Control program, control method, and processing circuit for controlling work assistance device

Abstract

This processing circuit: acquires a worker's instruction information for a work assistance device; inputs the instruction information to a learning model that learns an operation of the work assistance device, the operation being in accordance with the worker's instruction; acquires, from the learning model, control information used to control the work assistance device; and outputs the control information.

Description

Processing Circuit, Control Method, and Control Program for Controlling a Work Support Device 【0001】 The present disclosure relates to a processing circuit, a control method, and a control program for controlling a work support device. 【0002】 Patent Document 1 discloses a teaching operation panel for a robot. The teaching operation panel includes a voice input unit for an operator to input an instruction voice, a voice recognition unit for recognizing the instruction voice acquired by the voice input unit, and a processing unit for transmitting at least one command of operation of the robot's operation program and cancellation of an alarm to a robot control device based on a result of recognition of the instruction voice by the voice recognition unit. 【0003】 WO2023 / 162248A1 【0004】 In the configuration of Patent Document 1, since the robot performs standardized operations, it may be difficult for some operators to move the robot as intended. 【0005】 One aspect of the present disclosure aims to facilitate moving the work support device as intended by the operator. 【0006】 A processing circuit according to one aspect of the present disclosure is a processing circuit for controlling a work support device, which acquires instruction information indicating an instruction of the operator for the work support device, inputs the instruction information to a learning model for learning an operation of the work support device according to the instruction of the operator, acquires control information used for controlling the work support device from the learning model, and outputs the control information. 【0007】 A control method according to one aspect of the present disclosure is a method for controlling a work support device, which acquires instruction information indicating an instruction of the operator for the work support device, inputs the instruction information to a learning model for learning an operation of the work support device according to the instruction of the operator, acquires control information used for controlling the work support device from the learning model, and outputs the control information. 【0008】 A control program according to one aspect of the present disclosure causes at least one processor to execute the above method. The program can be stored in a non-temporary and tangible storage medium readable by a computer. 【0009】 According to one aspect of this disclosure, it is possible to facilitate the operation of a work support device as intended by the worker. 【0010】 Figure 1 is a schematic diagram of a control system for a work support device according to an embodiment. Figure 2 is a block diagram of the work support device and controller of Figure 1. Figure 3 is a block diagram of the server of Figure 1. Figure 4 is a flowchart showing the processing of the control system of Figure 1. Figure 5 is a flowchart showing the processing of the work support device of Figure 1. Figure 6 is a diagram illustrating the processing using the learning model of Figure 3. Figure 7 is a diagram showing a modified version of the learning model of Figure 6. Figure 8 is a block diagram showing a modified version of the controller of Figure 2. Figure 9 is a block diagram of a first modified version of the server of Figure 3. Figure 10 is a block diagram of a second modified version of the server of Figure 3. 【0011】 The embodiments will be described below with reference to the drawings. 【0012】 Figure 1 is a schematic diagram of a control system 1 for a work support device according to an embodiment. As shown in Figure 1, the control system 1 comprises a plurality of controllers 3 and a server 4 that can communicate with the plurality of controllers 3 via a communication network N. Each of the plurality of controllers 3 controls a corresponding work support device 2. The communication network N includes, for example, the Internet, but may also include an intranet. The work support device 2 is a device that supports the work of a worker 5. In this embodiment, the work support device 2 is a device that supports the work of a worker 5 who supports the movements of a person being worked on 6. 【0013】The work support device 2 is used in places where human work may occur, such as nursing homes, hospitals, factories, and warehouses. For example, the work support device 2 is a care support device, the worker 5 is the caregiver, and the person receiving care 6 is the person being cared for. As a care support device, the work support device 2 may, for example, be a care chair for moving the person receiving care 6 to a desired position and orientation. If the person receiving care 6 uses the work support device 2 without the assistance of the worker 5, the person receiving care 6 can become the worker using the work support device 2. The work support device 2 may also be a device that assists the person receiving care 6 in performing tasks without the assistance of a caregiver. 【0014】 The multiple work support devices 2 include, for example, a first work support device 2A and a second work support device 2B. Since the first work support device 2A and the second work support device 2B are the same device, the first work support device 2A will be described as representative of work support device 2. The configuration of the work support device described below is merely an example, and various configurations can be adopted as long as they are equipped with controllable actuators. 【0015】 The work support device 2 comprises a plurality of wheels 11, a base frame 12 supported by the wheels 11, and a body 13 supported by the base frame 12. The body 13 has, for example, a seat and a backrest. The body 13 is connected to the base frame 12 so as to be rotatable around a vertical axis. 【0016】 The work support device 2 includes a first actuator 14 connected to each wheel 11. The first actuator 14 includes, for example, an electric motor. The first actuator 14 directly drives the corresponding wheel 11. Each wheel 11 may be, for example, a Mecanum wheel capable of changing the direction of travel. One first actuator 14 may drive two or more wheels 11. The type of wheel 11 is not particularly limited, and an omniwheel may be used instead of a Mecanum wheel. The wheels 11 may be steered by an active steering mechanism. 【0017】The work support device 2 includes a second actuator 15 that drives the body 13 to rotate the body 13 around a vertical axis relative to the base frame 12. The second actuator 15 includes, for example, an electric motor. The second actuator 15 makes it possible to change the orientation of the body 13 relative to the worker 6 without moving the wheels 11. 【0018】 The multiple controllers 3 include a first controller 3A and a second controller 3B. The first controller 3A is attached to the first work support device 2A, and the second controller 3B is attached to the second work support device 2B. Since the first controller 3A and the second controller 3B are the same device, the first controller 3A will be described as controller 3 in the following description. 【0019】 Figure 2 is a block diagram of the work support device 2 and controller 3 shown in Figure 1. As shown in Figure 2, the work support device 2 comprises a first actuator 14, a second actuator 15, an input interface 16, and a control circuit 17. The first actuator 14 and the second actuator 15 are as described above. The input interface 16 receives signals from the controller 3. The control circuit 17 controls the first actuator 14 and the second actuator 15 based on the signals input to the input interface 16. The input interface 16 includes at least one of a digital interface, an analog interface, and a wireless interface. The control circuit 17 includes a processor, system memory, storage memory, and driver circuitry. 【0020】The controller 3 is detachably attached to the work support device 2 via a mounting bracket 7. That is, the controller 3 is externally attached to the work support device 2. The controller 3 comprises a housing 3a. The mounting bracket 7 includes fasteners such as bolts, or bands. The housing 3a comprises, for example, a flange having holes. The housing 3a is fixed to the frame of the work support device 2 by inserting bolts into the holes in the flange and sandwiching the flange between the bolt heads and the frame of the work support device 2. The controller 3 can be replaced and used with any work support device 2 that corresponds to the controller 3, thus improving versatility. 【0021】 The controller 3 is powered by the work support device 2. The housing 3a may be provided with a power input terminal and a power cord. The controller 3 may also have a battery. By providing a battery, the power supply wires between the controller 3 and the work support device 2 can be omitted, and the attachment of the controller 3 to the work support device 2 becomes easier. The controller 3 may also be built into the work support device 2. The actuators 14 and 15 may also be attached externally to the work support device 2. 【0022】 Controller 3 includes a processor 21, system memory 22, and storage memory 23. The processor 21 may include a CPU (Central Processing Unit). The system memory 22 may include volatile memory. The storage memory 23 may include non-volatile memory. The storage memory 23 includes a hard disk, flash memory, or a combination thereof. The storage memory 23 stores a control program P1. An example of a processing circuit 20 is a configuration in which the processor 21 executes the control program P1 read from the storage memory 23 to the system memory 22. The processing circuit 20 of the first controller 3A may be referred to as the first processing circuit, and the processing circuit 20 of the second controller 3B may be referred to as the second processing circuit. 【0023】The controller 3 includes a microphone 24, a camera 25, a user interface 26, an IC reader 27, a wireless communication device 28, and an output interface 29. The microphone 24 collects the voice emitted by the worker 5. The microphone 24 outputs voice data indicating the collected voice. The microphone 24 serves as an input interface that receives voice instructions from the worker 5. When the worker 5 is identified by voice characteristics as described later, the microphone 24 can also serve as an input interface into which information for identifying the worker 5 is input. When the person being worked on 6 uses the work support device 2 without the assistance of the worker 5, the person being worked on 6 is the worker using the work support device 2, and the microphone 24 collects the voice emitted by the person being worked on 6. 【0024】 Camera 25 photographs the area around the work support device 2 and outputs image data showing the captured image. Camera 25 also photographs the worker 5. For example, camera 25 is positioned to photograph the worker 5's face, iris, or fingerprints. User interface 26 receives identification information, such as a worker number, from the worker 5. For example, user interface 26 may be a touch panel display or an operation panel. When an IC tag carried by the worker 5 is brought close to the IC reader 27, it reads the worker 5's identification information from the IC tag contactlessly. Camera 25, user interface 26, and IC reader 27 can serve as input interfaces into which information for identifying the worker 5 is input. In particular, microphone 24 and camera 25 can serve as input interfaces for biometric authentication. 【0025】 The control program P1 includes a worker identification program 50. The processor 31 executes the worker identification program 50 and identifies which worker 5 the information belongs to based on the information about the worker 5 input to the microphone 24, camera 25, user interface 26, or IC reader 27, thereby identifying the worker 5's identification information. 【0026】The processor 31 may execute the worker identification program 50 to extract speaker-specific features from the voice input to the microphone 24 to identify the worker 5 who is the speaker, and obtain the identification information of the identified worker 5. The processor 31 may execute the worker identification program 50 to extract unique features from the face, iris, or fingerprint of the worker 5 captured in the image taken by the camera 25 to identify the worker 5, and obtain the identification information of the identified worker 5. The processor 31 may execute the worker identification program 50 to obtain the identification information input by the worker 5 to the user interface 26 as the identification information of the worker 5. The processor 31 may execute the worker identification program 50 to obtain the identification information output from the IC reader 27 that reads the worker 5's IC tag as the identification information of the worker 5. 【0027】 The camera 25, user interface 26, and IC reader 27 are merely illustrative examples and can be selectively adopted. If the worker 5 is identified by voice characteristics, as described later, the controller 3 does not need to include the camera 25, user interface 26, and IC reader 27. The type of input interface into which information for identifying the worker 5 is input is determined according to design requirements. 【0028】 The wireless communication device 28 wirelessly connects to the communication network N to which the server 4 is connected. The wireless communication device 28 may include wireless communication devices for LTE communication and mobile communication such as local 5G, or it may include wireless communication devices for Wi-Fi communication. The output interface 29 outputs a signal to the input interface 16 of the work support device 2. The output interface 29 includes at least one of a digital interface, an analog interface, and a wireless interface. The output interface 29 of the controller 3 may be connected to the input interface 16 of the work support device 2 by wire or by wireless communication. If the work support device 2 incorporates the controller 3, the input interface 16 and the output interface 29 may be omitted. 【0029】Figure 3 is a block diagram of the server 4 in Figure 1. As shown in Figure 4, the server 4 includes a processor 31, system memory 32, storage memory 33, and a communication interface 34. The communication interface 34 includes a communicator connected to the communication network N. The server 4 communicates with the controller 3 via the communication network N using its communication interface 34 and the wireless communication device 28 of the controller 3. 【0030】 The processor 31 may include a CPU (Central Processing Unit). The system memory 32 may include volatile memory. The storage memory 33 may include non-volatile memory. The storage memory 33 may include a hard disk, flash memory, or a combination thereof. The storage memory 33 stores the processing program P2. An example of a processing circuit 30 is a configuration in which the processor 31 executes the processing program P2 read from the storage memory 33 to the system memory 32. The processing circuit 30 of the server 4 is connected to the processing circuit 20 of the controller 3 via a communication network N in order to receive input from the processing circuit 20 of the controller 3. 【0031】 The processing program P2 includes a speech recognition program 40, multiple learning models 41, an identification information receiving program 42, a correction judgment program 43, a learning data program 44, and a feedback learning program 45. The processor 31 executes the speech recognition program 40 to convert the voice of the worker 5 received from the controller 3 via the communication network N into text using known speech recognition technology. Multiple learning models 41 are prepared for each identification information of the worker 5. Multiple learning models 41 are updated for each worker 5. 【0032】 The learning model 41 is a trained model that takes instruction information indicating instructions from the worker 5 as input and outputs control information indicating the control of the work support device 2 corresponding to those instructions. When the processor 31 that executes the learning model 41 receives instruction information indicating instructions from the worker 5, which has been converted into text by the speech recognition program 40, the processor 31 outputs control information indicating the control of the work support device 2 corresponding to that instruction information. 【0033】The processor 31 executing the learning model 41 may output NULL as control information if there is no learned instruction that matches the input information. The processor 31 may execute the processing program P2 and, if the voice from the worker 5 contains a predetermined trigger sound that indicates the start of an instruction, consider the voice immediately following that trigger sound as an instruction, and if there is no trigger sound, it may determine that the voice from the worker 5 is not an instruction. If the multiple learning models 41 corresponding to each identification information of the worker 5 are considered as a single comprehensive learning model as a whole, then this comprehensive learning model can be considered to take identification information and instruction information as input and output control information indicating the control of the work support device 2. 【0034】 The control information includes at least one of the following: the start and stop of the operation of the first actuator 14 and the second actuator 15 of the work support device 2, the amount of movement of the first actuator 14 and the second actuator 15, the direction of operation of the first actuator 14 and the second actuator 15, and the operating speed of the first actuator 14 and the second actuator 15. The learning model 41 is an initial model corresponding to the work support device 2 before being updated by feedback learning, which will be described later. In the initial model, the control information is set by default through prior machine learning. 【0035】 The processor 31 executes the identification information receiving program 42 to receive the identification information of worker 5 from the controller 3 via the communication network N. Instead of the controller 3's control program P1 having the worker identification program 50, the server 4's processing program P2 may have the worker identification program 50. For example, the server 4 may receive information about worker 5 input to the controller 3's microphone 24, camera 25, user interface 26, or IC reader 27 via the communication network N, identify which worker 5 the received information belongs to, and then identify the worker 5's identification information. 【0036】The processor 31 executes the correction determination program 43 to determine whether the instruction information input to the learning model 41 is an instruction that corrects the operation of the work support device 2 corresponding to the previous instruction information. The initial model of the learning model 41 has already learned the control information corresponding to the instruction to correct the control. The processor 31 may also execute the correction determination program 43 to determine whether the current instruction information input to the learning model 41 is an instruction that corrects the operation of the work support device 2 if it was input while the work support device 2 was operating in response to the previous instruction information. 【0037】 For example, the processor 31 may execute the correction determination program 43 and determine that the current instruction information input to the learning model 41 is a correction instruction if it indicates an instruction that is the opposite of the operation of the work support device 2 in response to the previous instruction information. The processor 31 may execute the correction determination program 43 and determine that the current instruction information input to the learning model 41 is a correction instruction if it indicates an instruction that corrects the amount of operation of the work support device 2 in response to the previous instruction information. The processor 31 may execute the correction determination program 43 and determine that the current instruction information input to the learning model 41 is a correction instruction if it indicates an instruction that corrects the direction of operation of the work support device 2 in response to the previous instruction information. The processor 31 may execute the correction determination program 43 and determine that the current instruction information input to the learning model 41 is a correction instruction if it indicates an instruction that corrects the speed of operation of the work support device 2 in response to the previous instruction information. 【0038】The processor 31 executes the correction determination program 43, and when it determines that the current instruction is a correction instruction, it executes the learning data program 44 to create updated learning data, using the previous instruction information as input and the correction operation result of the work support device 2 in response to the current instruction information as output. If multiple instruction pieces to correct the operation of the work support device 2 are input consecutively while the work support device 2 is operating in response to the previous instruction information, the processor 31 may create updated learning data, using the first instruction information as input and the correction operation result of the work support device 2 in response to the last instruction information as output. 【0039】 The processor 31 executes a feedback learning program 45 to perform feedback learning on the learning model 41 using the learning data program 44, which is created during operation of the work support device 2 using the learning model 41. The initial learning model 41 can be updated by feedback learning based on the next instruction information input from the worker 5 after outputting control information in response to the instruction information from the worker 5. In this way, the learning model 41 is automatically customized to become a learning model that reflects the characteristics of the worker 5. 【0040】 Feedback learning may be performed by reinforcement learning. For example, the learning model 41 may be reinforced to increase the reward if the instruction information input to the learning model 41 is not an instruction that corrects the operation of the work support device 2 corresponding to the previous instruction information, and to decrease the reward if the instruction information input to the learning model 41 is an instruction that corrects the operation of the work support device 2 corresponding to the previous instruction information. In this case, the reward may be set to decrease as the correction amount increases. The correction amount may be the difference between the control value indicated by the control information corresponding to the previous instruction information and the control value indicated by the control information corresponding to the current instruction information. 【0041】Figure 4 is a flowchart showing the processing of the control system 1 in FIG. 1. FIG. 6 is a drawing for explaining the processing using the learning model in FIG. 3. The processing of the control system 1 will be described along the flow of FIG. 4 with appropriate reference to FIGS. 1 to 3 and 6. Hereinafter, the processing of the controller 3 is executed by the processing circuit 20, and the processing of the server 4 is executed by the processing circuit 30. When the power of the server 4 is turned on, it determines whether or not to receive a session start signal in step S11. The server 4 may be activated from the standby state when receiving a session start notification from the controller 3. When the processing of the flowchart shown in FIG. 4 ends, the server 4 shifts to step S11 again. 【0042】 When the operator 5 turns on the power of the controller 3 or when the operator 5 inputs the start of work to the controller 3, in step S1, the controller 3 starts a session in which the specific operator 5 performs work and waits for an input from the operator 5. In step S1, the controller 3 transmits a session start notification to the server 4 and shifts to step S2. 【0043】 The server 4 determines whether or not it has received a session start notification from the controller 3 in step S11. If it is determined that the session start notification has been received, it shifts to step S12. 【0044】 In step S2, the controller 3 determines whether or not it has acquired the identification information of the operator 5 who uses the work support device 2 by executing the operator identification program 50. If it is determined that the identification information has been acquired, the controller 3 shifts to step S3. In step S3, the controller 3 transmits the acquired identification information to the server 4 via the communication network N to the server 4 and shifts to step S4. 【0045】 In step S12, the server 4 determines whether or not it has received the identification information. If it is determined that the identification information has been received, the server 4 shifts to step S13. 【0046】The controller 3 determines whether the instruction information for the operator 5 to instruct the operation of the work support device 2 is acquired by the microphone 24 in step S4. When it is determined that the instruction information has been acquired, the controller 3 transmits the acquired instruction information to the server 4 in step S5 and proceeds to step S6. 【0047】 The server 4 determines whether it has received the instruction information in step S13. When it is determined that the instruction information has been received, the server 4 calculates the control information indicating the control of the work support device 2 according to the identification information and instruction information received from the controller 3 in step S14, transmits the control information to the controller 3, and proceeds to step S15. 【0048】 As shown in FIG. 6, the server 4 has a plurality of learning models 41 respectively corresponding to the identification information of a plurality of operators 5. The learning model 41 includes a first learning model 41A corresponding to the identification information of the first operator 5A and a second learning model 41B corresponding to the identification information of the second operator 5B. When the server 4 receives the identification information corresponding to the first operator 5A from the controller 3, it uses the first learning model 41A in the calculation. When the server 4 receives the identification information corresponding to the second operator 5B from the controller 3, it uses the second learning model 41B in the calculation. 【0049】 The first learning model 41A and the second learning model 41B can output different control information for the same input. The plurality of learning models 41 have configurations corresponding to each identification information at the stage of the initial model. For example, the identification information of a plurality of operators 5 may be grouped, and a plurality of learning models 41 respectively corresponding to each group may be prepared as the initial model. When the operator 5 is initially registered in the server 4, any one of the groups may be selected, and the initial model corresponding to the selected group may be determined as the learning model 41 corresponding to the operator 5. The first learning model 41A and the second learning model 41B may have the same configuration as each other at the stage of the initial model which is the initial state. In this case, since it is not necessary to prepare a plurality of learning models 41 different from each other, the manufacturing becomes easy. 【0050】In step S14, the server 4 inputs instruction information indicating instructions from the worker 5 recognized using the speech recognition program 40 into the learning model 41 and acquires control information output from the learning model 41. In step S14, the server 4 further transmits the control information to the controller 3 and proceeds to step S15. 【0051】 In step S6, the controller 3 determines whether or not it has received control information from the server 4. If it is determined that it has received control information from the server 4, in step S7, the controller 3 controls the first actuator 14 or the second actuator 15 based on the control information received from the server 4 and proceeds to step S8. In step S7, the controller 3 may operate the first actuator 14 and the second actuator 15 simultaneously. 【0052】 In step S15, the server 4 uses the correction determination program 43 to determine whether the instruction information input to the learning model 41 is an instruction to correct the control indicated by the control information corresponding to the previous instruction information. If the server 4 determines in step S15 that the current instruction information is not a correction instruction, the server 4 skips step S16. If the server 4 determines in step S15 that the current instruction information is a correction instruction, the server 4 uses the learning data program 44 in step S16 to create updated learning data with the previous instruction information as input and the correction operation result of the work support device 2 in response to the current instruction information as output. In step S16, the server 4 uses the created learning data to perform feedback learning on the learning model 41 and proceeds to step S17. 【0053】 For example, if worker 5 gives the voice command "move a little to the right" and then gives the voice command "move back a little to the left," the learning model 41 is updated so that the amount of movement that worker 5 means by "a little" is reduced through feedback learning. The amount of movement refers to, for example, the operating angle of an actuator. When the amount of movement that "a little" means is reduced through feedback learning, for example, the amount of movement is reduced by 1 degree. 【0054】For example, if worker 5 gives the voice command "rotate to the right" and then gives the voice command "no, rotate to the left," the learning model 41 will be updated so that the direction of movement that worker 5 means by "right" is corrected to the opposite direction through feedback learning. If worker 5 intended to instruct the work support device 2 to move to the right from their perspective, but the work support device 2 moves to the left from worker 5's perspective to move to the right from the work support device 2's perspective, the learning model 41 will be updated so that the "right" and "left" that worker 5 says will be interpreted in reverse from the next time onward. 【0055】 For example, if worker 5 gives the voice command "move to the right" and then gives the voice command "slower," the learning model is updated so that the speed of the movement that worker 5 means is reduced through feedback learning. 【0056】 Through this feedback learning, the work support device 2 will operate in a manner that matches the intent of the voice instructions given by worker 5 from the next time onward. Feedback learning is performed individually for each of the first learning model 41A and the second learning model 41B, which correspond to the first worker 5A and the second worker 5B, respectively. The first learning model 41A and the second learning model 41B are updated in different ways according to the characteristics of the first worker 5A and the second worker 5B, respectively. Therefore, the operation of the work support device 2 is appropriately determined to match the characteristics of each individual worker 5A and 5B according to the identification information of workers 5A and 5B. 【0057】 After the first worker 5A updates the first learning model 41A by using the first work support device 2A, when the first worker 5A uses the second work support device 2B, the second controller 3B that controls the second work support device 2B will utilize the updated first learning model 41A. Therefore, even when the first worker 5A uses the second work support device 2B for the first time, the second work support device 2B will operate in a manner that matches the characteristics of the first worker 5A's voice instructions. 【0058】If the controller 3 determines in step S8 that the session in which a specific worker 5 is performing work has not ended, it proceeds to step S4. The session ends, for example, when worker 5 turns off the power to the controller 3 or when worker 5 inputs "work finished" to the controller 3. If the controller 3 determines in step S8 that the session in which a specific worker 5 is performing work has ended, in step S9 it sends a session end notification to the server 4, terminates processing, and transitions to a standby state in response to power off or input of work finished. 【0059】 If Server 4 has not received a session termination notification from Controller 3, in step S17, it determines that the session in which the specific worker 5 is performing work has not ended, and proceeds to step S13 to wait for the next instruction from Controller 3. If Server 4 receives a session termination notification from Controller 3, it terminates processing and transitions to a standby state. 【0060】 Figure 5 is a flowchart showing the processing of the work support device 2. Hereafter, the processing of the work support device 2 is executed by the control circuit 17. When the power of the work support device 2 is turned on, in step S21, it determines whether or not to receive a session start notification. If the work support device 2 receives a session start notification from the controller 3, it may start up from the standby state. When the processing of the flowchart shown in Figure 5 is completed, the work support device 2 returns to step S21. The work support device 2 may also be configured to supply power to the controller 3 even when the power is off. 【0061】 In step S21, the work support device 2 determines whether it has received a notification from the controller 3 that a session has started. If it determines that it has received a notification that a session has started, it proceeds to step S22. In step S22, the work support device 2 determines whether it has received a control signal to control the first actuator 14 or the second actuator 15. The control signal includes, for example, a signal given to the processor of the control circuit 17. If it determines that it has received a control signal to control the first actuator 14 or the second actuator 15, it executes the process in step S23. 【0062】 In step S23, the work support device 2 operates the first actuator 14 or the second actuator 15 based on the control signal and proceeds to step S24. The work support device 2 may operate the first actuator 14 and the second actuator 15 simultaneously based on the control signal. If the work support device 2 has not received a session end notification from the controller 3, in step S24 it determines that the session in which the specific worker 5 is performing work has not ended, and proceeds to step S22 to wait for the next control signal from the controller 3. If the work support device 2 receives a session end notification from the controller 3, it terminates processing and proceeds to a standby state. 【0063】 Figure 7 is a diagram showing a modified version of the learning model 41 in Figure 6. As shown in Figure 7, the modified learning model 141 may be a trained model that takes the identification information of the worker 5 and instruction information indicating instructions from the worker 5 as inputs, and outputs control information indicating the control of the work support device 2. As described above, the update of the learning model 141 by feedback learning is also performed using the identification information as input, so the learning model 141 is feedback-learned in accordance with the identification information of the worker 5. The other configurations of the server 4 when using the learning model 141 are the same as in the embodiment described above, so their explanation is omitted. 【0064】Figure 8 is a block diagram showing a modified version of the controller 3 in Figure 2. In the modified version of Figure 8, components common to the above embodiment are denoted by the same reference numerals and their descriptions are omitted. As shown in Figure 8, the program P11 of the modified controller 203 includes a speech recognition program 40, a learning model 41, a worker identification program 50, a correction judgment program 43, a learning data program 44, and a feedback learning program 45. The processing circuit 220 of the controller 203 executes the speech recognition program 40, the learning model 41, the correction judgment program 43, the learning data program 44, and the feedback learning program 45. In the controller 203 in this modified version, the learning function is provided on the controller 203 side rather than the server 4 side, which speeds up the response time from when the worker 5 inputs instruction information to the controller 203 until the controller 203 acquires control information for the work support device 2. 【0065】 In the configuration described above, the server 4 had multiple learning models 41 corresponding to the identification information of multiple workers 5, but it may have only one learning model 41. In that case, the server 4 may input instruction information indicating the worker's instructions into the learning model 41 without receiving the worker's identification information and acquire control information output from the learning model 41. For example, if one worker uses the work support device 2 exclusively, the server 4 may execute the feedback learning program 45 without referring to the worker's identification information and update the learning model 41 through feedback learning so that it matches the worker's characteristics. 【0066】 In the configuration described above, the instruction information indicating worker 5's instructions was information related to worker 5's voice, but it may be other types of information. For example, the instruction information indicating worker 5's instructions may be information related to worker 5's body language. In that case, the instruction information that worker 5 inputs to the controller 3 would be information indicating worker 5's motion rather than worker 5's voice. 【0067】Figure 9 is a block diagram of a first modified version of the server 4 in Figure 3. As shown in Figure 9, the server 104 of the first modified version stores a motion recognition program 140 in the storage memory 33 instead of the voice recognition program 40 described above. The controller 3 shown in Figure 2 captures the motion of the worker 5 with the camera 25 and transmits the captured video showing the motion of the worker 5 to the server 4. Upon receiving the video, the server 104 has a processor 31 that executes the motion recognition program 140 and recognizes the motion of the worker 5 from the video using known motion recognition technology. 【0068】 The learning model 141 of the server 104 takes the motion recognized by the motion recognition program 140 as input and outputs control information indicating the control of the work support device 2. For example, when the learning model 141 receives input of a motion in which the worker 5 waves their hand, it outputs control information indicating the direction of operation of the work support device 2 according to the angle of the hand. The learning model 141 also outputs control information indicating the operating speed of the work support device 2 according to the speed at which the hand is waved. If the worker 5 moves only their hand and points to the side, the work support device 2 may be controlled to move a small amount in the direction the finger is pointing, and if the worker 5 also moves their arm and points to the side, the work support device 2 may be controlled to move a large amount in the direction the finger is pointing. 【0069】 Furthermore, the instruction information indicating the worker's instructions may be information input from an input device operated by the worker 5, instead of information related to voice or motion. Figure 10 is a block diagram of a second modified example of the server 4 in Figure 3. As shown in Figure 10, the server 204 of the second modified example does not have the aforementioned voice recognition program 40 or motion recognition program 140, and receives instruction information indicating the worker's instructions from an input device that the worker 5 can operate. The input device may be, for example, a touch panel device, a remote control device, etc. The touch panel device may be the user interface 26 of the controller 3, or it may be the touch panel display of a smartphone held by the worker 5. 【0070】The learning model 241 of the server 204 takes instruction information received from the input device as input and outputs control information indicating the control of the work support device 2. For example, when worker 5 presses the right movement button, the learning model 241 controls the work support device 2 to move to the right from the worker 5's perspective. If worker 5 presses the right movement button repeatedly after the initial press, the learning model 241 learns to increase the operating speed of the work support device 2 in response to this instruction from worker 5. 【0071】 The technology of this disclosure is not limited to the embodiments and modifications described above. For example, the work support device 2 as a care support device may include, for example, a lift device and may also include a bathing device. The lift device includes, for example, a transfer lift device, a standing lift device and a bathing lift device. The control system 1 includes a plurality of controllers 3 and a server 4, but may also be configured to include one controller 3 and a server 4. 【0072】 Controller 3 operates the first actuator 14 or the second actuator 15 by transmitting a control signal to the processor of the control circuit 17 of the work support device 2. However, the controller 3 may also directly control the first actuator 14 or the second actuator 15 by transmitting the control signal to the driver circuit of the control circuit 17. In this case, the processing in step S22 is omitted in the flowchart of Figure 5. When the controller 3 directly controls the first actuator 14 or the second actuator 15 by transmitting the control signal to the driver circuit of the control circuit 17, the first actuator 14 or the second actuator 15 can be operated even if the processing circuit 20 and the control circuit 17 cannot communicate. 【0073】 The controller 3 may further include a driver circuit, and the processing circuit may control the driver circuit to operate the first actuator 14 or the second actuator 15 of the work support device 2. If the controller 3 further includes a driver circuit, the configuration of the work support device 2 can be simplified. If the controller 3 further includes a driver circuit, there is no need to standardize the communication between the processing circuit 20 and the control circuit 17, making it easier to apply the controller 3 to various work support devices 2. 【0074】 As described above, the embodiments have been explained as examples of the technology disclosed in this application. However, the technology in this disclosure is not limited to these embodiments and can be applied to embodiments that have been modified, replaced, added, or omitted as appropriate. Furthermore, it is possible to combine the components described in the embodiments to create new embodiments. For example, some components or methods in one embodiment may be applied to other embodiments, and some components in an embodiment can be separated from other components in that embodiment and extracted as appropriate. Among the components described in the attached drawings and detailed description are not only components that are essential for solving the problem, but also components that are not essential for solving the problem, in order to illustrate the technology. 【0075】 The functions of the elements disclosed herein can be performed using circuits or processing circuits, including general-purpose processors, dedicated processors, integrated circuits, ASICs (Application Specific Integrated Circuits), FPGAs (Field Programmable Gate Arrays), conventional circuits, and / or combinations thereof, configured or programmed to perform the disclosed functions. A processor is considered a processing circuit or circuit because it includes transistors and other circuits. In this disclosure, a circuit, unit, or means is hardware that performs the enumerated functions, or hardware programmed to perform the enumerated functions. The hardware may be hardware disclosed herein, or other known hardware that is programmed or configured to perform the enumerated functions. If the hardware is a processor, which is considered a type of circuit, the circuit, means, or unit is a combination of hardware and software, and the software is used to configure the hardware and / or the processor. 【0076】 [Embodiment] The embodiments described above are specific examples of the following embodiments. 【0077】(Aspect 1) A processing circuit for controlling a work support device, the processing circuit comprising: acquiring instruction information indicating instructions from an operator to the work support device; inputting the instruction information into a learning model that learns the operation of the work support device in response to the operator's instructions; acquiring control information used to control the work support device from the learning model; and outputting the control information. 【0078】 According to the configuration of Embodiment 1, the operation of the work support device is determined in response to the worker's instructions. Therefore, in a configuration in which the work support device is operated by instruction information, it is possible to easily operate the work support device as intended by the worker. 【0079】 (Aspect 2) The processing circuit according to aspect 1, wherein the processing circuit acquires identification information of the worker using the work support device, the learning model includes a learning model corresponding to the identification information, and the processing circuit inputs the instruction information to the learning model corresponding to the identification information. 【0080】 According to the configuration of Embodiment 2, the operation of the work support device corresponding to the worker's instruction information is determined according to the worker's identification information. Therefore, when multiple workers use the same work support device, it is possible to easily operate the work support device as each individual worker intends. 【0081】 (Aspect 3) The processing circuit according to aspect 1 or 2, wherein the instruction information includes information relating to voice, information relating to body language, or information input from an input device. 【0082】 According to the configuration of Embodiment 3, instructions can be given to the processing circuit by a method that is easy for the operator to use, such as voice, body language, or operation of an input device. 【0083】 (Aspect 4) A processing circuit according to any one of aspects 1 to 3, which is connected to the learning model via a communication network. 【0084】 According to the configuration of Embodiment 4, the computational load on the processing circuit can be reduced. 【0085】(Aspect 5) A processing circuit according to any one of aspects 1 to 4, which is attached to the work support device and includes the learning model. 【0086】 According to the configuration of embodiment 5, the response speed from the input of instruction information to the acquisition of control information can be increased. 【0087】 (Aspect 6) The work support device includes an actuator, and the control information includes the amount of operation of the actuator, the direction of operation of the actuator, or the operating speed of the actuator, according to any one of aspects 1 to 5. 【0088】 According to the configuration of embodiment 6, the operation content of the work support device in response to voice instructions from the worker can be appropriately determined to suit the characteristics of the worker. 【0089】 (Aspect 7) The processing circuit according to any one of aspects 1 to 6, wherein the learning model is updated by feedback learning using the next instruction information input from the worker after outputting the control information in response to the instruction information from the worker. 【0090】 According to the configuration of Embodiment 7, if the work support device operates in response to a voice command from the worker, and then the worker issues a voice command to correct the operation of the work support device, the content of that correction can be used to provide feedback learning to the learning model, so that the work support device can operate in accordance with the intent of the worker's voice command from the next time onward. 【0091】 (Aspect 8) The processing circuit according to aspect 7, wherein the learning model includes an initial model corresponding to the work support device, and the initial model is updated by the feedback learning. 【0092】 According to the configuration of embodiment 8, the work support device can be operated by voice commands from an initial state using a pre-trained learning model. 【0093】(Aspect 9) The learning model receives from a second processing circuit, which is different from the first processing circuit, the following inputs: identification information of the worker using a second work support device different from the first work support device, which is the work support device; and instruction information from the worker to the second work support device; and outputs control information for the second work support device in accordance with the identification information and instruction information input from the second processing circuit, and then performs feedback learning using the next instruction information input from the worker, which is input from the first processing circuit along with the same identification information. 【0094】 According to the configuration of embodiment 9, the learning model updated by feedback learning when a specific worker uses one work support device is also used when that specific worker uses another work support device. Therefore, even when that specific worker uses the other work support device for the first time, the work support device can be operated in a way that suits the worker's characteristics. 【0095】 (Aspect 10) A processing circuit according to any one of aspects 1 to 9, which identifies the worker from the input image or sound and acquires the identification information. 【0096】 According to the configuration of embodiment 10, special input operations for identifying the worker issuing voice instructions can be omitted, improving convenience and reducing hardware costs. 【0097】 (Aspect 11) A method for controlling a work support device, comprising: acquiring instruction information indicating instructions from an operator to the work support device; inputting the instruction information into a learning model that learns the operation of the work support device in response to the operator's instructions; acquiring control information used to control the work support device from the learning model; and outputting the control information. 【0098】 (Aspect 12) A control program that causes at least one processor to execute the above method. 【0099】1 Control system 2 Work support device 3 Controller 4 Server 5 Worker 6 Person being worked on 7 Mounting device 14 First actuator 15 Second actuator 20 Processing circuit 21 Processor 30 Processing circuit 41, 141 Learning model 42 Identification information receiving unit 45 Feedback learning unit N Communication network P1 Control program P2 Processing program

Claims

1. A processing circuit for controlling a work support device, comprising: acquiring instruction information indicating instructions from an operator to the work support device; inputting the instruction information into a learning model that learns the operation of the work support device in response to the operator's instructions; acquiring control information used to control the work support device from the learning model; and outputting the control information.

2. The processing circuit according to claim 1, wherein the processing circuit acquires identification information of the worker using the work support device, the learning model includes a learning model corresponding to the identification information, and the processing circuit inputs the instruction information to the learning model corresponding to the identification information.

3. The processing circuit according to claim 1, wherein the instruction information includes information relating to voice, information relating to body language, or information input from an input device.

4. The processing circuit according to claim 1, which is connected to the learning model via a communication network.

5. The processing circuit according to claim 1, which is attached to the work support device and comprises the learning model.

6. The processing circuit according to claim 1, wherein the work support device includes an actuator, and the control information includes the amount of operation of the actuator, the direction of operation of the actuator, or the operating speed of the actuator.

7. The processing circuit according to claim 1, wherein the learning model is updated by feedback learning using the next instruction information input from the operator after outputting the control information in response to the instruction information from the operator.

8. The processing circuit according to claim 7, wherein the learning model includes an initial model corresponding to the work support device, and the initial model is updated by the feedback learning.

9. The processing circuit according to claim 4, wherein the learning model receives input from a second processing circuit, which is different from the first processing circuit, which is the processing circuit, of identification information of a worker using a second work support device different from the first work support device, and instruction information of the worker to the second work support device, and after outputting control information of the second work support device in accordance with the identification information and instruction information input from the second processing circuit, it is updated by feedback learning using the next instruction information input from the first processing circuit along with the same identification information.

10. The processing circuit according to claim 1, which identifies the worker from the input image or sound and obtains the identification information.

11. A method for controlling a work support device, comprising: acquiring instruction information indicating instructions from an operator to the work support device; inputting the instruction information into a learning model that learns the operation of the work support device in response to the operator's instructions; acquiring control information used to control the work support device from the learning model; and outputting the control information.

12. A control program that causes at least one processor to execute the method described in claim 11.

Images