JP7208254B2 - Work optimization system and work optimization device - Google Patents

Description

The present invention relates to a work optimization system and a work optimization device, and is applied, for example, to a work optimization system and a work optimization device for optimizing work (collaborative work) performed by a device and a worker in cooperation. It is suitable for

The number of cases in which predetermined tasks are executed while working in an environment where equipment and workers cooperate is on the rise due to the development of robot technology and the need for automation due to labor shortages. In such a situation, it is required that the equipment should be mastered by a highly skilled worker by controlling the equipment, and that the operator should act in accordance with the optimized equipment.

As for the former, the actions of highly skilled workers are often difficult for unskilled workers to master, and it is conceivable that the same efficiency as skilled workers cannot be expected. Regarding the latter, there is concern about the burden of matching the behavior of the operator to the operation of the equipment, and the decrease in yield due to the decrease in concentration, etc., the physical load, and the mental load. etc. is a problem.

In this respect, a robot control method for appropriately assisting a person has been disclosed (see Patent Document 1).

JP 2017-30137 A

In the technique described in Patent Document 1, by controlling the robot so as not to impose a burden on the worker, a part of the above-mentioned problems can be solved, but the burden on the worker is reduced, and the worker's behavior is slowed down. There is a suboptimal problem.

The present invention has been made in consideration of the above points, and is intended to propose a work optimization system or the like capable of controlling the actions of workers while maximizing the efficiency of equipment.

In order to solve such a problem, in the present invention, a device control unit that controls the device based on a device control pattern of the device and parameters of the device control pattern, and the device control unit controls the device. a worker measurement unit that measures the work of the worker from sensor information of a worker who works in cooperation with the device when the device is being used, and the work measured by the worker measurement unit. based on a predetermined target value; and an evaluation unit that determines whether the evaluation by the evaluation unit is higher than a threshold value. and a behavior control pattern generator for generating behavior control patterns.

According to the above configuration, the action control pattern is generated based on the device control pattern. In other words, since action control is performed based on device control, for example, by using the device control pattern of an expert, it is possible to maximize the efficiency of the device and control the action of the worker. Become.

According to the present invention, operator behavior can be controlled while maximizing equipment efficiency.

It is an example of the configuration related to the work optimization system according to the first embodiment. 1 is a diagram showing an example of a working environment according to a first embodiment; FIG. FIG. 2 is a diagram illustrating an example of a configuration related to generation of device control patterns according to the first embodiment; FIG. FIG. 2 is a diagram showing an example of a model for optimizing device control according to the first embodiment; FIG. FIG. 7 is a diagram showing an example of a flowchart relating to device control pattern generation processing according to the first embodiment; FIG. 4 is a diagram showing an example of a model for optimizing behavior control based on device control according to the first embodiment; FIG. 7 is a diagram showing an example of a flowchart relating to action control pattern generation processing according to the first embodiment; FIG. 10 is a diagram showing an example of an index indicating a state related to work performed by a worker according to the first embodiment; FIG. 4 is a diagram showing an example of parameter index related information according to the first embodiment; FIG.

One embodiment of the present invention will be described in detail below with reference to the drawings.

TECHNICAL FIELD The present invention relates to a technology for optimizing a worker's behavior by controlling it. More specifically, in an environment where equipment and workers coexist, when a certain task is executed, the equipment control on the equipment side (optimal equipment control pattern) is used to optimally control the behavior of related workers (optimal control pattern). The present invention relates to a technology for generating an action control pattern to control a worker.

For example, in the present embodiment, work evaluation is performed based on behavior measurement data, which is data obtained by measuring the behavior of a worker, and target values, reward evaluation is performed based on the results of work evaluation and behavior control patterns, and work It is characterized by storing behavior control patterns with high evaluation, selecting behavior control patterns for optimizing the behavior of workers, and controlling equipment based on the selected behavior control patterns.

Further, for example, the present embodiment is characterized by learning an optimum device control pattern from behavior of a worker (for example, a skilled worker) who is regarded as a model.

Further, for example, in the present embodiment, by presenting equipment control patterns that have been acquired in advance to the worker, it is possible to approach the action pattern of an expert who is capable of performing optimal work. It is characterized by being able to control the behavior of

Further, for example, the present embodiment is characterized in that by defining device control patterns and action control patterns for each individual, it is possible to present the optimum work for each individual.

In the present embodiment, an example of collaborative work between a robot device and a worker in a factory environment will be described. Details will be described below.

(1) First Embodiment In FIG. 1, 100 generally indicates a work optimization system according to the first embodiment.

FIG. 1 is a diagram showing an example of the configuration of a work optimization system 100. As shown in FIG. The work optimization system 100 measures the behavior of the worker, the operation of the equipment, and the like, and optimizes the collaborative work between the equipment and the worker. The work optimization system 100 includes an imaging device 110, processing units (worker measurement unit 120, evaluation unit 130, target instruction unit 140, equipment measurement unit 150, behavior control pattern generation unit 160, behavior control planning unit 170, equipment control 180, etc.) and various information (device control pattern 191, action control pattern 192, etc.). Note that part of each component may be provided outside the work optimization system 100 .

The processing unit is, for example, realized as a function of a work optimization device (computer). Memory), ROM (Read Only Memory), HDD (Hard Disk Drive), communication device, and the like.

The function (processing unit) of the work optimization device may be implemented by, for example, having the CPU read out a program stored in the ROM into the RAM and execute it (software), or by hardware such as a dedicated circuit. may be implemented, or may be realized by combining software and hardware. Also, part of the functions of the work optimization device may be implemented by another computer that can communicate with the work optimization device.

The imaging device 110 is a video camera, a network camera, or the like. The photographing device 110 includes a lens, an image sensor, and the like, and can acquire a predetermined image (image information). The image sensor is a mechanism including an imaging element (none of which is shown) such as a CMOS (Complementary Metal Oxide Semiconductor) or a CCD (Charge Coupled Device). By using an image (video) acquired from such a photographing device 110, it is possible to sense the situation of a worker or the like in the work environment. Note that the work environment will be described later with reference to FIG.

For example, when the device control unit 180 controls the device, the worker measurement unit 120 obtains image information of the worker working in cooperation with the device acquired by the imaging device 110 (an example of a sensor). (an example of sensor information), the worker's work is measured.

The evaluation unit 130, for example, evaluates the work measured by the worker measurement unit 120 based on a predetermined target value (target value set by the target instruction unit 140). For example, the evaluation unit 130 determines whether or not the index indicating the state related to the worker's work exceeds the target value. In addition, the index indicating the state of the worker's work will be described later with reference to FIG. 8 .

The target instruction unit 140 sets, for example, a target value. Note that the target value is the amount of work to be achieved in a predetermined period (for example, moving boxes from a first location to a second location and stacking them in one day).

For example, when the device control unit 180 is controlling the device, the device measurement unit 150 measures the operation of the device from the image information of the device acquired by the imaging device 110 .

For example, the behavior control pattern generation unit 160 determines whether or not the evaluation by the evaluation unit 130 is higher than a threshold value, and if it is determined to be high, generates a behavior control pattern including parameters of the device control pattern determined to be high. do.

Here, when considering the phase of learning the behavior control pattern (learning phase) and the phase of using the behavior control pattern (operation phase), in the learning phase, for example, the behavior control planning unit 170 is not provided. may In this case, the action control pattern generation unit 160, for example, exploratoryly changes the parameter of the device control pattern being learned to another parameter so that the evaluation becomes higher, and the device control pattern and the changed other parameter are combined. is used to instruct the device control unit 180 to control the device. According to such a configuration, it becomes possible to appropriately obtain an optimum action control pattern. In addition, the work optimization system 100 may learn while operating without dividing into a learning phase and an operating phase. In this case, it becomes possible to obtain an optimum action control pattern while operating.

The action control planner 170 selects one action control pattern from the action control patterns generated by the action control pattern generator 160, for example. Further, for example, the behavior control planning unit 170 determines whether or not the index indicating the current state exceeds the threshold. may be used to specify an action control pattern whose index indicating the current state exceeds a threshold value, and change the selected action control pattern to the specified action control pattern.

The device control unit 180 specifies, for example, the device control pattern of the device and the parameters of the device control pattern based on the behavior control pattern selected by the behavior control planning unit 170, and controls the device.

FIG. 2 is a diagram showing an example of a working environment. It shows a bird's-eye view of a robot device 200 and a worker 201 working in the same environment. More specifically, the robot device 200 picks up the workpiece 202 and moves it to a target position, and the worker 201 appropriately supports the work that the robot device 200 cannot do before, during, or after the fact, It shows how they work together to complete a task. While observing their appearance with the photographing device 110, the actions and work situations between the robot device 200 and the worker 201 are measured using recognition technology.

For example, the robot device 200 is typically a 6-axis picking robot, but it may be a dual-arm type, an autonomously moving type, or a mobile body for transportation. There is no particular limitation as long as it has a function to support the work of the operator.

Various devices can be applied as the sensor of this embodiment, but the imaging device 110 will be described as an example.

In the work optimization system 100 , the worker measurement unit 120 measures the work (work state) of the worker 201 using the image information acquired from the imaging device 110 . The evaluation unit 130 evaluates the work. The evaluation index corresponds to a preset target value (for example, the number of movements of the workpiece 202 per minute) input to the target instruction unit 140 .

In addition, the device measurement unit 150 measures the motion (operation state) of the robot device 200 based on the image information obtained from the imaging device 110 . Here, information such as the posture and position of the robot device 200 directly obtained from the robot device 200 may be utilized. As a premise, the device control unit 180 transmits a robot control signal (device control signal) to the robot device 200 based on the device control pattern 191 learned in advance, and causes the robot device 200 to operate.

Here, the device control pattern 191 stores an optimum device control pattern generated by working with a skilled person, teaching, or the like. The device control pattern 191 is, for example, information that defines the operation of the device such as the trajectory of the arm of the robot device 200 . The machine control patterns 191 include a machine control pattern to move the work 202 by grasping one, a machine control pattern to move the work 202 by grasping two, a machine control pattern to move the work 202 to the front of the worker 201, and the work 202 to move. to the final destination. It should be noted that the device control pattern 191 is not limited to the one illustrated for the purpose of explanation.

FIG. 3 is a diagram showing an example of a configuration related to generation of the device control pattern 191. As shown in FIG. As shown in FIG. 3, the work optimization system 100 may be configured including an equipment control pattern generation section 300 as a processing section. The device control pattern generation unit 300 generates an optimal device control pattern 191 from, for example, a behavior of a worker as a model.

Regarding the generation of the device control pattern 191, it is assumed that an expert or the like decides, for example, the most efficient work in accordance with the target values presented by the target instruction unit 140 through trial and error with the expert. , there is a way to get it by machine learning method.

As a machine learning method, for example, a method called reinforcement learning can be used.

FIG. 4 is a diagram showing an example of a machine learning method (a model for optimizing device control). As shown in FIG. 4, when a simple model such as a device 400 and a work environment 401 is defined, the change (value) of the work environment 401 is evaluated by the action a of the device 400 and fed back as a reward r. This is the method of constructing the optimal action a with the reward r. Various such methods have been proposed, and a method such as Q-learning typified by "Watkins, C.J.C.H. (1989). Learning from Delayed Rewards." may be employed.

According to obtaining by making full use of these methods, work evaluation is performed for the target value set by the target instruction unit 140, and this work is evaluated as high (for example, close to the target value, higher than the target value high, etc.) can be acquired as a role on the equipment side for a highly skilled worker. Next, a method of generating a device control pattern will be explained using FIG.

FIG. 5 is a diagram showing an example of a flowchart relating to processing (device control pattern generation processing) for the device control pattern generation unit 300 to generate the device control pattern 191 .

First, the evaluation unit 130 acquires action data (work) obtained by the worker measurement unit 120 using data obtained from the photographing device 110 and a predetermined target value (target value 500) (step S501). .

Subsequently, the evaluation unit 130 evaluates (behavior evaluation) the work measured by the worker measurement unit 120 based on the target value (step S502).

Here, the work (behavior) includes the distance traveled by the worker, squatting which puts a heavy load on the body, a holding posture in a high position, and stressful work. In behavioral evaluation, assuming product stuffing work, how much time was required for each case and how much physical load (physical impossibility) was applied was evaluated. It is possible to define an evaluation criterion (an index indicating the state of a worker's work) such that an action with a low evaluation is highly evaluated. Since the evaluation criteria depend on the installation environment, etc., it is preferable to define them according to the work environment. It should be noted that, in addition to the imaging device 110, other devices such as a stress measurement sensor and an electroencephalogram sensor can be used to measure the physical load. Furthermore, it is good to take the form which utilized sensitivity evaluation, such as questionnaire evaluation. In addition, the operations are not limited to those exemplified for explanation. Behavioral evaluation is not limited to the examples given for explanation.

Subsequently, the device control pattern generation unit 300 determines whether or not to end the device control pattern generation process (step S503). The device control pattern generation unit 300 shifts the process to step S505 when determining to end, and shifts the process to step S504 when determining not to end. Termination is determined according to predetermined criteria, such as no change in action evaluation value, execution of a predetermined number of times, error from target value less than a threshold value, and the like.

In step S504, the device control pattern generation unit 300 updates, as an ideal device control pattern, the states of devices that have obtained evaluations higher than the threshold as a result of behavior evaluation. It should be noted that the equipment control pattern generation unit 300 repeats successive evaluations for improvements made by the equipment and/or the worker through trial and error toward the target value 500 .

In step S<b>505 , the device control pattern generation unit 300 saves the device control pattern optimal for the expert (for example, one or more device control patterns exceeding the target value and/or threshold value) as the device control pattern 191 .

Note that the device control patterns 191 may be narrowed down based on constraint conditions (safety viewpoint, maintenance viewpoint, etc.).

Next, a configuration for implementing action control based on device control will be described with reference to FIGS. 6 and 7. FIG. As described above, in the learning of the device control pattern 191, the method of learning how to operate the device to achieve efficient movement has been described with respect to the purpose such as work efficiency. This is, so to speak, a task for the equipment side to learn the control method of the equipment suitable for the expert.

On the other hand, in the evaluation unit 130 and the behavior control pattern generation unit 160, based on the learned behavior of the equipment, how the worker who performs the work learns efficient behavior and what kind of equipment is used. If the movement is executed, it becomes a procedure to find out whether the combination of the worker and the equipment is a highly efficient movement.

FIG. 6 is a diagram showing an example of a model for optimizing behavior control based on device control. First, the device ₆₀₀ performs an action am as a predetermined operation. Then, in response to this, the worker 602 performs the action _aw in concert. On the other hand, a state change s occurs in the work environment 601 .

Considering the worker 602 as the center, what kind of action am of the device ₆₀₀ , what kind of action _aw of the worker 602 occurred, and what change s occurred in the work environment 601 is the action Assuming that a _m is a reward r, it can be considered in the framework of inverse reinforcement learning how the worker 602 approaches the ideal change s with what kind of reward r (action a _m ).

Inverse Reinforcement Learning is a method of estimating rewards (which state is better) from the actions of an expert. . What kind of reward (accuracy of action) is judged from the actions of an expert who can perform ideal work and what kind of judgments and actions are caused by the maturity of the expert? is to learn.

In the present embodiment, instead of directly utilizing inverse reinforcement learning such that actions generated by experts are used as action guidelines for inexperienced people, optimal device control on the device side defined by experts is performed. Predicting, through patterns, which action am can be the optimal _reward for the worker 602 when these machine control patterns, that is, action am, are given to the worker 602 (inexperienced worker, _unskilled worker, etc.). , which can be rephrased as controlling actions through the device 600 .

As for the effect of direct instruction, there are many cases in which the instruction does not match the individual, and it takes time to learn and the behavior is not optimal for the individual. By encouraging ideal behavior, the operator's behavior is optimized, and at the same time, the control on the equipment side evolves and adapts.

Furthermore, it can be said that an effect such as mutual evolution can be obtained by repeating the learning of device control on the device side and the learning of action control on the worker side.

Here, in the work optimization system 100, the action _am is changed by changing the parameter of the equipment control pattern 191 (the parameter for tuning the method of moving the equipment such as speed), and the equipment control pattern 191 at that time is changed. and parameters are used as behavior control patterns. Next, a method of generating an action control pattern will be described with reference to FIG.

FIG. 7 is a diagram showing an example of a flowchart relating to processing (behavior control pattern generation processing) in which the behavior control pattern generation unit 160 generates the behavior control pattern 192 .

First, the evaluation unit 130 acquires action data (work) obtained by the worker measurement unit 120 using data obtained from the photographing device 110 and a predetermined target value (target value 700) (step S701). . Note that the target value 500 and the target value 700 may be the same or different.

Subsequently, the evaluation unit 130 evaluates the work measured by the worker measurement unit 120 based on the target value (behavior evaluation) (step S702). For example, the evaluation unit 130 performs evaluation in the same manner as in step S502 and compares the superiority or inferiority of actions.

Subsequently, the action control pattern generation unit 160 determines whether or not to end the action control pattern generation process (step S703). If the action control pattern generation unit 160 determines to end the process, it moves the process to step S705, and if it determines not to end the process, it moves the process to step S704. Termination is determined according to predetermined criteria, such as no change in action evaluation value, execution of a predetermined number of times, error from target value less than a threshold value, and the like.

In step S704, the behavior control pattern generation unit 160 updates the behavior control pattern, which has been evaluated higher than the threshold as the result of the behavior evaluation, as an ideal behavior control pattern. For example, the behavior control pattern generation unit 160 generates a behavior control pattern including a device control pattern determined to be higher than a threshold value (which may be an identifier by which the device control pattern can be identified) and a parameter.

Note that the behavior control pattern generation unit 160 repeats the evaluation sequentially for those with changed behavior control patterns (parameters of device control patterns that enable the behavior control of the worker). For example, in the case of a behavior control pattern in which the movement speed of the equipment arm is the first speed, the workpieces are carefully stacked, but the time required for movement is long, and the movement speed of the equipment arm is the second speed, which is faster than the first speed. In the case of the action control pattern with a speed of , the workpieces are piled up roughly, but the time required for movement is short. In the former case, sufficient work check and adjustment time is required while the worker is waiting, but efficiency is low, and monotonous work induces mistakes. Sometimes not. On the other hand, in the latter case, although the speed is very high, it is also possible to work efficiently for workers who are highly adaptable to such work.

The above describes a very simple case. Here, some action control patterns are considered to be general device accuracy such as speed and accuracy, but different action control patterns are also conceivable. In general, there is a tendency to work in a precise rhythm ideal for factory machines, but by introducing random fluctuations and noise-like movements, it is possible to call attention to the workers and diversify the work. It can also be expected to prevent stress and loss of concentration by giving. On the other hand, there are cases where such work is not desired.

Furthermore, a more appropriate action control pattern can be applied by taking into account information related to the worker, such as the time zone of the work, the duration of the work, the current environment, and the state of the body. For example, if there is advance information that the accuracy of work will be lower in the afternoon due to fatigue than in the morning, it is conceivable to appropriately control the work speed.

In this way, inverse reinforcement learning is performed to determine which action control pattern is a reward (a pattern that leads to a good state) for the worker.

Further, in step S704, the action control pattern generation unit 160 converts information (parameter index related information) associated with the parameter of the device control pattern used for controlling the device and the index indicating the state related to the work of the worker into the action control pattern. may be stored in the HDD in association with . A physical strength index, a mental index, and the like are provided as indices that indicate the worker's working condition.

FIG. 8 is a diagram showing an example of indicators (indicator table 800) indicating the status of work performed by a worker. As shown in FIG. 8, an index table 800 stores information about what kind of action plan is effective for a certain device control pattern. This differs depending on the type of business and is defined for each device control pattern.

More specifically, the index table 800 stores information relating to indices such as a proficiency level 802, movement amount 803, concentration 804, precision 805, stress 806, and work performance 807 for each device control pattern 801. there is Note that the index table 800 may include information related to workers as described above.

Also, FIG. 9 shows an example of the parameter index related information (approximate curve 900). Note that the parameter index related information is generated for each index and parameter.

For example, when the worker gets tired, the approximate curve 900 of the movement amount 803 can be used to change the parameters within the range that satisfies the threshold value. Further, for example, when the work of the worker becomes rough, the approximate curve 900 of the precision 805 can be used to change the parameters within the range that satisfies the threshold value. Also, for example, when the daily quota is changed, the approximated curve 900 of the work record 807 can be used to change the parameters within a range that satisfies the threshold value. In this way, according to the parameter index related information, it becomes possible to change the parameter within a range that satisfies the threshold according to the current state.

In step S<b>705 , the action control pattern generation unit 160 saves the optimum action control pattern for the worker as the action control pattern 192 .

Here, the action control pattern is different for each worker, and the reward is learned as to what kind of movement of the equipment will be the optimum work when executing a certain work.

According to the above-described configuration, for example, by acquiring the optimum device control pattern of the device and utilizing the device control pattern for behavior control of another worker, the performance of the device can be sufficiently brought out. In addition, the work can be optimized for each individual, and the overall efficiency can be maximized.

In addition, according to the above-described configuration, for example, the device side can also optimize the device control, so it is possible to generate the optimum action control pattern for the combination of the device and the worker and maximize the efficiency.

In addition, according to the above-described configuration, for example, it is possible to acquire the optimum device control pattern and also to control human behavior, so that it is possible to make the worker master the work in a manner similar to that of an expert.

Further, according to the above-described configuration, for example, since the action control pattern is an item related to personal preference or action pattern, the action control pattern and/or the parameter index-related information generated by one device can be used by another device. Ease of deployment, such as being able to be used for

As described above, according to the present embodiment, the action of an expert as an example of the norm and the device control pattern with the highest target such as work efficiency are acquired in advance, and the optimal work such as that of an unskilled person is obtained. can not be executed immediately, by guiding the work of unskilled workers to the work of skilled workers through equipment control patterns obtained in advance, as a result, the behavior of workers is optimized, and production targets and high efficiency are achieved. It is possible to obtain the effect of being close to a work such as Also, for example, by providing an index indicating the state of the worker, it is possible to obtain the effect of performing action control suitable for the individual. Also, for example, it is possible to obtain the effect of performing action control according to the current state.

In this way, based on machine control patterns that have been learned in advance, a work optimization system that can encourage the worker's proficiency level by changing the machine side by performing action evaluation and reward learning for the worker, and It is possible to provide a work optimization device.

In the present embodiment, the form of repetitive mutual deepening, such as updating of device control patterns, is not illustrated, but a repetitive configuration may of course be employed to enable mutual evolution.

(2) Other Embodiments In the above-described embodiments, the case where the present invention is applied to a work optimization system was described, but the present invention is not limited to this, and various other systems, It can be widely applied to devices, methods and programs.

In the above description, information such as programs, tables, and files that implement each function is stored in a memory, hard disk, SSD (Solid State Drive), or other storage device, or recorded on an IC card, SD card, DVD, or the like. You can put it on the medium.

Moreover, the above-described configurations may be appropriately changed, rearranged, combined, or omitted within the scope of the present invention.

According to the configuration described above, it is possible to control the actions of the operator while maximizing the efficiency of the equipment.

100 work optimization system 110 imaging device 120 worker measurement unit 130 evaluation unit 140 target instruction unit 150 equipment measurement unit 160 action control pattern generation unit , 170... action control plan part, 180... device control part, 191... device control pattern, 192... action control pattern.

Claims (2)

a device control unit that controls the device based on a device control pattern that controls the operation of the device and parameters of the device control pattern;
Worker measurement for measuring behavior of a worker from image information of a worker working in cooperation with the device acquired by a photographing device when the device is controlled by the device control unit. Department and
an evaluation unit that evaluates the behavior of the worker measured by the worker measurement unit based on a predetermined target value;
Determining whether or not the evaluation by the evaluation unit is higher than a threshold value, and generating an action control pattern for controlling the action of the worker, including the parameter of the device control pattern determined to be high when it is determined to be higher. a behavior control pattern generator that
An action control plan for selecting one action control pattern from among the action control patterns of the worker generated by the action control pattern generation unit based on the result of reward evaluation for evaluating the correctness of the action of the worker. Department and
with
The action control pattern generation unit changes the parameter of the device control pattern to another parameter based on a predetermined approximate curve, and controls the device using the device control pattern and the changed other parameter. giving an instruction to the device control unit to
The device control unit specifies a device control pattern of the device and a parameter of the device control pattern based on the behavior control pattern selected by the behavior control planning unit, and controls the device;
The action control pattern generation unit associates parameter index related information, which is information that associates a parameter of the device control pattern of the device with an index indicating a state related to the work of the worker, with the action control pattern including the parameter. and remember
The behavior control planning unit determines whether or not the index indicating the current state exceeds the threshold. identifying an action control pattern exceeding the value and changing the selected action control pattern to the identified action control pattern;
A work optimization system characterized by: a device control unit that controls the device based on a device control pattern that controls the operation of the device and parameters of the device control pattern;
Worker measurement for measuring behavior of a worker from image information of a worker working in cooperation with the device acquired by a photographing device when the device is controlled by the device control unit. Department and
an evaluation unit that evaluates the behavior of the worker measured by the worker measurement unit based on a predetermined target value;
Determining whether or not the evaluation by the evaluation unit is higher than a threshold value, and generating an action control pattern for controlling the action of the worker, including the parameter of the device control pattern determined to be high when it is determined to be higher. a behavior control pattern generator that
An action control plan for selecting one action control pattern from among the action control patterns of the worker generated by the action control pattern generation unit based on the result of reward evaluation for evaluating the correctness of the action of the worker. Department and
with
The action control pattern generation unit changes the parameter of the device control pattern to another parameter based on a predetermined approximate curve, and controls the device using the device control pattern and the changed other parameter. giving an instruction to the device control unit to
The device control unit specifies a device control pattern of the device and a parameter of the device control pattern based on the behavior control pattern selected by the behavior control planning unit, and controls the device;
The action control pattern generation unit associates parameter index related information, which is information that associates a parameter of the device control pattern of the device with an index indicating a state related to the work of the worker, with the action control pattern including the parameter. and remember
The behavior control planning unit determines whether or not the index indicating the current state exceeds the threshold. identifying an action control pattern exceeding the value and changing the selected action control pattern to the identified action control pattern;
A work optimization device characterized by:

Images