally learned for each controlled object. Figure 9 This is a block diagram illustrating the structure of the control device 100 according to the second embodiment. Figure 9 The diagram only illustrates the structure of the control device 100, and omits the illustrations of the device 30, the operating device 40, and the sensor 60 connected to the control device 100.

[0176] The storage unit 112 of the control device 100 stores a first learning model 11M for controlling the transmission device 31 and a first learning model 13M for controlling the suspension 33. The storage unit 112 also stores a first learning model 15M for controlling the seat post 35, a first learning model 17M for controlling the braking device 37, and a first learning model 19M for controlling the auxiliary device 39.

[0177] The first learning model 11M, the first learning model 13M, the first learning model 15M, the first learning model 17M, and the first learning model 19M are models that can be copied and stored on the non-temporary storage medium 900, namely the first learning model 91M, the first learning model 93M, the first learning model 95M, the first learning model 97M, and the first learning model 99M.

[0178] The control unit 110 of the control device 100 in the second embodiment performs the operations on each device 30 that were performed on the transmission device 31 in the first embodiment. Figure 8The processing sequence is shown in the flowcharts for both the first and ninth steps.

[0179] Therefore, the control device 100 uses at least one of the suspension 33, seat post 35, braking device 37, and auxiliary device 39 as the controlled object. For driving scenarios that have not been learned, the second learning model 82M is also used to promote the learning of the first learning models 11M, 13M, 15M, 17M, and 19M. The learning time required for controlling any device 30 can be shortened.

[0180] (Third Implementation)

[0181] In the third embodiment, the learning method differs when the condition is determined to be unlearned. Except for the details of the processing sequence of the learning method described below, the structure of the control device 100 in the third embodiment is the same as that of the control device 100 in the first embodiment. The same reference numerals are used for structures in the control device 100 of the third embodiment that are common to those in the control device 100 of the first embodiment, and their detailed descriptions are omitted.

[0182] Figure 10 This is a flowchart illustrating an example of the learning method of the first learning model 11M in the third embodiment. The control unit 110 executes the method shown in the first embodiment. Figure 7 The flowchart processing sequence is as follows: multiple second learning models 82M are received from the information processing device 8 (S111), and any one of them is selected (step S113) to supplement the first learning model 11M as shown below.

[0183] The control unit 110 stores the selected second learning model 82M in the storage unit 112 (step S501). The control unit 110 inputs the input information acquired and stored in step S101 to the selected second learning model 82M (step S503). The control unit 110 acquires the output information output from the second learning model 82M (step S505).

[0184] The control unit 110 stores the input information and output information as learning data in correspondence (step S507). The control unit 110 uses the stored learning data to learn the first learning model 11M (step S509), and then ends the learning process.

[0185] In the third embodiment, such as Figure 10 As shown, the control device 100 uses the input information acquired by the control unit 110 and the output information output when the input information is input to the second learning model 82M as learning data to learn the first learning model 11M.

[0186] Even in scenarios where the control device 100 of the manually driven vehicle 1 has insufficient learning, it can still use the already learned second learning model 82M for control, and can use the second learning model 82M to create learning data to learn the first learning model 11M. This can shorten the learning time for unlearned scenarios.

[0187] (Fourth Implementation)

[0188] The control device 100 of the fourth embodiment is identical to the control device 100 of the first embodiment, except for the structure of the first learning model 11M and the details of the processing order using the first learning model 11M. For structures in the control device 100 of the fourth embodiment that are common to the control device 100 of the first embodiment, the same reference numerals are used, and detailed descriptions are omitted.

[0189] Figure 11 This is a schematic diagram of the first learning model 11M in the fourth embodiment. In the fourth embodiment, the first learning model 11M is also a learning model that is learned by using deep learning trained with a neural network. The first learning model 11M in the fourth embodiment learns using input information acquired in multiple different driving scenarios. Driving scenarios are divided into at least one of highway, off-road, and urban. Driving scenarios are divided into at least one of uphill, flat, and downhill. In the following description, driving scenarios are divided into nine types: uphill on highways, flat roads on highways, downhill on highways, uphill on off-road, flat roads on off-road, downhill on off-road, uphill in urban areas, flat roads in urban areas, and downhill in urban areas. Driving scenarios are not limited to nine types. Driving scenarios are not limited to the above division. Driving scenarios can be divided by driving scenarios related to acceleration and deceleration, such as starting, accelerating, decelerating, and stopping. Driving scenarios can be categorized based on road shape, such as when going straight, entering a curve, exiting a curve, entering an intersection, and entering a road with decreasing width. Driving scenarios can also be categorized based on vehicle type, such as when there are no other vehicles around, when a vehicle is approaching from behind, when traveling with other human-powered vehicles, and when attempting to overtake other human-powered vehicles.

[0190] The first learning model 11M includes an input layer 111 that receives input information related to the driving of the manually driven vehicle 1 obtained by the sensor 60 and data representing the driving scene. Apart from this, the structure is the same as the first learning model 11M described in the first embodiment. In the following description, the first learning model 11M learns to reproduce the gear ratio indicated by the transmission device 31, which is one of the devices 30, when it receives input information related to the driving of the manually driven vehicle 1 obtained by the sensor 60 and data representing the driving scene.

[0191] In the fourth embodiment, the control unit 110, based on a portion of the learning function of the supplementary processing unit 118, creates learning data by marking the actual gear ratio of the indicated gear shifter 31 in the corresponding input information. The control unit 110 inputs the created learning data and a driving scenario determined based on information obtained from the sensor 60 to the input layer 111, learning the parameters in the intermediate layer 113 to reduce the error between the gear ratio output from the output layer 115 and the actual gear ratio indicated by the rider. Thus, based on the input information obtained from the sensor 60, a first learning model 11M is learned to reproduce the gear ratio indicated by the rider to the gear shifter 31 based on the driving scenario of the manually driven vehicle 1 and the speed, acceleration, etc., in that driving scenario.

[0192] Figure 11 The first learning model 11M outputs numerical values ​​of the control data for the gear shifter 31 as output information related to the control of the device 30. In the fourth embodiment, the output layer 115 of the first learning model 11M can also output the probability of instructing the gear shifter 31. In this case, the intermediate layer 113 is learned to output the probability that the rider instructs the gear shifter 31 to shift gears based on the riding scenario.

[0193] In the fourth embodiment, the first learning model 11M learned in each control device 100 is also collected and stored by the information processing device 8.

[0194] Figure 12 as well as Figure 13 This is a flowchart illustrating an example of the learning method of the first learning model 11M in the fourth embodiment. The control device 100 performs the following processing in learning mode.

[0195] The control unit 110 of the control device 100 acquires and stores input information related to the driving of the manually driven vehicle 1 from the sensor 60 (step S131). In step S131, the control unit 110 acquires data from at least one of the speed sensor 61, acceleration sensor 62, torque sensor 63, cadence sensor 64, gyroscope sensor 65, seating sensor 66, and camera 67.

[0196] The control unit 110 determines the driving scenario based on the input information obtained from the sensor 60 (step S133).

[0197] In step S133, the control unit 110 determines the driving scenario based on information obtained from at least one of the speed sensor 61, acceleration sensor 62, torque sensor 63, cadence sensor 64, gyroscope sensor 65, seating sensor 66, camera 67, and position information sensor 68. In the first example, the control unit 110 determines whether the road the human-powered vehicle 1 is traveling on is off-road, urban, or highway based on information related to the position of the human-powered vehicle 1 obtained from the position information sensor 68 and pre-existing map information of the human-powered vehicle 1. In the first example, the control unit 110 may be configured to acquire map information from an external source via communication. In the second example, the control unit 110 determines whether the road the human-powered vehicle 1 is traveling on is off-road, urban, or highway based on vibration information relative to the human-powered vehicle 1 obtained from the acceleration sensor 62 and posture information of the human-powered vehicle 1 obtained from the gyroscope sensor 65. In the second example, if the frequency of vibration exceeding a predetermined value is higher than a predetermined frequency, the control unit 110 can determine that the driving scenario is off-road. In the third example, when the frequency of driving in a non-seated state exceeds a predetermined frequency, the control unit 110 can determine that the driving scenario is off-road based on the seating sensor 66. In the fourth example, when the number of stops and starts relative to the travel distance exceeds a predetermined number, the control unit 110 can determine that the driving scenario is urban. In the fifth example, when the torque and cadence are constant, the control unit 110 can determine that the driving scenario is highway. In the sixth example, when neither off-road nor urban driving occurs, the control unit 110 can determine that the driving scenario is highway.

[0198] In step S133, the control unit 110 can determine uphill, downhill, and flat terrain based on the tilt of the pitch direction of the manually driven vehicle 1 using the gyroscope sensor 65.

[0199] The control unit 110 determines whether the driving scenario identified in step S133 is an unlearned driving scenario (step S135). In step S135, the control unit 110 determines whether it is an unlearned driving scenario based on information stored in the storage unit 112 corresponding to the data used to identify the driving scenario, indicating whether the learning has been completed.

[0200] If the driving scenario is determined to be unlearned ("Yes" in S135), the control unit 110 sends a request for the second learning model 82M to the information processing device 8 via the wireless communication device 114 (step S137). In step S137, the control unit 110 specifies the data used to identify the determined driving scenario and sends the request.

[0201] If the information processing device 8 receives a request from the second learning model 82M (step S821), the control unit 800 determines the identification data of the human-powered vehicle 1 of the control device 100, which is the source of the request (step S823).

[0202] The control unit 800 extracts a second learning model 82M that has learned the driving scenario specified in the request from a plurality of second learning models 82 stored in the information processing device 8 (step S825). From the extracted second learning models 82M, the control unit 800 selects a second learning model 82M corresponding to at least one of the human-powered vehicle 1 of the control device 100 as the request source and the rider as a backup (step S827).

[0203] In step S827, the control unit 800 can use the first embodiment Figure 7 The method described in step S805 of the processing sequence shown in the flowchart is used to select at least one of the first to third examples.

[0204] The control unit 800 sends the complement of the selected second learning model 82M from the communication unit 804 to the control device 100 (step S829).

[0205] The control unit 110 receives from the information processing device 8 multiple complements of a second learning model 82M that has been learned using input information from at least one of the human-powered vehicles 1 and the rider (step S139). The control unit 110 selects the second learning model 82M from the multiple complements (step S141).

[0206] In step S141, in the first example, the control unit 110 selects from multiple alternatives a second learning model 82M whose output information when the input information obtained in step S101 is input is closest to the actual operation performed by the rider on the gear shift indicator 40B.

[0207] In step S141, in the second example, the control unit 110 selects from multiple alternatives the output information output when the first learning model 11M has acquired and stored input information in other driving scenarios that it has already learned, and a second learning model 82M that is similar to the output information output when the input information is input to the first learning model 11M.

[0208] In step S141, in the third example, the control unit 110 selects from multiple backups a second learning model 82M that is similar to the control data determined by the device control unit 116 based on the output information obtained when the same input information is input. More specifically, the control unit 110 inputs the input information obtained in step S131 into the backups of the second learning model 82M. The control unit 110 determines from the control data determined by the device control unit 116 based on the output information output from the backups of the second learning model 82M the control data that is most similar to the actual operation performed by the rider on the gear indicator device 40B. The control unit 110 selects the second learning model 82M that outputs the output information as the source of the determined control data.

[0209] In step S141, in the fourth example, the control unit 110 selects a second learning model 82M from among multiple complements. This second model is similar to the control data determined by the device control unit 116 based on the output information obtained when input information acquired in other learned driving scenarios is input. More specifically, the control unit 110 inputs the input information acquired and stored in other learned driving scenarios, and the data used to identify the driving scenario, into the complement of the second learning model 82M. The control unit 110 inputs the stored input information and the data used to identify the driving scenario into the first learning model 11M that has already been learned. For the learned driving scenario, the control data determined by the device control unit 116 based on the output information obtained from the first learning model 11M is acquired. This control data is not used for control. The control unit 110 acquires the control data determined by the device control unit 116 based on the output information output from the complements of the second learning model 82M. The control unit 110 selects from the supplement of the second learning model 82 to output the second learning model 82M, which is similar to the control data determined based on the output information from the first learning model 11M, as the source of the control data.

[0210] In step S141, in the fifth example, the control unit 110 can arbitrarily select one from the multiple received backups.

[0211] The control unit 110 updates at least a portion of the first learning model 11M using the second learning model 82M selected in step S141, thereby performing supplementary processing (step S143). In step S143, the control unit 110 uses the output information obtained from the second learning model 82 as learning data for learning, following the order shown in the third embodiment. In step S143, the control unit 110 can replace the entire first learning model 11M with the selected second learning model 82M. In step S143, the control unit 110 can also update some parameters of the first learning model 11M using the parameters of the selected second learning model 82M.

[0212] The control unit 110 relearns the supplemented first learning model 11M using learning data obtained from the input information acquired in step S131, which marks the actual operations performed by the rider on the gear shift indicator 40B. (Step S145) The control unit 110 stores the learned information in the storage unit 112 in a manner corresponding to the data used to identify the riding scenario determined in step S133. (Step S147)

[0213] In step S135, if the control unit 110 determines that it is not a driving scenario that has not been learned (in S135, it is "No"), the control unit 110 sends the first learning model 11M that has been learned and the data used to identify the driving scenario from the wireless communication device 114 to the information processing device 8 (step S149).

[0214] The information processing device 8 receives the first learning model 11M after learning (step S831), and stores it as the second learning model 82M together with the data used to identify the driving scenario in the storage unit 802 (step S833), and then ends the processing.

[0215] Figure 12 as well as Figure 13 The processing sequence shown in the flowchart is not limited to the driving of the manually driven vehicle 1, but can also be executed at the end of each journey based on the input information stored in step S131. Figure 12 as well as Figure 13 As shown in the flowchart, the first learning model 11M uses the second learning model 82M to supplement the driving scenarios that have not been learned. Therefore, the probability of a scenario being identified as unlearned in step S107 is reduced, and the first learning model 11M can still be used for automatic control even in unlearned scenarios.

[0216] Figure 14 This is a flowchart illustrating an example of control processing using the first learning model 11M of the fourth embodiment. The control device 100 repeatedly executes the following processing in control mode.

[0217] The control unit 110 of the control device 100 acquires input information related to the driving of the manually driven vehicle 1 from the sensor 60 (step S221). Based on the input information acquired from the sensor 60, the control unit 110 determines the driving scenario (step S223).

[0218] The control unit 110 inputs the acquired input information related to the driving of the human-powered vehicle 1 and the data used to identify the determined driving scenario to the first learning model 11M (step S225). The control unit 110 acquires the output information output from the first learning model 11M (step S227).

[0219] Based on the output information obtained in step S227, the control unit 110 determines the control data of the transmission device 31 through the device control unit 116 (step S229). In step S229, the control unit 110 can determine the transmission ratio itself or whether to change speed.

[0220] The control unit 110 controls the transmission device 31 using the determined control data (step S231), and then ends the process. The control unit 110 repeatedly executes the processes of steps S221 to S231.

[0221] After performing the control in step S231, if there is an intervention operation by the rider through the operating device 40, the control unit 110 can re-store the riding scenario as unlearned and then execute it. Figure 12 as well as Figure 13 The processing sequence is shown in the flowchart. In this case, the control unit 110 performs relearning on the first learning model 11M, which has been supplemented using the second learning model 82M, using input information obtained from the sensor 60 of the human-powered vehicle 1, data for recognizing the driving scene, and learning data from the rider's operation.

[0222] Through the processing of the control device 100 in the fourth embodiment, the learning time of the first learning model 11M that obtains suitable output information related to control data can be shortened depending on whether it is a driving scenario that has not been learned.

[0223] (Fifth Implementation)

[0224] The control device 100 of the fifth embodiment is identical to the control device 100 of the first embodiment in structure, except for the structure of the first learning model 11M and the details of the processing order using the first learning model 11M. For structures in the control device 100 of the fifth embodiment that are common to those in the control device 100 of the first embodiment, the same reference numerals are used, and detailed descriptions are omitted.

[0225] Figure 15 This diagram illustrates the first learning model 11M of the fifth embodiment. In the fifth embodiment, the first learning model 11M includes multiple learning models 11MA, 11MB, ... stored according to various driving scenarios. Each learning model 11MA, 11MB, 11MC, ... includes the models from the first embodiment. Figure 3 or Figure 4 The learning model shown comprises the input layer, output layer, and intermediate layers. Detailed explanations of the input layer, output layer, and intermediate layers are omitted.

[0226] Each learning model 11MA, 11MB, 11MC... learns according to different driving scenarios, so as to output control data for device 30 when given input information related to the driving of the manually driven vehicle 1 obtained through sensor 60. For example... Figure 14 As shown, in the fifth embodiment, the driving scenario is divided into at least one of highway, off-road, and urban, and the driving scenario is further divided into at least one of uphill, flat, and downhill. In the following description, the driving scenario is divided into a total of nine driving scenarios: uphill on a highway, flat road on a highway, downhill on a highway, uphill on an off-road vehicle, flat road on an off-road vehicle, downhill on an off-road vehicle, uphill in an urban area, flat road in an urban area, and downhill in an urban area.

[0227] For example, learning model 11MA learns from input information of an uphill driving scenario on a road, acquired by the human-powered vehicle 1 from sensor 60, and from the rider's actions, as learning data. Learning model 11MB learns from input information of a flat road driving scenario, acquired by the human-powered vehicle 1 from sensor 60, and from the rider's actions, as learning data. Similarly, learning model 11MC learns from input information of a downhill driving scenario, acquired by the human-powered vehicle 1 from sensor 60, and from the rider's actions, as learning data. Learning models 11MD, 11ME, 11MF, 11MG, 11MH, and 11MI learn from uphill driving in off-road terrain, flat road driving in off-road terrain, downhill driving in off-road terrain, uphill driving in urban areas, flat road driving in urban areas, and downhill driving in urban areas, respectively.

[0228] As described above, the processing of the control device 100, which uses a first learning model 11M that has been learned according to various driving scenarios, will be explained below.

[0229] Figure 16 as well as Figure 17 This is a flowchart illustrating an example of the learning method of the first learning model 11M in the fifth embodiment. Figure 16 as well as Figure 17 The processing sequence shown in the flowchart is consistent with that of the fourth embodiment. Figure 12 as well as Figure 13 The flowcharts show a common processing sequence, with the same step numbers appended and detailed descriptions omitted.

[0230] The control unit 110 acquires and stores the input information (S131) ​​and determines the driving scenario (S133).

[0231] The control unit 110 determines whether the first learning model 11M that has been learned in the storage unit 112 includes the learning models 11MA, 11MB, ..., which have been learned in step S133 (step S151).

[0232] If it is determined that the first learning model 11M includes learning models 11MA, 11MB... that have already learned the determined driving scenario ("Yes" in S151), the control unit 110 executes the processing in step S153. The control unit 110 sends the data used to identify the determined scenario and the learned models 11MA, 11MB... that have already been learned for the determined driving scenario from the wireless communication device 114 to the information processing device 8 (step S153).

[0233] The information processing device 8 receives data for identifying the determined scene and the learned first learning model 11M (step S851). The control unit 800 stores the received first learning model 11M as a second learning model 82M together with the data for identifying the driving scene in the storage unit 802 (step S853), and then ends the processing. The storage unit 802 stores the first learning model 11M, which includes multiple learning models, as the second learning model 82M.

[0234] If it is determined that the first learning model 11M does not include the learning models 11MA, 11MB, etc., that have already been learned for the determined driving scenario (in S151, this is "No"), the control unit 110 sends a request for the second learning model 82M to the information processing device 8 via the wireless communication device 114 (step S155). In step S155, the control unit 110 specifies the data used to identify the determined driving scenario and sends the request.

[0235] If the information processing device 8 receives a request from the second learning model 82M (step S855), the control unit 800 determines the identification data of the human-powered vehicle 1 of the control device 100, which is the source of the request (step S857).

[0236] The control unit 800 extracts a second learning model 82M from a plurality of second learning models 82 stored in the information processing device 8, including a learning model that has already learned the driving scenario specified by the request (step S859). From the extracted second learning models 82M, the control unit 800 selects a second learning model 82M corresponding to at least one of the human-powered vehicle 1 of the control device 100 that is the request source and the rider, as a backup (step S861).

[0237] In step S861, the control unit 800 may use the first embodiment. Figure 7The method described in step S805 of the processing sequence shown in the flowchart is used to select at least one of the first to third examples.

[0238] The control unit 800 sends the complement of the selected second learning model 82M from the communication unit 804 to the control device 100 (step S863).

[0239] The control unit 110 receives from the information processing device 8 multiple complements of a second learning model 82M that has been learned using input information from at least one of the human-powered vehicles 1 and the rider (step S157). The control unit 110 selects the second learning model 82M from the multiple complements (step S159).

[0240] In step S159, the control unit 110 selects from a plurality of backups the output information output when the input information of the driving scenario already learned by the first learning model 11M is input to the learning model for that driving scenario, and a second learning model 82M that is similar to the output information output when the input information is input to the learning models 11MA, 11MB, etc. More specifically, the control unit 110 selects from the first learning model 11M any of the learning models 11MA, 11MB, etc., for other driving scenarios that are different from the driving scenario determined in step S133. The control unit 110 selects the same learning model for the same driving scenario from the backups of the second learning model 82M. The control unit 110 stores the output information output when the input information of the learned driving scenario is input to any of the selected models, such as learning model 11MA. The control unit 110 stores the output information output when the input information of the learned driving scenario is input to the learning models selected from the backups. The control unit 110 selects a learning model from the supplementary learning models that outputs information similar to the output information output by the selected learning model 11MA. The control unit 110 then determines a supplementary second learning model 82 that includes the selected learning model.

[0241] In step S159, the control unit 110 selects from a plurality of backups a second learning model 82M that is similar to control data determined based on output information when input information of a driving scenario already learned by the first learning model 11M is input, and control data determined based on output information when the input information is input to the first learning model 11M. More specifically, the control unit 110 selects from the first learning model 11M any of the learning models 11MA, 11MB, ... of other driving scenarios already learned that are different from the driving scenario determined in step S133. The control unit 110 selects from the backups of the second learning model 82M a learning model of a driving scenario that is the same as any of the selected models, such as learning model 11MA. The control unit 110 inputs the input information of the learned driving scenario to the selected learning model 11M. For the learned driving scenario, the acquisition device control unit 116 determines control data based on the output information obtained from the selected learning model 11MA. This control data is not used for control. The control unit 110 inputs the learned driving scenario input information to a learning model selected from the second learning model 82M that is the same as the learning model 11MA for the driving scenario. The control unit 110 acquires control data determined by the device control unit 116 based on the output information output from the learning model for the same driving scenario as the learning model 11MA. The control unit 110 selects a second learning model 82M from the second learning model 82 as a supplement, which includes a learning model that outputs output information similar to the control data determined based on the output information from the learning model 11MA, serving as the source of the control data.

[0242] The control unit 110 selects a learning model from among the multiple learning models included in the selected second learning model 82, which is different from the already learned driving scenario and corresponds to the unlearned driving scenario determined in step S133 (step S161).

[0243] The control unit 110 performs supplementary processing (step S163) by storing the acquired learning model as the learning model in the driving scenario determined in step S133 in the first learning model 11M, and then ends the processing.

[0244] Figure 18 This diagram illustrates the processing performed by the supplementary processing unit 118 in the fifth embodiment. Figure 18 The first learning model 11M initially learned the models 11MA, 11MB, and 11MC for 3 out of 9 driving scenarios, including highway driving scenarios, but did not learn the models for other driving scenarios. Figure 17In the diagram, solid lines represent the learned models 11MA, 11MB, and 11MC for driving scenarios that have been learned, while dashed lines represent the learned models for driving scenarios that have not been learned.

[0245] In storage Figure 18 When the manually driven vehicle 1 of the control device 100 of the first learning model 11M shown starts off-road driving, the control unit 110 determines the driving scenario as off-road based on the tilting of the manually driven vehicle 1, etc. The control unit 110 determines that the first learning model 11M does not include the learning model for off-road driving scenarios that has already been learned ("Yes" in S151). The control unit 110 requests the information processing device 8 to obtain the complement of the second learning model 82M that includes the learning model for off-road driving scenarios that has already been learned. From the complement of the obtained second learning model 82M, the control unit 110 selects the second learning model 82M that outputs the most similar output information in the road scenario. The control unit 110 obtains the learning model for off-road driving scenarios included in the selected second learning model 82M. Thus, as Figure 18 As shown, the first learning model 11M is supplemented by learning models 11MD, 11ME, and 11MF, which have already learned the off-road driving scenario.

[0246] Figure 19 This is a flowchart illustrating an example of control processing using the first learning model 11M of the fifth embodiment. The control device 100 repeatedly executes the following processing in control mode.

[0247] The control unit 110 of the control device 100 acquires input information related to the driving of the manually driven vehicle 1 from the sensor 60 (step S251). Based on the input information acquired from the sensor 60, the control unit 110 determines the driving scenario (step S253).

[0248] The control unit 110 selects a learning model for the determined driving scenario from the first learning model 11M (step S255). The control unit 110 inputs the acquired input information related to the driving of the human-powered vehicle 1 to the learning model selected in step S255 (step S257). The control unit 110 acquires the output information output from the first learning model 11M (step S259).

[0249] Based on the output information obtained in step 259, the control unit 110 determines the control data of the transmission device 31 through the device control unit 116 (step S261). In step S261, the control unit 110 can determine the transmission ratio itself or whether to change speed.

[0250] The control unit 110 controls the transmission device 31 using the determined control data (step S263), and then ends the process. The control unit 110 repeatedly executes the processes of steps S251-S263.

[0251] In the fifth embodiment, as described above, for driving scenarios that have not been learned, other second learning models 82M can be used to supplement the first learning model 11M. For driving scenarios that have not been learned, the first learning model 11M, supplemented by other second learning models 82M, can also be used to achieve automatic control.

[0252] (Sixth Implementation Method)

[0253] Figure 20 This diagram illustrates the control device 100 and information processing device 8 according to the sixth embodiment. The structure of the control device 100 in the sixth embodiment is the same as that of the control device 100 in the first embodiment. Except for the processing sequence shown below in the information processing device 8, the control device 100 and information processing device 8 of the sixth embodiment are otherwise identical to those of the control device 100 and information processing device 8 of the first embodiment. For structures common to the control device 100 and information processing device 8 of the sixth embodiment shown below, the same reference numerals are used, and detailed descriptions are omitted.

[0254] In the sixth embodiment, the second learning model 82 includes learning models for each driving scenario. In the sixth embodiment, the information processing device 8 performs statistical processing on parameters including at least one of the weights and biases in the learning models for each driving scenario collected from each control device 100, thereby creating each learning model.

[0255] Figure 21 This is a flowchart illustrating an example of the processing sequence in the information processing apparatus 8 according to the sixth embodiment. If the first learning model 11M is sent from the control device 100, the information processing apparatus 8 performs the following processing.

[0256] If the control unit 800 receives the first learning model 11M (step S601), it determines the identification data of the human-powered vehicle 1 of the control device 100, which is the transmission source (step S603).

[0257] The control unit 800 extracts other second learning models 82M corresponding to the human-powered vehicle or rider that are similar to at least one of the human-powered vehicle 1 and the rider of the control device 100 that is the request source (step S605) for the received first learning model 11M. The second learning models 82M are stored in the information processing device 8 according to the type of each rider and the type of human-powered vehicle 1.

[0258] In step S605, the control unit 800 uses the first embodiment Figure 7 At least one of the first to third examples described in step S805 of the flowchart is used to extract the second learning model 82M.

[0259] For the first learning model 11M included in the learning model for each driving scenario and the second learning model 82M included in the learning model for each driving scenario, the control unit 800 performs statistical processing on the parameters including at least one of the weights and biases, and then creates each learning model (step S607).

[0260] The control unit 800 uses the second learning model 82M, which includes the created learning models, to update the second learning model 82M extracted in step S605 (step S609), and then ends the process.

[0261] Figure 22 This is a schematic diagram of the second learning model 82 in the sixth embodiment. In the sixth embodiment, as... Figure 22 As shown, the second learning model 82 includes learning models for each driving scenario. The learning models for each driving scenario are as follows: Figure 21 As described herein, statistical processing is performed on the parameters of at least one of the weights and biases of multiple learning models, including those collected from the control device 100 of the same type of human-powered vehicle 1 driven by the same type of rider, for the same driving scenario.

[0262] Through the processing of the information processing device 8 of the sixth embodiment, the second learning model 82M stored in the information processing device 8 is a model obtained by statistically summarizing the models learned in multiple human-powered vehicles 1. In this way, the control device 100 can supplement the learning model with a general second learning model 82M instead of using the learning model of a specific human-powered vehicle 1 and a specific rider.

[0263] The embodiments disclosed above are illustrative in all respects and are not intended to limit the invention. The scope of the invention is defined by the claims, including all modifications within the meaning and scope of the claims.

[0264] Symbol explanation:

[0265] 1…Human-driven vehicle, 10…Vehicle body, 10A…Frame, 10B…Front fork, 12…Handlebars, 14…Front wheel, 16…Rear wheel, 18…Seat, 20…Drive mechanism, 21…Crank, 21A…Crankshaft, 21B…Right crank, 21C…Left crank, 23…First sprocket assembly, 23A…Sprocket, 25…Second sprocket assembly, 25A…Sprocket, 27…Chain, 29…Pedals, 30…Components, 31…Transmission, 33…Suspension, 35…Seatpost, 37…Brake, 371…Front brake, 372…Rear brake 39…Auxiliary device, 40…Operating device, 40A…Operating unit, 40B…Gear shift indicator, 40C…Suspension indicator, 40D…Seat post indicator, 40E…Brake indicator, 40F…Auxiliary indicator, 50…Battery, 51…Battery body, 53…Battery bracket, 60…Sensor, 61…Speed ​​sensor, 62…Acceleration sensor, 63…Torque sensor, 64…Peak cadence sensor, 65…Gyroscope sensor, 66…Seating sensor, 67…Camera, 68…Position information sensor, 7…Information terminal device 100…Control device, 110…Control unit, 112…Storage unit, 114…Wireless communication device, 116…Device control unit, 118…Supplementary processing unit, 10P…Device control program, 12P…Supplementary processing program, 11M…First learning model, 111…Input layer, 113…Intermediate layer, 115…Output layer, 11MA…Learning model, 11MB…Learning model, 11MC…Learning model, 11MD…Learning model, 11ME…Learning model, 11MF…Learning model, 11MG…Learning model, 11MI…Learning Model, 13M… First learning model, 15M… First learning model, 17M… First learning model, 19M… First learning model, 8… Information processing device, 800… Control unit, 802… Storage unit, 822… Model database, 804… Communication unit, 82M… Second learning model, 900… Non-temporary storage medium, 90P… Device control program, 92P… Supplementary processing program, 91M… First learning model, 93M… First learning model, 95M… First learning model, 97M… First learning model, 99M… First learning model.

Claims

1. A human-powered vehicle control device, comprising: The acquisition department acquires input information related to the operation of the human-powered vehicle; The storage unit stores a first learning model, which learns based on the acquired input information to output output information related to the control of the device mounted on the human-powered vehicle. The control unit controls the device using control data determined based on output information obtained by inputting the input information into the first learning model; as well as The supplementary processing unit performs processing to supplement the first learning model in the storage unit with a second learning model, which has been learned using input information from at least one of the human-powered vehicles and the rider of the human-powered vehicle. The control unit requests the second learning model based on a comparison between the output information from the first learning model and the actual operations performed by the cyclist on the device corresponding to the input information.

2. The human-powered vehicle control device according to claim 1, wherein, The supplementary processing unit uses the second learning model to update at least a portion of the first learning model in the storage unit.

3. The human-powered vehicle control device according to claim 1, wherein, The supplementary processing unit uses the input information acquired by the acquisition unit and the output information output when the input information is input into the second learning model as learning data to learn the first learning model.

4. The human-powered vehicle control device according to any one of claims 1 to 3, wherein, The first learning model learns by utilizing input information acquired by the acquisition unit in multiple different driving scenarios. For a driving scenario that is different from the driving scenario that the first learning model has already learned but has not been learned, and for a driving scenario that the second learning model has already learned, the supplementary processing unit uses the second learning model to supplement the first learning model.

5. The human-powered vehicle control device according to claim 4, wherein, The first learning model includes multiple learning models stored according to each driving scenario. The supplementary processing unit uses a portion of a second learning model that has already learned the unlearned driving scenario as a learning model corresponding to an unlearned driving scenario that is different from the driving scenario that has already been learned by the first learning model.

6. The human-powered vehicle control device according to claim 4, wherein, The driving scenarios are divided into at least one of highway, off-road, and urban driving.

7. The human-powered vehicle control device according to claim 4, wherein, The driving scenario is divided into at least one of uphill, flat, and downhill.

8. The human-powered vehicle control device according to any one of claims 1 to 3, wherein, The supplementary processing unit uses a second learning model that is similar to the output information output when the same input information is input to the first learning model.

9. The human-powered vehicle control device according to claim 8, wherein, The first learning model learns using input information acquired by the acquisition unit in multiple different driving scenarios. The supplementary processing unit uses, among multiple second learning models, the output information output when input information of a driving scenario already learned by the first learning model is input, and the second learning model that is similar to the output information output when the input information is input to the first learning model.

10. The human-powered vehicle control device according to claim 8, wherein, The first learning model and the second learning model include multiple learning models stored according to each driving scenario. The supplementary processing unit is configured as follows: Among multiple second learning models, a second learning model is used where the output information of the driving scenario already learned by the first learning model is used, and the output information is similar to that of the first learning model when the input information is input into the first learning model. From the multiple learning models included in the second learning model, obtain the learning model corresponding to the unlearned driving scenario that is different from the already learned driving scenario.

11. The human-powered vehicle control device according to any one of claims 1 to 3, wherein, The supplementary processing unit uses a second learning model, which is a model used in similar human-powered vehicle control devices based on the same input information to determine control data.

12. The human-powered vehicle control device according to claim 11, wherein, The first learning model learns using input information acquired by the acquisition unit in multiple different driving scenarios. The supplementary processing unit uses, among multiple second learning models, control data determined based on output information when input information of a driving scenario already learned by the first learning model is input, and second learning models similar to control data determined based on output information when the input information is input to the first learning model.

13. The human-powered vehicle control device according to claim 11, wherein, The first learning model and the second learning model include multiple learning models stored according to each driving scenario. The supplementary processing unit is configured as follows: Among multiple second learning models, a second learning model is used that uses control data determined based on the output information when input information of a driving scenario already learned by the first learning model is input, and control data that is similar to the control data determined based on the output information when the input information is input to the first learning model. Obtain the learning model corresponding to the unlearned driving scenario that is different from the already learned driving scenario from among the multiple learning models included in the second learning model.

14. The human-powered vehicle control device according to any one of claims 1 to 3, wherein, The supplementary processing unit uses a second learning model among multiple second learning models to learn from input information from other human-powered vehicles that are the same as or similar to the human-powered vehicle in terms of at least one of type and size.

15. The human-powered vehicle control device according to any one of claims 1 to 3, wherein, The supplementary processing unit uses, among multiple second learning models, a second learning model that has learned from input information from other human-powered vehicles equipped with devices of the same or similar category as the device and the manufacturer.

16. The human-powered vehicle control device according to claim 15, wherein, The device is identified as belonging to at least one of the following categories: transmission, suspension, seat post, braking device, and auxiliary device.

17. The human-powered vehicle control device according to any one of claims 1 to 3, wherein, The supplementary processing unit uses, among multiple second learning models, a second learning model that has learned from input information from human-powered vehicles driven by riders of the same or similar type as the riders of the human-powered vehicle.

18. The human-powered vehicle control device according to any one of claims 1 to 3, wherein, The first learning model stored in the storage unit is sent to other human-powered vehicle control devices.

19. The human-powered vehicle control device according to any one of claims 1 to 3, wherein, The second learning model is a model obtained by statistically processing the following parameters, which include one of the weights and biases of multiple models learned in multiple other human-powered vehicles.

20. A learning method, wherein, The computer mounted on the human-powered vehicle executes: Based on the output information of a first learning model that learns from input information related to the driving of the human-powered vehicle to output output information related to the control of the device mounted on the human-powered vehicle, and the actual operation of the device by the rider corresponding to the input information, a second learning model that learns from input information from at least one of the human-powered vehicle and the rider of the human-powered vehicle is selected from the outside. Perform processing that supplements the first learning model with the selected second learning model; Control data is determined based on the output information obtained by inputting the input information into the supplemented first learning model; as well as The device is controlled based on the determined control data.

21. A control method for a manually driven vehicle, wherein, The computer mounted on the human-powered vehicle executes: Based on the output information of a first learning model that learns from input information related to the driving of the human-powered vehicle to output output information related to the control of devices mounted on the human-powered vehicle, and a comparison with the actual operation of the device by the rider corresponding to the input information, a second learning model that learns from input information from at least one of the human-powered vehicles and the rider of the human-powered vehicle is selected from external sources. The process of supplementing the first learning model with the selected second learning model is then performed. Control data is determined based on the output information obtained by inputting the input information into the supplemented first learning model. The device is controlled according to the determined control data.

22. A computer program product that causes a computer mounted on a human-powered vehicle to perform the following processing: Based on the output information of a first learning model that learns from input information related to the driving of the human-powered vehicle to output output information related to the control of devices mounted on the human-powered vehicle, and a comparison with the actual operation of the device by the rider corresponding to the input information, a second learning model that learns from input information from at least one of the human-powered vehicles and the rider of the human-powered vehicle is selected from external sources. The process of supplementing the first learning model with the selected second learning model is then performed. Control data is determined based on the output information obtained by inputting the input information into the supplemented first learning model; and The device is controlled based on the determined control data.

Images