Learning device, learning method, and program

Abstract

This enables the automatic generation of operational plans that increase the number of autonomous mobile devices that can operate within a given space at the same time. [Solution] The operation plan generation device 1 includes a control unit 1a that generates an operation plan including multiple waypoints for an autonomous mobile device in response to an operation plan generation request including operation preference information, and a storage unit 1b. The storage unit 1b stores the relationship between the executed operation plan and the actual operation results at the time of execution. The control unit 1a generates candidate operation plans including multiple waypoints based on the operation preference information, and predicts the positional and temporal deviations that may occur if the operation is carried out according to the candidate operation plan based on the above relationship. Based on the prediction results, the control unit 1a sets an exclusive area for the candidate operation plan that prohibits competition with other autonomous mobile devices. The control unit 1a determines the candidate operation plan that does not compete with operation plans that have been generated but are not yet completed for other autonomous mobile devices as the response operation plan.

Description

Technical Field

[0001] The present disclosure relates to a learning device, a learning method, and a program.

Background Art

[0002] In recent years, research on devices that move autonomously (autonomous mobile devices) such as drones and self-driving vehicles has been actively conducted, and some have been actually used. In addition, due to the decrease in the working population and cost reduction and work efficiency improvement, further spread of autonomous mobile devices is expected in the future. For example, it is expected that tasks such as luggage transportation, delivery, and distribution will be replaced from humans and automobiles driven by humans to autonomous mobile devices, and autonomous mobile devices will be used in scenarios closely related to people's lives. Then, since autonomous mobile devices will be operated in people's living spaces, it is required that autonomous mobile devices be operated safely, and stable operation such as reliable operation and compliance with arrival times is also required.

[0003] Patent Document 1 describes a traffic control support system having a storage unit, an instruction responsiveness estimation unit, a geographic E-map generation unit, and a movement prediction unit for the purpose of improving the control accuracy of a ship's route while suppressing the calculation amount. The storage unit has action data regarding the actions of a moving object, geographic data in which geographic attribute information, which is information regarding the movement reference of the moving object, is given to each section obtained by dividing a map into meshes, and responsiveness data regarding the responsiveness of the moving object to an instruction. The instruction responsiveness estimation unit estimates the ideal action of the moving object based on the responsiveness data, calculates the difference between the estimated action and the action of the action data, and updates the responsiveness data based on the calculation result. The geographic E-map generation unit estimates the probability that the moving object exists at a certain coordinate at each time based on the action data, the geographic data, and the responsiveness data, and generates a geographic E-map. The movement prediction unit predicts the future coordinates of the moving object based on the geographic E-map.

Prior Art Documents

Patent Documents

[0004] [Patent Document 1] Japanese Patent Publication No. 2018-36958 [Overview of the Initiative] [Problems that the invention aims to solve]

[0005] The inventors of this invention considered how to operate autonomous mobile devices safely and stably. First, autonomous mobile devices are often operated by moving along a given route and for a set duration, meaning they autonomously control their path to stay in line with the planned route. Even if disturbances occur during movement, the autonomous mobile device will autonomously correct its trajectory to stay in line with the planned route and duration. For example, for devices that operate in the air, such as drones, or devices that move on the sea, such as ships, a typical disturbance that could cause them to deviate from their route is wind. Even with disturbances such as wind, the autonomous mobile device will autonomously control its body to stay as close to the planned route as possible.

[0006] A typical example of an autonomous mobile device is a drone that flies autonomously beyond the line of sight. Autonomous drones can move along a given path by making their own decisions, rather than receiving step-by-step instructions from an operator. However, while autonomous mobile devices possess a certain degree of autonomous navigation capability, it is desirable that they be monitored remotely via communication from a flight management system to determine their current position and ensure safe operation, in order to respond to unforeseen circumstances. Monitoring the flight path should also be one of the functions of the flight management system.

[0007] Furthermore, some autonomous mobile devices are equipped with sensors and other features to avoid collisions with each other. However, since they are used in urban areas, it is unacceptable for any collisions to occur between autonomous mobile devices. Therefore, even when multiple autonomous mobile devices are operating simultaneously, it is desirable to adjust and plan their routes and times in advance to prevent collisions. The adjusted and planned routes and times can be called an operation plan, and the formulation of such an operation plan is one of the functions of the operation management system.

[0008] Our investigation into methods for formulating operational plans for autonomous mobile devices revealed two main problems. The first problem is that there is no automated method for designing routes and generating operational plans that coordinate with the routes of other autonomous mobile devices simply by setting the start and end points; this process is currently performed manually via the web. While manual adjustments to operational plans are possible when there are few drones in operation and the operational density in a given airspace is not high, this process becomes unsustainable as the number of operating devices increases.

[0009] The second problem is that if flight plans are created through manual procedures, the spatial and temporal efficiency of use will be poor, making it impossible to efficiently operate a large number of autonomous mobile devices in the same space. For example, one could consider designing flight paths and collision detection using a uniformly fixed width path (protected airspace), but even with such a uniform protected airspace in mind, it is difficult to say that efficient operation is possible.

[0010] To explain in more detail, as mentioned above, even if an autonomous mobile device has a predetermined route, it may temporarily deviate from the planned route due to disturbances or other factors. Therefore, it is generally advisable to design the route with a large margin to avoid collisions with other autonomous mobile devices. For example, in the case of an autonomous mobile device flying in the air, such as a drone, in the horizontal plane, it is advisable to leave a distance margin large enough so that even if it is blown off course by the wind at the maximum wind speed permitted for operation, it will not stray into other operating areas. Furthermore, this large margin also needs to be secured in terms of time. For example, in the case of an autonomous mobile device flying in the air, such as a drone, collisions will occur if multiple drones enter the same area in the same space at the same time, so the time that a drone can occupy a certain space needs to be long enough to account for delays. However, if these margins are too large, the efficiency of space utilization will decrease, and the number of autonomous mobile devices that can operate simultaneously in a given space at the same time will be limited.

[0011] Due to these issues, when formulating operational plans for autonomous mobile devices, it is necessary to increase the number of autonomous mobile devices that can operate simultaneously within a given space and time period, and to automate the process of formulating such plans.

[0012] On the other hand, the technology described in Patent Document 1 is designed with human-operated vessels in mind and allows multiple vessels to occupy a single large mesh section. This is because, since humans are on board, even if a collision is likely to occur within the section, it can be avoided by human judgment. Therefore, it is desirable to automatically generate an operation plan that increases the number of autonomous mobile devices that can operate simultaneously within a certain space at the same time, without requiring human judgment to avoid collisions.

[0013] The purpose of this disclosure is to provide a flight plan generation device, a learning device, and methods and programs thereof that can automatically generate a flight plan that increases the number of autonomous mobile devices that can operate within a certain space at the same time, in order to solve the above-mentioned problems. [Means for solving the problem]

[0014] The operation plan generation device according to the first aspect of the present disclosure comprises a control unit that, in response to an operation plan generation request including operation preference information, generates an operation plan including multiple passing points for an autonomous mobile device included as an operation target in the operation preference information, and a storage unit that stores the relationship between an executed operation plan, which is an executed operation plan, and the actual operation results at the time of execution of the executed operation plan. The control unit generates candidate operation plans including multiple passing points based on the operation preference information, predicts positional and temporal deviations that may occur when operating according to the candidate operation plan based on the relationship, sets an exclusive area as an area where other autonomous mobile devices are prohibited from operating in competition with the candidate operation plan based on the prediction results, and determines a candidate from among the candidate operation plans that does not conflict with an operation plan that has been generated for another autonomous mobile device but is not yet completed, as the operation plan to respond to the operation plan generation request.

[0015] A learning device according to a second aspect of this disclosure includes an acquisition unit that, in response to a request for generation of an operation plan including operation request information, acquires information including an executed operation plan that has already been executed, which is generated to include multiple waypoints for an autonomous mobile device included as an operation target in the operation request information, and the operation results at the time of execution of the executed operation plan; and a control unit that generates a trained model that predicts the positional and temporal deviations that may occur when operating according to an operation plan candidate generated to include multiple waypoints based on other operation request information, based on the information acquired by the acquisition unit. The operation plan is a plan in which, for each section route connecting adjacent waypoints in the multiple waypoints included in the operation plan, an exclusive area is set as an area that prohibits other autonomous mobile devices from operating in competition by surrounding the section route.

[0016] A third aspect of the present disclosure provides a method for generating an operation plan, which, in response to an operation plan generation request including operation preference information, generates an operation plan including multiple passing points for an autonomous mobile device included as an operation target in the operation preference information. The process generates candidate operation plans including multiple passing points based on the operation preference information, predicts positional and temporal deviations that may occur when operating according to the candidate operation plan based on the relationship between the executed operation plan (an executed operation plan) and the actual operation results at the time of execution of the executed operation plan, sets an exclusive area for the candidate operation plan as an area where other autonomous mobile devices are prohibited from operating in competition with it based on the prediction results, and determines a candidate among the candidate operation plans that does not conflict with an operation plan that has been generated for another autonomous mobile device but is not yet completed, as the operation plan to respond to the operation plan generation request.

[0017] A learning method relating to a fourth aspect of this disclosure responds to a request for the generation of a route plan including route preference information by acquiring information including an executed route plan that has already been executed, which is generated to include multiple waypoints for an autonomous mobile device included as a target for operation in the route preference information, and the actual operation results at the time of execution of the executed route plan, and generates a trained model that predicts the positional and temporal deviations that may occur when operating according to a candidate route plan that is generated to include multiple waypoints based on other route preference information, based on the acquired information. The route plan is a plan in which an exclusive region is set for each section route connecting adjacent waypoints in the multiple waypoints included in the route plan, which is a region that prohibits other autonomous mobile devices from operating in competition by surrounding the section route.

[0018] A fifth aspect of the present disclosure is a program that causes a computer to perform a process to generate an operation plan including operation request information, for an autonomous mobile device included as an operation target in the operation request information, the process being to generate an operation plan including multiple passing points, the process being to generate candidate operation plans including multiple passing points based on the operation request information, the process being to predict positional and temporal deviations that may occur when operating according to the candidate operation plan based on the relationship between the executed operation plan, which is an executed operation plan, and the actual operation results at the time of execution of the executed operation plan, the process being to set an exclusive area for the candidate operation plan as an area where other autonomous mobile devices are prohibited from operating in competition with it based on the prediction results, and the process being to determine the candidate operation plan that does not conflict with an operation plan that has been generated for another autonomous mobile device but is not yet completed, as the operation plan to respond to the operation plan generation request.

[0019] A program relating to a sixth aspect of this disclosure is a program that causes a computer to perform a process in response to a request to generate an operation plan including operation preference information, by acquiring information including an executed operation plan that has already been executed, which has been generated to include multiple waypoints for an autonomous mobile device included as an operation target in the operation preference information, and the operation results at the time of execution of the executed operation plan, and generating a trained model that predicts the positional and temporal deviations that may occur when operating according to an operation plan candidate that has been generated to include multiple waypoints based on other operation preference information, based on the acquired information. The operation plan is a plan in which an exclusive area is set for each section route connecting adjacent waypoints in the multiple waypoints included in the operation plan, which is an area that prohibits other autonomous mobile devices from operating in competition by surrounding the section route. [Effects of the Invention]

[0020] This disclosure provides an operation plan generation device, a learning device, and methods and programs thereof that can automatically generate an operation plan that increases the number of autonomous mobile devices that can operate within a certain space at the same time.

Brief Description of Drawings

[0021] [Figure 1] It is a block diagram showing a configuration example of an operation plan generation device according to Embodiment 1. [Figure 2] It is a block diagram showing a configuration example of an autonomous mobile device that is an operation target in an operation plan generated by the operation plan generation device of FIG. 1. [Figure 3] It is a flowchart for explaining a processing example in the operation plan generation device according to Embodiment 1. [Figure 4] It is a block diagram showing a configuration example of a learning device that generates a learned model that can be used in the operation plan generation device according to Embodiment 1. [Figure 5] It is a flowchart for explaining a processing example in the learning device of FIG. 4. [Figure 6] It is a schematic diagram showing a configuration example of an operation management system according to Embodiment 2. [Figure 7] It is a block diagram showing a configuration example of an operation management device according to Embodiment 2. [Figure 8] It is a schematic diagram for explaining section management according to a comparative example. [Figure 9] It is a schematic diagram for explaining an example of management in the operation management device of FIG. 7. [Figure 10] It is a conceptual diagram for explaining the relationship between waypoints, edges, and path deviations, as well as exclusive areas. [Figure 11] It is a diagram showing an example of the data format of an operation plan issuance request. [Figure 12] It is a diagram showing an example of the data format of the data held in the operation plan issuance request holding unit for the request of FIG. 11. [Figure 13] It is a diagram showing an example of the data format of an operation plan candidate created during operation plan formulation. [Figure 14] It is a diagram showing an example of the data format of an operation plan issued for the operation plan candidate of FIG. 13. [Figure 15] It is a diagram showing an example of the data format of an issued operation plan. [Figure 16] Figure 15 shows an example of a data format where multiple issued operation plans are displayed side by side. [Figure 17] This figure shows an example of the data format for waypoint coordinate information. [Figure 18] This figure shows an example of the data format for edge information. [Figure 19] Figure 7 is a sequence diagram showing an example of operation in the operation management system. [Figure 20] This is a sequence diagram following Figure 19. [Figure 21] This is a flowchart illustrating an example of the registration process for requesting the issuance of a train schedule. [Figure 22] This figure shows an example of the data format for parameters in a flight plan request. [Figure 23] This is a flowchart illustrating an example of the process for determining the spatial information acquisition range. [Figure 24] This is a schematic diagram showing an example of the spatial information acquisition range. [Figure 25] This figure shows an example of the data format for spatial information acquisition request parameters. [Figure 26] This is a flowchart illustrating an example of the process for determining the spatial information acquisition range. [Figure 27] This figure shows an example of the data format for acquired spatial information. [Figure 28] This is a schematic diagram illustrating an example of acquired spatial information. [Figure 29] This is a flowchart illustrating an example of the process for formulating candidate operation plans. [Figure 30] This is a flowchart following Figure 29. [Figure 31] This figure shows an example of the data format for predicted values. [Figure 32] This is a schematic diagram to explain the prediction results. [Figure 33] This figure shows an example of the data format of the calculated exclusive area. [Figure 34] This figure shows an example of the data format for the scope of issued operation plans. [Figure 35] Figure 34 is a schematic diagram showing an example of the spatial information acquisition range determined in Figure 34. [Figure 36] This figure shows an example of the data format for candidate operation plans, with the result of calculating the exclusive area of ​​the candidate operation plan added. [Figure 37] This is a schematic diagram illustrating an example of a race condition. [Figure 38] This is a schematic diagram illustrating another example of a competitive state. [Figure 39] This is a schematic diagram illustrating another example of a competitive state. [Figure 40] This is a diagram to explain the competition for domains. [Figure 41] This figure shows an example of the format of operation plan data. [Figure 42] This figure shows an example of the format of operation plan data. [Figure 43] This figure shows an example of the format of operation plan data. [Figure 44] This figure shows an example of the device's hardware configuration. [Modes for carrying out the invention]

[0022] The embodiments will be described below with reference to the drawings. In the embodiments, the same or equivalent elements may be denoted by the same reference numerals, and redundant explanations will be omitted as appropriate.

[0023] <Embodiment 1> The operation plan generation device according to Embodiment 1 will be described with reference to Figures 1 and 2. Figure 1 is a block diagram showing an example configuration of the operation plan generation device according to Embodiment 1, and Figure 2 is a block diagram showing an example configuration of an autonomous mobile device that is the target of operation in the operation plan generated by the operation plan generation device of Figure 1.

[0024] As shown in Figure 1, the operation plan generation device 1 according to this embodiment comprises a control unit 1a that controls the entire device and a storage unit 1b, and generates operation plans for autonomous mobile devices. Operation plans are usually generated for multiple autonomous mobile devices, but it is sufficient for the operation plan to be generated in response to an operation plan generation request for a particular autonomous mobile device, and it is not necessary to generate operation plans for all managed autonomous mobile devices at the same time.

[0025] Before describing the configuration of the operation plan generation device 1, we will first describe an example configuration of an autonomous mobile device that is the target of operation in the generated operation plan. As shown in Figure 2, the autonomous mobile device (autonomous mobile equipment) 2 that is the target of operation in the operation plan may be equipped with a plurality of sensors (sensor groups) 2a, a communication unit 2b, a movement control unit 2c, and a drive unit 2d.

[0026] The autonomous mobile device 2 possesses information processing capabilities and can control actuators and other components to move autonomously, thus encompassing all types of equipment responsible for transportation and movement.

[0027] Examples of autonomous mobile devices 2 include self-driving cars, self-driving trains, self-navigating ships, self-flying aircraft, drones, AGVs (automated guided vehicles) used in factories and warehouses, robots with driving or leg-based mobility functions, robotic electric wheelchairs, and motorcycles. Here, drones are not limited to aircraft such as unmanned aerial vehicles that fly in the air, but can operate in any environment, including on the ground, in the air, on water, underwater, and in space. Robots can move in any way, including by running or walking. Furthermore, examples of autonomous mobile devices 2 include forklifts, construction vehicles such as construction machinery and heavy equipment, railway vehicles, vehicles used for logistics such as taxis and trucks, police vehicles, and fire vehicles. Here, railway vehicles can move in any way, including light rail, steel-wheeled vehicles, new transit systems, monorails, and magnetic levitation vehicles. Furthermore, autonomous mobile device 2 also includes devices that can switch between modes, sometimes operating autonomously and sometimes operating under human instruction or control, as well as devices that move in coordination with human operation or instruction for part of their autonomous operation. In other words, autonomous mobile device 2 only needs to be capable of moving autonomously, and may also have the function of moving according to control from an external controller or according to the operation of a rider.

[0028] Sensor group 2a consists of multiple sensors installed at various locations on the autonomous mobile device 2. The results detected by each sensor are passed to the mobile control unit 2c and can be used to control its movement. In other words, the autonomous mobile device 2 is a device capable of moving autonomously, and can detect its state (current state) while moving (operating) based on the detection results (sensor data) output from sensor group 2a. Here, the current state can refer to the sensor value obtained by measuring the detection target at the installation location for each sensor included in sensor group 2a, or the state indicated by that sensor value, and mainly includes the movement state (operation state). The above movement state may include the tilt and position of the autonomous mobile device 2. Note that the above installation locations and detection targets will generally differ depending on the type and shape of the autonomous mobile device 2. Furthermore, sensor group 2a can consist of multiple sensors necessary for controlling the autonomous mobile device 2, and can include not only a group of sensors that collect the state of the autonomous mobile device 2 itself, but also a group of sensors that collect the state of the surroundings (surrounding environment).

[0029] An example of sensor group 2a will be explained using a drone, a type of autonomous mobile device, as an example. Sensor group 2a may include a velocity sensor to observe the aircraft's speed, an acceleration sensor to observe the aircraft's acceleration, a gyroscope sensor to observe the aircraft's roll, pitch, and yaw, and a rotation sensor to observe the rotation speed of each rotor used for flight. Sensor group 2a may also include a barometric pressure sensor or altitude sensor to observe the aircraft's altitude, a geomagnetic sensor to observe the direction from the aircraft, and a temperature sensor to detect heat generation from circuits and rotors. Of course, sensor group 2a does not need to include all of these examples; it may include only some, and it may also include other types of sensors not limited to these examples.

[0030] For example, other sensors included in sensor group 2a include sensors that acquire position information in GNSS (Global Navigation Satellite System). GNSS is a general term for satellite positioning systems such as GPS (Global Positioning System), GLONASS (Global Navigation Satellite System), Galileo, and the Quasi-Zenith Satellite System. Here, the Quasi-Zenith Satellite System is QZSS (Quasi-Zenith Satellite System). With this sensor, the autonomous mobile device 2 can receive GNSS signals, determine its position in space, and obtain position information. Furthermore, the sensors included in sensor group 2a can also be optical cameras or stereo cameras, and the position and attitude of the device can be determined by processing the images acquired by the cameras. Additionally, the sensors included in sensor group 2a can also be acoustic sensors, ultrasonic sensors, 2D-LiDAR, 3D-LiDAR, etc. In this case, the autonomous mobile device 2 can process waveform data and point cloud data obtained from sensing devices using sound waves and lasers to determine its own position, attitude, and surrounding conditions. The information obtained in this way can also be called sensor information. Therefore, sensors that require information processing in this manner can also be included in the sensor group 2a.

[0031] Furthermore, the operating mode of the autonomous mobile device 2 (such as the mode's On / Off status or the current mode) and other mounted devices (such as the status of the rotor, or log data from the control device used for autonomous control) can also be considered sensor information, and therefore, devices that provide such information can also be included in the sensor group 2a. Here, the status of the rotor can be defined as enabled / disabled, normal / abnormal, rotational speed, output, temperature, etc.

[0032] The communication unit 2b is the part that communicates with external devices such as the operation plan generation device 1, receives operation plan data generated by the operation plan generation device 1, and passes it to the mobile control unit 2c. Depending on the type of autonomous mobile device 2, the timing of data communication with the outside, etc., the communication unit 2b can be at least one of a wireless communication unit and a wired communication unit.

[0033] The movement control unit 2c is a control unit that controls the movement of the autonomous mobile device 2, and can also be called the drive control unit. The movement control unit 2c is the brain that performs autonomous control, controlling the drive unit 2d (for example, actuators such as rotors) mounted on the autonomous mobile device 2 based on sensing information provided by the sensor group 2a, according to the operation plan data received by the communication unit 2b. The autonomous mobile device 2 performs autonomous control to operate according to the route, coordinates, attitude, and speed given as an operation plan at the start of operation. Therefore, even if the position of the autonomous mobile device 2 is slightly shifted due to control errors or disturbances such as wind, it can maintain its original position, attitude, and speed on its own by controlling the actuators based on the sensing information.

[0034] In this manner, the movement control unit 2c controls the drive unit 2d according to the detection results of the sensor group 2a, thereby moving the autonomous mobile device 2. Alternatively, the movement control unit 2c transmits some or all of the detection results of the sensor group 2a to an external server device (which may be the operation plan generation device 1, as long as it has the function of handling movement control) via the communication unit 2b. The movement control unit 2c then receives information for movement control from the server device via the communication unit 2b, and controls the drive unit 2d according to this information. The drive unit 2d drives the autonomous mobile device 2 according to the control from the movement control unit 2c, thereby moving the autonomous mobile device 2.

[0035] The movement control unit 2c can be configured as a control unit (not shown) that controls the entire autonomous mobile device 2. This control unit can be implemented, for example, by a CPU (Central Processing Unit), working memory, and a non-volatile storage device that stores a program. This program can be a program that executes processing related to autonomous movement control. This control unit can also be implemented, for example, by an integrated circuit.

[0036] Next, we will describe the various components of the operation plan generation device 1 that generates operation plans for autonomous mobile devices such as the autonomous mobile device 2.

[0037] The operation plan generation device 1 functions as a server device that provides operation plans and can be implemented as a server computer. The control unit 1a controls the entire operation plan generation device 1. Therefore, the control unit 1a can be implemented by, for example, a CPU, working memory, and a non-volatile storage device that stores a program. This program can be a program that executes processing related to the generation of operation plans. Furthermore, this control unit can also be implemented by, for example, an integrated circuit.

[0038] The control unit 1a, in response to a request for generating a route plan that includes route request information received from an external source, generates a route plan that includes multiple waypoints for the autonomous mobile device included as the target of the route in the route request information. The route request information may include, for example, a departure point, a destination, and a (possible or desired) departure time, and may also include a desired arrival time.

[0039] The memory unit 1b stores the relationship between the executed operation plan, which is an executed operation plan, and the actual operation results at the time of execution of that operation plan. Here, the executed operation plan is basically a plan generated by the control unit 1a, but it may also be a plan generated by another device, and can include both types of plans. The above relationship only needs to be stored in a state that allows for the prediction of deviations described later (for example, a calculation formula that shows the relationship), but it is desirable that it be stored as a trained model, as will be described later.

[0040] If the execution operation plan is generated by the control unit 1a, it can be stored in the storage unit 1b at the time of generation; otherwise, it can be obtained from an external source via communication or a portable recording medium. The operation record can be obtained directly from the autonomous mobile device 2 via communication, but the timing is not restricted, and it can also be received via other devices or via a portable recording medium. However, the storage unit 1b only needs to store the above-mentioned relationships obtained from these sources, and it does not need to store either the execution operation plan or the operation record.

[0041] The control unit 1a first generates a candidate route plan including multiple passing points based on the desired route information, and then, based on the above relationship, performs a prediction of the positional and temporal deviations that may occur if the train is operated according to this candidate route plan.

[0042] For example, a pre-trained model can be used for this prediction. In other words, the above relationship can be a pre-trained model generated by machine learning based on the executed operation plan and the actual operation, which predicts the positional and temporal deviations that may occur when operating according to the candidate operation plan. In this case, this pre-trained model is stored in the memory unit 1b, and the control unit 1a uses this pre-trained model to perform the above prediction. This pre-trained model is a model that has learned the deviations in destination and arrival time between past operation plans and actual results. The destination here can include not only the destination but also a set number of waypoints.

[0043] The control unit 1a sets an exclusive region for the above-mentioned candidate operation plan, based on the predicted results, which is an area where other autonomous mobile devices are prohibited from operating in competition. This exclusive region can be set as an area where the operation of other autonomous mobile devices that have not yet generated an operation plan or autonomous mobile devices with generated but incomplete operation plans is prohibited. Therefore, this exclusive region corresponds to the margin range of the operation path of the autonomous mobile devices, which is determined based on the deviation prediction results.

[0044] The control unit 1a then selects from the candidate operation plans that do not conflict with operation plans that have already been generated but are not yet completed for other autonomous mobile devices, and uses this as the operation plan to respond to the operation plan generation request. The operation plan generation device 1 can then return the operation plan thus determined as a response to the operation plan generation request. The determined operation plan can also be transmitted to the autonomous mobile device 2 sequentially during its operation, for example, in the amount necessary for control at that time.

[0045] Furthermore, as described above, this embodiment can be constructed as an autonomous mobile system comprising a route plan generation device 1 and an autonomous mobile device 2 equipped with a group of sensors that collect information to be included in the above-mentioned operational performance and capable of communicating with the route plan generation device 1. In addition, the route plan generation device 1 can be configured as a single device, but it can also be configured as a system by distributing its functions among multiple devices.

[0046] Next, with reference to Figure 3, the operation plan generation method performed in the operation plan generation device 1 will be described. Figure 3 is a flowchart illustrating an example of processing in the operation plan generation device according to Embodiment 1.

[0047] As described above, the operation plan generation device 1, in response to an operation plan generation request that includes operation preference information, performs the process of generating an operation plan that includes multiple waypoints for the autonomous mobile device 2 included as the target of operation in the operation preference information.

[0048] In this process, first, the operation plan generation device 1 generates a candidate operation plan that includes multiple passing points based on the desired operation information (step S1). Next, the operation plan generation device 1 predicts the positional and temporal deviations that may occur if the operation is carried out according to the candidate operation plan, based on the relationship between the actual operation plan and the actual operation results when the actual operation plan is carried out (step S2).

[0049] Next, the operation plan generation device 1 sets an exclusive area for the operation plan candidates based on the predicted results (step S3). Then, the operation plan generation device 1 selects a candidate from the operation plan candidates that does not conflict with operation plans that have already been generated but are not yet completed for other autonomous mobile devices, and sets that candidate as the operation plan to respond to the operation plan generation request (step S4), and terminates the process. The operation plan determined here will be set in the autonomous mobile device 2 that is designated as the device to be operated in the operation plan generation request. Furthermore, as described above, the operation plan generation device 1 can be configured as a computer, and in that case, the program can be a program that causes the computer to execute the above-described process.

[0050] As explained above, in this embodiment, exclusive zones are set in candidate operation plans based on predicted deviations based on past performance. Therefore, the set exclusive zones are variable and based on past performance. Accordingly, according to this embodiment, it is possible to automatically generate operation plans that increase the number of autonomous mobile devices that can operate in a certain space at the same time, thereby increasing the capacity density and number of simultaneous operations of autonomous mobile devices in space while avoiding collisions between autonomous mobile devices. In other words, this embodiment can achieve both collision prevention and improved space utilization efficiency. Furthermore, in this embodiment, exclusive zones can be assumed to be not only two-dimensional, such as on the ground, but also three-dimensional, such as in the airspace.

[0051] Furthermore, in this embodiment, as described as an exclusive area where other operations are prohibited, the autonomous mobile device 2 can be operated based on the concept of an operating area where, for example, only one device is allowed in one area. From this perspective, this embodiment can be said to be a safe and useful concept for route design in areas where collision avoidance is difficult for the autonomous mobile device on its own, or where there are limitations to its avoidance capabilities, including judgment and detection, such as small drones flying beyond visual line of sight without a human on board. In addition, according to this embodiment, an operation plan can be automatically generated simply by specifying the desired operation information.

[0052] Next, with reference to Figure 4, we will describe a learning device that generates a trained model usable by the operation plan generation device 1. Figure 4 is a block diagram showing an example configuration of a learning device that generates a trained model usable by the operation plan generation device 1.

[0053] As shown in Figure 4, the learning device 4 may include an acquisition unit 4a and a control unit 4b. Here, the control unit 4b may include a storage unit 4c.

[0054] The acquisition unit 4a acquires information including the execution operation plan and the actual operation results based on that plan. If the execution operation plan is a plan generated by the control unit 1a, it can be acquired after generation from the operation plan generation device 1 or from the autonomous mobile device 2 that acquired the generated plan, via communication or a portable recording medium. The actual operation results can also be acquired, for example, via the operation plan generation device 1, directly from the autonomous mobile device 2 via communication, or via a portable recording medium. In addition, the learning data may also include various environmental data acquired from sensor groups 2a and weather information websites.

[0055] The control unit 4b generates a trained model as described above based on the information acquired by the acquisition unit 4a. Specifically, the control unit 4b inputs the information acquired by the acquisition unit 4a as training data into the untrained model 4d stored in the memory unit 4c, performs machine learning, obtains a trained model 4e, and stores it in the memory unit 4c. Alternatively, it updates the untrained model 4d to create the trained model 4e. The algorithm used in the untrained model 4d is irrelevant as long as the generated trained model 4e can perform the predictions described above. The algorithm (learner) to be trained can be arbitrarily determined by the analyst. The analyst can then select an appropriate trained model by running the trained models, or it can be automatically selected to meet predetermined conditions that are considered appropriate.

[0056] The learning device 4 can be configured as a computer. Therefore, the control unit 4b can be implemented, for example, by a CPU, working memory, and a non-volatile storage device that stores the program. This program can be a program for generating trained models.

[0057] Furthermore, the learning device 4 can be configured as a single device, but its functions can also be distributed among multiple devices to form a system. The learning device 4 can also be installed in or connected to the operation plan generation device 1. However, the learning device 4 can also be configured as an independent device not connected to the operation plan generation device 1; in this case, for example, the trained model generated by the learning device 4 can be copied or moved to the operation plan generation device 1 using a portable recording medium or the like.

[0058] Next, the learning method performed in the learning device 4 will be explained with reference to Figure 5. Figure 5 is a flowchart illustrating an example of processing in the learning device 4 shown in Figure 4. First, the learning device 4 acquires information including the execution operation plan and the actual operation results based on that execution operation plan (step S11). Then, the learning device 4 generates a trained model based on the acquired information (step S12).

[0059] With this learning device 4, the operation plan generation device 1 can generate a trained model that can be used as the above relationship, and the operation plan generation device 1 can obtain appropriate prediction results obtained through machine learning and determine the exclusive region corresponding to the actual deviation based on those prediction results.

[0060] <Embodiment 2> Embodiment 2 will be explained with reference to Figures 6 to 43, focusing on the differences from Embodiment 1, although various examples described in Embodiment 1 can be applied. The following explanation assumes that the flight plan generated by the flight plan generation device according to this embodiment is a plan generated for a drone, but the essence remains the same even if the drone is replaced with other types of autonomous mobile devices.

[0061] This embodiment will be explained using Figure 6 as an example, with reference to the case where a delivery company (transportation company) Tr uses a drone for parcel delivery purposes. Figure 6 is a schematic diagram showing one example configuration of the operation management system according to this embodiment. However, it can be similarly applied to autonomous mobile devices other than drones, and can also be similarly applied to purposes other than parcel delivery. For example, the use of drones can be applied not only to parcel delivery purposes, but also to purposes such as mediating, managing, and allocating flight routes for companies that perform beyond visual line of sight (BVLOS) flights for tasks such as cargo transport, photography, and surveillance, as performed by the Ministry of Land, Infrastructure, Transport and Tourism. Furthermore, this embodiment is not limited to the operation management of drones and the operations using them, but can be applied to the operation management of autonomous mobile devices in general, such as robots, autonomous vehicles, automated guided vehicles in factories, and ships and aircraft equipped with autonomous navigation functions, as well as operations using these devices. In addition, this embodiment can be applied to the transportation industry, manufacturing industry, infrastructure inspection and maintenance management, construction and civil engineering, public services (flood control, dam management, disaster prevention, disaster response, police, fire), security, etc.

[0062] As shown in Figure 6, the operation management system according to this embodiment includes an operation management device 10 that performs operation management. A delivery company Tr, which receives an order from a shipper sh, becomes the requester for the issuance of an operation plan and requests the operation management device 10 to issue an operation plan, thereby obtaining an operation plan. The operation management device 10 functions as a server device that provides the operation plan.

[0063] The flight management device 10 is an example of the flight plan management device 1 according to Embodiment 1, and in this example, it has a mechanism for managing the operation of the drone 3 and an air traffic control function for an aircraft. The flight management device 10 has a function to issue a flight plan in response to a request from a flight management requester Tr, and a function to collect and manage the flight history of the drone 3 via communication at all times or online and offline after the completion of the operation.

[0064] The shipper sh requests the dispatch plan issuer Tr to transport the goods (operate drone 3). The shipper sh provides the dispatch plan issuer Tr with requests regarding the operating conditions (start time (loading time or time slot), end time (desired delivery time or time slot), start location (loading location), end location (desired destination)). In Figure 6, the shipper sh is represented by a human icon, but the actions of this component can be automated by replacing them with a computer system, etc. For example, the shipper sh can place an order with a server device used by the delivery company Tr using a terminal device such as a PC (Personal Computer), smartphone, or dedicated terminal.

[0065] The requester Tr, based on the order from the shipper sh, requests the operation management device 10 to issue an operation plan, and the device then issues the operation plan. The requester Tr registers the received operation plan with the drone 3 to be operated and is responsible for operating the drone 3 in accordance with the operation plan.

[0066] In Figure 6, the requester of the operation plan issuance, Tr, is represented by a human icon. However, it merely mediates the exchange of information between the shipper sh, drone 3, and operation management device 10. The functions of this component can be automated by replacing them with a computer system or the like. For example, the delivery company Tr uses a terminal device such as a PC, smartphone, or dedicated terminal to access the server device used by the delivery company Tr and sends an operation plan issuance request to the operation management device 10, receiving the operation plan in response. This operation plan is generated and issued by the operation management device 10 based on the operation plan issuance request from the delivery company Tr.

[0067] Drone 3 transports the cargo of shipper sh according to the registered flight plan. Although this embodiment describes a transportation use case, it can be similarly applied to passenger transport, photography, or any other case in which Drone 3 is operated.

[0068] Next, with reference to Figures 7 to 9, an example of the configuration of the operation management device 10 according to this embodiment will be described in general terms. Figure 7 is a block diagram showing an example of the configuration of the operation management device according to Embodiment 2, Figure 8 is a schematic diagram for explaining the section management according to the comparative example, and Figure 9 is a schematic diagram for explaining an example of management in the operation management device of Figure 7.

[0069] As shown in Figure 7, the operation management device 10 according to this embodiment may include a performance learning unit 100, a performance management unit 200, a plan formulation unit 300, a plan management unit 400, and a space management unit 500. Each component shown in Figure 7 may be a program on a computer, an electrical or electronic circuit, or a combination of a program and an electrical or electronic circuit. Information can be exchanged between components connected by lines in Figure 7. If it is a program, information can be exchanged between elements via shared memory or storage, or if the units are separated in implementation, information can be exchanged via communication.

[0070] In general terms, the flight management device 10 according to this embodiment may include a flight plan formulation unit 304 in the planning unit 300 that sets exclusive areas based on deviation predictions and formulates a flight plan. The above deviation prediction can refer to learning the route deviation between the executed flight plan and the actual flight, and using the learned results to predict the route and passing time deviations. The flight plan formulation unit 304 can automatically generate a flight plan based on the prediction results. In the case of the drone 3, the flight plan may be called a flight plan.

[0071] Furthermore, the operation management device 10 according to this embodiment manages the operation route of the autonomous mobile device 2, such as the drone 3, not by managing exclusive access in sections as in the comparative example shown in Figure 8, but as follows. That is, the operation management device 10 manages exclusive access using waypoints (passage points) as shown by circles in Figure 9, edges (sections connecting the passage points), and variable exclusive access areas set around those edges.

[0072] To explain in more detail, in this embodiment, the system learns the discrepancies between past planned and actual destinations and arrival times, and uses these learning results to predict these discrepancies when formulating the operation plan, thereby determining the margin of safety in the autonomous mobile device's route (hereinafter referred to as the "risk range" in this embodiment).

[0073] Furthermore, in this embodiment, when designing a route, the method of selecting from a space where the margin range of the entire space (all sections) is a continuous arrangement of rectangular prisms of a fixed, uniform size is not adopted, and the exclusion time of the dangerous area is not to be applied to the entire section from the start to the end of operation of a certain autonomous mobile device.

[0074] In the comparative example's path design, as shown in Figure 8, a rectangular prism (cube) larger than the danger zone is defined in advance to fill the space, and as shown in the shaded area of ​​Figure 8, the path for the autonomous mobile device is set by selecting this cube, and the autonomous mobile device is made to travel as close to the center as possible. In the comparative example, this path design sets a path that does not cause collisions with the autonomous mobile device.

[0075] In contrast, in this embodiment, as shown in Figure 9, multiple points called waypoints are densely defined in space, and a route is set by selecting the waypoints to be passed through. A waypoint is a passing point, and it is a point that defines the place to be passed through and the time or time period to be passed through in the operation plan. Furthermore, in this embodiment, a dangerous range (an exclusive range in which routes for other autonomous mobile devices are not set) is set around the lines connecting the waypoints, within a range corresponding to the predicted value of the route deviation.

[0076] More specifically, in areas where the predicted deviation from the planned route is small based on operational data, the exclusive zone called the "hazard zone" is reduced. Conversely, if the predicted deviation is large, the exclusive zone of the hazard zone is increased.

[0077] Furthermore, in the operation management device 10, if there is a large discrepancy between the planned arrival time and the actual arrival time at a waypoint, and an advance or delay in arrival time is predicted, the device will process it as follows: In other words, the operation management device 10 will lengthen the time for which the section is excluded as a hazard zone (lifetime of the hazard zone) according to the magnitude of the discrepancy, and conversely, if the discrepancy is small, it will shorten the exclusion time (lifetime of the hazard zone).

[0078] Thus, in this embodiment, it is possible to design a route that passes through smaller (or conversely, larger) sections rather than large, fixed-size sections.

[0079] In this way, the operation management device 10 can set the routes of autonomous mobile devices more flexibly. In particular, the operation management device 10 can reduce the area and time that the hazard zone occupies in the space compared to the comparative example, allowing for the setting of more routes and increasing the number of autonomous mobile devices that can be accommodated in the space.

[0080] Thus, in this embodiment, instead of partitions in space, a high-density network of points called waypoints is defined, and the route is designed by setting a route that passes through these waypoints. In particular, in this embodiment, not only is a danger zone (an area that excludes other autonomous mobile devices) set to surround the route, but it can also be set as follows: That is, its size (the size of the surface perpendicular to the edge of the exclusion zone) and exclusion time (exclusion period) can be determined and set based on predicted values ​​of deviation predicted from learning past route deviation information. Here, the danger zone can be set to a different size for each edge between waypoints in the designed route, based on predictions.

[0081] In other words, the size of the plane perpendicular to the direction of travel (the plane perpendicular to the path) in the exclusive area is set to differ for at least two edges connecting adjacent waypoints. Thus, in this embodiment, the exclusive area is set according to the predicted displacement, and it is permitted to set the above size differently, so it is possible to have an uneven division with different sizes depending on the section.

[0082] Furthermore, the exclusive time (exclusive period) can be set to differ for at least two edges, thereby enabling a more heterogeneous partitioning with varying temporal sizes.

[0083] Next, we will explain each component in Figure 7. The coordinated operation between each component will be described later as the operation of the operation management device 10, referring to Figures 19 and 20, etc.

[0084] As explained with reference to Figure 6, the flight management device 10 has a mechanism for managing the operation of the drone 3 and an air traffic control function for aircraft. Specifically, the flight management device 10 has the function of issuing an operation plan in response to a request from the flight management requester Tr, and the function of collecting and managing the operation history of the drone 3 via communication at all times when it is operating, or online and offline after the operation is completed. To achieve this, as described above, the flight management device 10 comprises a performance learning unit 100, a performance management unit 200, a plan formulation unit 300, a plan management unit 400, and a space management unit 500.

[0085] (Achievement Learning Department 100) When drone 3 is actually operated, a discrepancy occurs between the operation plan and the actual operation route. Therefore, the performance learning unit 100 learns from the operation performance information collected from drone 3 by the performance management unit 200. This provides a trained model that can be used to predict the deviation of drone 3's route from the operation plan. To achieve this, the performance learning unit 100 comprises a trained model management unit 101, an operation performance learning unit 102, and a trained model holding unit 103.

[0086] The trained model management unit 101 manages the trained model data stored in the trained model holding unit 103 (such as data retrieval, loading, and searching).

[0087] The flight performance learning unit 102 primarily learns the differences (positional differences, time differences) between the flight plan and the actual flight performance of the drone 3. The trained model obtained through this learning process is stored in the trained model holding unit 103 via the trained model management unit 101.

[0088] The trained model storage unit 103 stores the trained model learned by the operation performance learning unit 102. This trained model is used by the planning unit 300, described later, to predict the generated operation plan and the positional and temporal deviations that may occur when operating according to that operation plan.

[0089] The flight performance learning unit 102 reads the drone's flight performance information from the flight performance retention unit 202 of the performance management unit 200 via the flight performance management unit 201. The drone's flight performance information can be information that shows the history of how the drone actually flew between waypoints, for example, as shown by the path deviation (dashed line) in the left diagram of Figure 10. Figure 10 is a conceptual diagram to explain the relationship between waypoints, edges, and path deviations, as well as the exclusive area.

[0090] The operational performance learning unit 102 compares this information with the issued operational plans stored in the issued operational plan storage unit 402, which is read via the issued operational plan management unit 401 of the planning management unit 400. The operational performance learning unit 102 then learns the route deviation between the plan and the actual results obtained from the comparison, the deviation in arrival time at waypoints, the time period in which the deviation occurred, and the edge (and waypoints at both ends) in which the deviation occurred, and builds a trained model. The above route deviation may be the maximum value or mode of the deviation in the section. Furthermore, the time period in which the deviation occurred can be, for example, a tag representing the month, a tag representing each week when the year is divided into one-week periods, a tag representing the time when the day is divided into six parts, or a combination thereof, since the deviation is greatly influenced by the season, temperature, climate, and weather. In other words, the time period in which the deviation occurred means a time period divided into units in which the same trend can be obtained for the deviation.

[0091] The operation performance learning unit 102 stores the constructed trained model in the trained model holding unit 103 via the trained model management unit 101. When the trained model stored in the trained model holding unit 103 is given the location (edge ​​(or waypoints at both ends)) where the deviation is to be predicted and the time at which the train is scheduled to travel to that location, a predicted deviation value can be obtained. The obtained predicted value is the predicted deviation of the route and the deviation from the time of passing the endpoint (the time of passing the end of the edge) that will occur at that time and location. However, in this embodiment, the explanation will be based on the premise that this trained model is transferred to the planning unit 300 for use.

[0092] (Performance Management Department 200) The performance management unit 200 collects, stores, and manages the flight history of drone 3 via communication, either continuously or online and offline after the completion of the flight. To achieve this, the performance management unit 200 comprises a flight history storage unit 202 and a flight history management unit 201.

[0093] The flight record retention unit 202 stores the flight history collected and accumulated by the drone 3 during its operation, as well as the information on its retrieval. The flight record management unit 201 manages the flight history stored in the flight record retention unit 202. In addition, the flight record management unit 201 searches for and retrieves the flight history from the flight record retention unit 202 in response to a request from the performance learning unit 100, and passes it to the performance learning unit 100.

[0094] (Planning Department 300) The planning unit 300 formulates an operation plan in cooperation with other functional units in response to a request from the operation plan requester Tr. For this purpose, the planning unit 300 comprises a trained model holding unit 103, an exclusive area calculation unit 301, an operation plan issuance request management unit 302, an operation plan issuance request holding unit 303, an operation plan formulation unit 304, and an operation plan candidate temporary holding unit 305.

[0095] The pre-trained model holding unit 103 of the planning unit 300 holds the pre-trained model generated by the performance learning unit 100. The pre-trained model held here may be the pre-trained model generated by the performance learning unit 100 itself, or a copy obtained via the pre-trained model management unit 101. It is used to predict the operation plan and the expected deviation from that plan when formulating the operation plan.

[0096] The exclusive area calculation unit 301 calculates an exclusive area to be set around the operation plan. The exclusive area calculation unit 301 uses a trained model managed by the trained model holding unit 103 to predict the positional and temporal deviations that occur on the designed operation route, and calculates the exclusive area as the range obtained by adding a safety margin (danger zone) to the predicted range.

[0097] An exclusive area calculation unit 301 is provided to design the flight path so that other drones 3 do not enter this exclusive area. The planning unit 300 designs the flight path so that the exclusive areas of one drone 3 and the exclusive areas of another drone 3 do not overlap, as there is a risk of collision if they overlap.

[0098] The Operation Plan Issuance Request Management Unit 302 receives requests from Operation Plan Issuance Requesters Tr, registers them in the Operation Plan Issuance Request Holding Unit 303, and requests the Operation Plan Formulation Unit 304 to formulate an operation plan.

[0099] The operation plan issuance request storage unit 303 stores the content of the request from the operation plan issuance requester Tr, as well as the status of whether or not an operation plan has been issued in response to the request. The operation plan issuance request storage unit 303 can store this information in a format that allows searching by any item, similar to a relational database.

[0100] Here, an example of the data format (information items) held by the operation plan issuance request holding unit 303 will be explained with reference to Figures 11 and 12. Figure 11 is a diagram showing an example of the data format of an operation plan issuance request (operation plan generation request including operation preference information), and Figure 12 is a diagram showing an example of the data format held by the operation plan issuance request holding unit 303 in response to the request in Figure 11.

[0101] As illustrated in Figure 11, a request from the requester Tr for issuing a route plan can include route preference information such as departure point (latitude, longitude, altitude), arrival point (latitude, longitude, altitude), possible departure time (year, month, day, hour, minute, second), and desired arrival time (year, month, day, hour, minute, second). It is also possible to configure the system so that the requester Tr accesses the route management device 10 from a PC or other device and makes a request, for example, by specifying the departure and arrival points using waypoint IDs.

[0102] Then, when the Operation Plan Issuance Request Management Unit 302 receives a request from the Operation Plan Issuance Requester Tr, as shown in Figure 12, it registers the request management number and the operation route issuance number, in addition to the information included in the request, with the Operation Plan Issuance Request Holding Unit 303. In other words, when a request is received, the Operation Plan Issuance Request Holding Unit 303 registers the request management number, departure point (latitude, longitude, altitude), arrival point (latitude, longitude, altitude), possible departure time (year, month, day, hour, minute, second), desired arrival time (year, month, day, hour, minute, second), and operation route issuance number.

[0103] Furthermore, when the operation plan issuance request management unit 302 has finished generating the operation plan and returns a response to the operation plan issuance requester Tr, it updates the operation plan issuance request holding unit 303 by adding an issued flag and an operation route issuance number to the row with the matching request management number.

[0104] As described above, the planning unit 300 formulates an operation plan in cooperation with other functional units in response to a request from the operation plan requester Tr. The operation plan formulation unit 304 then executes this formulation. The operation plan formulation unit 304 designs the operation route so that other drones 3 do not enter the exclusive area calculated by the exclusive area calculation unit 301. Since there is a risk of collision if the exclusive area of ​​one drone 3 and the exclusive area of ​​another drone 3 overlap, the operation plan formulation unit 304 designs the operation route so that they do not overlap.

[0105] The temporary operation plan candidate storage unit 305 temporarily stores candidate operation plans that are created during the operation plan formulation process. The temporary operation plan candidate storage unit 305 can store information indicating candidate operation plans in a format that allows searching by any item, similar to a relational database.

[0106] Here, an example of the data format (information items) held by the temporary storage unit 305 for candidate operation plans will be explained with reference to Figures 13 and 14. Figure 13 shows an example of the data format for candidate operation plans created during the planning of an operation plan, and Figure 14 shows an example of the data format for an operation plan issued for the candidate operation plan in Figure 13.

[0107] As shown in Figure 13, the temporary storage unit 305 for proposed route plans temporarily stores the departure waypoint ID, possible departure time (year, month, day, hour, minute, second), arrival waypoint ID, arrival deadline time (year, month, day, hour, minute, second), and a list of route lines. Here, the list of route lines has columns for section sequence number, starting waypoint ID, ending waypoint ID, target destination arrival time, and edge ID.

[0108] Furthermore, the temporary operation plan candidate storage unit 305 also temporarily stores a conflict flag indicating whether or not this candidate conflicts with others. The conflict flag item can be used when determining whether the operation plan candidate does not conflict with others, that is, whether it is acceptable to adopt it as an operation plan. Then, if the operation plan formulation unit 304 determines that it is acceptable to adopt it as an operation plan, it removes the conflict flag item stored in the temporary operation plan candidate storage unit 305, as shown in Figure 14, and decides it as an operation plan.

[0109] The operation plan formulation unit 304, in this manner, first generates a draft (candidate) of the operation plan when formulating an operation plan and temporarily stores it in the operation plan candidate temporary storage unit 305. The operation plan formulation unit 304 then calculates the exclusive area for the stored operation plan and checks if it overlaps with the exclusive area of ​​an already issued operation plan. If there is no overlap, it issues it as an official operation plan. On the other hand, if there is an overlap, the operation plan formulation unit 304 generates another operation plan draft, and therefore has a mechanism to temporarily store the operation plan draft, such as the operation plan candidate temporary storage unit 305.

[0110] (Planning and Management Department 400) The planning management unit 400 holds and manages issued operation plans. The issued operation plans managed here are used to learn from discrepancies between operation plans and actual operation results, and to check whether the exclusive domains of the formulated operation plan overlap with the exclusive domains of already issued operation plans when formulating operation plans. To this end, the planning management unit 400 comprises an issued operation plan management unit 401 and an issued operation plan holding unit 402.

[0111] The issued operation plan management unit 401 manages issued operation plans in the issued operation plan storage unit 402, that is, it handles the loading, unloading, and retrieval of issued operation plans. The issued operation plan management unit 401 is also responsible for providing issued operation plans to the performance learning unit 100 and for exchanging issued operation plans with the plan formulation unit 300.

[0112] The issued operation plan storage unit 402 stores issued operation plans and exclusive areas for those operation plans. The issued operation plan storage unit 402 can store information about issued operation plans and exclusive areas for those operation plans in a format that allows searching by any item, similar to a relational database.

[0113] Here, an example of the data format (information items) held by the issued operation plan retention unit 402 will be explained with reference to Figures 15 and 16. Figure 15 is a diagram showing an example of the data format of an issued operation plan, and Figure 16 is a diagram showing an example of the data format when multiple issued operation plans from Figure 15 are arranged side by side.

[0114] The issued operation plan storage unit 402 can store information as shown in Figure 15 as an issued operation plan. Specifically, an issued operation plan may include waypoint IDs for the departure and arrival points, possible departure time (year, month, day, hour, minute, second), arrival deadline time (year, month, day, hour, minute, second), request management number, operation route issuance number, operation route, and exclusive area (exclusive area radius and exclusive end time). This information can be grouped together, and this group can be added for each issued operation plan, resulting in a data format as shown in Figure 16.

[0115] (Space Management Department 500) The spatial management unit 500 manages information about the space in which the drone 3 operates. This information (spatial information) can consist of coordinate information of points in space called waypoints, which are guidelines for the route the drone 3 will take when operating, and edge information, which are lines (vectors) connecting those points. To manage spatial information, the spatial management unit 500 includes a spatial information management unit 501 and a waypoint (WP) information holding unit 502.

[0116] The spatial information management unit 501 manages and searches waypoint and edge information. The waypoint information storage unit 502 stores coordinate information of waypoints, which are spatial points that serve as guidelines for the route when the drone 3 is operated and as intermediate points during operation. The waypoint information storage unit 502 can store the coordinate information of waypoints in a format that allows searching by any item, similar to a relational database.

[0117] Here, an example of the data format (information items) held by the waypoint information holding unit 502 will be explained with reference to Figure 17. Figure 17 is a diagram showing an example of the data format for waypoint coordinate information. As shown in Figure 17, the waypoint holding unit 502 holds a list of waypoint IDs and waypoint coordinates (latitude, longitude, altitude) as waypoint coordinate information. For example, it holds the coordinates of a point in space as shown in Figure 9. In the example of Figure 9, the waypoints are arranged evenly vertically and horizontally, but it is not necessary for the points to be evenly arranged; they can be arranged irregularly without any problem.

[0118] The edge information storage unit 503 stores edge information, which is a line (vector) connecting waypoints, and a flag (unavailable flag) indicating whether or not the edge can be used as part of the route. The edge information storage unit 503 can store edge information in a format that allows it to be searched by any item, similar to a relational database.

[0119] Here, an example of the data format (information items) held by the edge information holding unit 503 will be explained with reference to Figure 18. Figure 18 is a diagram showing an example of the data format of edge information.

[0120] As shown in Figure 18, the waypoint holding unit 502 can store information about the edges connecting waypoints, including the edge ID, the starting waypoint ID of the edge, the ending waypoint ID, and a list of unavailable flags. The unavailable flags are set to indicate that the waypoint cannot be used in sections that should not be used as part of the route, for example, if there is an obstacle on the line connecting the waypoints.

[0121] (Example of operation of the operation management device 10) An example of the internal operation of the operation management device 10 will be explained with reference to Figures 19 to 43. First, let's explain the operation of the performance learning unit 100. As described above, the operation performance learning unit 102 of the performance learning unit 100 reads the drone's operation performance information from the operation performance retention unit 202 of the performance management unit 200 via the operation performance management unit 201. The drone's operation performance information can be information that shows the history of how the drone actually operated between waypoints, for example, as shown by the deviation of the route (dashed line) in the left diagram of Figure 10. The operation performance learning unit 102 also compares this information with the issued operation plans held in the issued operation plan retention unit 402, which is read via the issued operation plan management unit 401 of the planning management unit 400.

[0122] The operational performance learning unit 102 then learns the route deviation between the plan and the actual route obtained from the comparison, the deviation in arrival time at waypoints, the time period in which the deviation occurred, and the edge (and the waypoints at both ends) in which the deviation occurred, and builds a trained model. The above route deviation may be the maximum value or mode of the deviation in the section. Also, the time period in which the deviation occurred means a time period divided into units such that the same trend can be obtained for the deviation, as described above.

[0123] Next, the operation performance learning unit 102 stores the constructed trained model in the trained model holding unit 103 via the trained model management unit 101. When the trained model stored in the trained model holding unit 103 is given the location (edge ​​(or waypoints at both ends)) where the deviation is to be predicted and the time at which the train is scheduled to travel to that location, a predicted deviation value is obtained. The obtained predicted value is the predicted deviation of the route that will occur at that time and place, and the predicted deviation relative to the time of passing the endpoint (the time of passing the end of the edge).

[0124] Next, we will explain the operation of the planning unit 300. The interactions between each element are shown in the sequence diagrams in Figures 19 and 20, and the processing within each functional unit will be explained individually with reference to the flowcharts. Figures 19 and 20 show an example of sequential execution of the operation plan formulation process. Figures 19 and 20 are sequence diagrams showing an example of operation in a series of operation management devices 10. Here, the entire sequence is represented using UML (Unified Modeling Language) 2.0 notation.

[0125] First, the person requesting the issuance of a route plan requests the issuance of a route plan from the route plan issuance request management unit 302, specifying the information for the items shown in Figure 11 (step S21). As shown in Figure 11, the items of information to be specified (desired route information) can be the departure point (latitude, longitude, altitude), arrival point (latitude, longitude, altitude), possible departure time (year, month, day, hour, minute, second), and desired arrival time (year, month, day, hour, minute, second).

[0126] The operation plan issuance request management unit 302 registers the received operation plan issuance request in the operation plan issuance request holding unit 303 (step S22). The process in step S22 will be explained with reference to Figure 21. Figure 21 is a flowchart illustrating an example of the operation plan issuance request registration process.

[0127] The operation plan issuance request management unit 302 receives an operation plan issuance request from the operation plan issuance requester, which includes the information shown in Figure 11 (step S51). The operation plan issuance request management unit 302 adds a request management number and an issued flag (False) to the contents of the operation plan issuance request, as shown in Figure 12, and registers and stores it in the operation plan issuance request holding unit 303 (step S52). Here, the information specified in Figure 11 in step S21 is assigned and added to the operation plan issuance request holding unit 303 as shown in Figure 12. At this point, the issued flag is false and the operation route issuance number is left blank.

[0128] Next, the Operation Plan Issuance Request Management Unit 302 issues an operation plan formulation request to the Operation Plan Formulation Unit 304, using the parameters for the operation plan formulation request as shown in Figure 22 (step S53). This step S53 corresponds to step S23 in Figure 19. Here, Figure 22 is a diagram showing an example of the data format for the parameters of the operation plan request.

[0129] Thus, in step S23, the operation plan issuance request management unit 302 requests the operation plan formulation unit 304 to formulate an operation plan, specifying parameters. As shown in Figure 22, the parameters for this operation plan request include the request management number, departure point (latitude, longitude, altitude), arrival point (latitude, longitude, altitude), possible departure time (year, month, day, hour, minute, second), and desired arrival time (year, month, day, hour, minute, second). The request management number is added when responding with the result, enabling the operation plan issuance request held by the operation plan issuance request holding unit 303 to associate with the responded operation plan.

[0130] Following step S23, the operation plan formulation unit 304 receives a request for operation plan formulation and determines the spatial information acquisition range (step S24). An example of the process in step S24 will be explained with reference to Figure 23. Figure 23 is a flowchart illustrating an example of the process for determining the spatial information acquisition range.

[0131] The operation plan formulation unit 304 receives the operation plan formulation request transmitted in step S23 from the operation plan issuance request management unit 302 using the parameters of the operation plan formulation request shown in Figure 22 as parameters (step S61). The operation plan formulation unit 304 then calculates the start point and end point of the requested range based on the departure point coordinates and arrival point coordinates of the parameters of the operation plan formulation request shown in Figure 22, as shown in Figure 24 (step S62). Figure 24 is a schematic diagram showing an example of the spatial information acquisition range.

[0132] Next, the Operation Planning Department 304 makes a spatial information acquisition request to the Spatial Information Management Department 501 using the spatial information acquisition request parameters shown in Figure 25 (step S63). Figure 25 is a diagram showing an example of the data format of the spatial information acquisition request parameters.

[0133] As shown in Figure 24, the operation plan formulation unit 304 defines the spatial information acquisition range as the range of possible routes for the operation plan. Here, as shown in Figure 24, the spatial information acquisition range is calculated to be wider than the range of a rectangular prism with the starting point and arrival point as diagonals. This is because there are cases where an operation route may conflict with an already issued operation route, and a route is created to take a detour to avoid it. In this way, by making the spatial information acquisition range wider by a predetermined distance than the rectangle with the coordinates of the starting point and arrival point on the horizontal plane as diagonals, a detour can be secured. Also, since a drone is used as an example here, the upper and lower altitude limits shown in Figure 24 are legally determined and are therefore not specified. However, when applying this to other autonomous mobile devices, the spatial information acquisition range may be determined by specifying the altitude as well.

[0134] In step S25, the spatial information acquisition request, the operation plan formulation unit 304 requests the spatial information management unit 501 to acquire spatial information by providing search parameters. These search parameters include information on items such as the departure point (latitude, longitude, altitude), arrival point (latitude, longitude, altitude), start point of the requested range (latitude, longitude), and end point of the requested range (latitude, longitude), as shown in Figure 25.

[0135] The spatial information management unit 501 receives the spatial information acquisition request in step S25 and performs a search for spatial information (step S26). An example of the process in step S26 will be explained with reference to Figure 26. Figure 26 is a flowchart illustrating an example of the process for determining the spatial information acquisition range.

[0136] In step S26, the spatial information management unit 501 receives a spatial information acquisition request from the operation plan formulation unit 304, with the requested range as a parameter (search parameter) (step S71). Next, the spatial information management unit 501 reads spatial information about the requested range (Figure 25), which is a parameter of the spatial information acquisition request in step S25, from the waypoint information in Figure 17 and the edge information in Figure 18 (step S72). The spatial information read here includes waypoints and edges included in the range determined by the start point of the requested range, the end point of the requested range vertex, the upper limit of the permitted operation, and the lower limit of the permitted operation, as shown in Figure 27, and this becomes the spatial information of the requested acquisition range. Figure 27 is a diagram showing an example of the data format of the acquisition range spatial information. The acquisition range spatial information acquired in this way will look like the image shown in Figure 28. Figure 28 is a schematic diagram showing an example of the acquisition range spatial information.

[0137] Then, the spatial information management unit 501 responds to the operation plan formulation unit 304 with the spatial information of the acquisition range shown in Figure 27 as the spatial information of the acquisition request range (steps S73, S27).

[0138] The operation plan formulation unit 304 receives spatial information of the acquisition request range from the spatial information management unit 501 (step S27), formulates a candidate operation plan based on that spatial information (step S28), and makes a request for exclusive area calculation to the exclusive area calculation unit 301 (step S29). In the process of step S28, a candidate operation plan is generated, not an actual operation plan. In subsequent processing, it is checked whether there are any conflicts in the operation route with an already issued operation plan before it becomes an actual operation plan, so what is generated at this point is a candidate operation plan.

[0139] The processing example for steps S27 to S29 will be explained with reference to Figures 29 and 30. Figures 29 and 30 are flowcharts illustrating an example of the process for formulating a candidate operation plan. First, the operation plan formulation unit 304 receives spatial information within the acquisition request range (Figure 27) from the spatial information management unit 501 (step S81). If the formulation in step S28 is the first execution, the operation plan formulation unit 304 registers the information shown in Figure 13 in the temporary operation plan candidate holding unit 305. That is, it registers the waypoint ID of the departure point, the possible departure time, the waypoint ID of the arrival point, and the arrival deadline time, as well as registering an empty table of operation routes and conflict flags. An empty table of operation routes and conflict flags is generated each time a candidate is formulated.

[0140] Next, the operation plan formulation unit 304 performs a loop process between steps S82a and S82b. The loop condition for this loop process is to repeat until the waypoint ID of the destination and the waypoint ID of the destination match.

[0141] In this loop process, the operation planning unit 304 first determines the starting waypoint as follows (step S83): For the first time, it searches for and reads the waypoint ID of the starting point in the spatial information within the acquisition request range received in step S81 (Figure 27). From the second time onward, it searches for and reads an edge where the endpoint waypoint ID and the starting waypoint ID from the preprocessing match, and where the unavailable flag is false.

[0142] Next, the operation planning unit 304 determines the target endpoint arrival time for each selected edge as follows (step S84). Using the starting waypoint ID and ending waypoint ID of the multiple edges searched and read in step S83 as keys, it obtains the waypoint coordinates of the waypoints whose spatial information waypoint IDs match within the acquisition request range (Figure 27). Then, from the distance between the two acquired waypoints and the standard speed of the drone to be used, it calculates the time required to travel along that edge (route section). Determining the time required means that the target endpoint arrival time for the edge can be determined.

[0143] Next, the operation plan formulation unit 304 selects a route section to proceed to the next waypoint from among the multiple edges searched and read in step S84 (step S85). At this time, edges whose edge ID matches the operation route and conflict flag table (which can be multiple) of the operation route candidate temporary storage unit 305 shown in Figure 13, and whose target endpoint time is close and whose conflict flag is true are excluded from selection. After such exclusions, edges are selected according to a route search algorithm such as a greedy algorithm or an A-star algorithm so that the route of the final operation plan is as short as possible. This determines the edge ID for one section and the endpoint waypoint (= next starting waypoint ID).

[0144] Next, the operation plan formulation unit 304 adds a record to the operation route and conflict flag table of the operation route candidate temporary storage unit 305, as shown in Figure 13 (step S86). Here, the assigned section sequence number (sequential number starting from 0), starting waypoint ID, ending waypoint ID, and conflict flag value false are added as one record.

[0145] In step S82b, the loop process is exited if the loop conditions described above are met. At the point of exiting the loop, one table of route information and conflict flags is completed.

[0146] Next, the operation plan formulation unit 304 requests the exclusive area calculation unit 301 to calculate the exclusive area of ​​the operation plan candidate, using the operation plan candidate that was last registered as information in Figure 13 in the operation plan candidate temporary holding unit 305 as a parameter (steps S87, S29).

[0147] As a result of this processing, the items shown in Figure 13 are formulated as candidates for the operation plan. In this step, the candidate operation plans generated and added to the temporary operation plan candidate storage unit 305 are retained until the operation plan is finalized, exiting the loop shown in Figure 20, and the registration and issuance of the operation plan is completed. After that, they are deleted.

[0148] In step S29, the request for calculation of exclusive areas for the candidate route plan is made to the exclusive area calculation unit 301 by specifying the items shown in Figure 14. The items shown in Figure 14 are the departure point waypoint ID, possible departure time (year, month, day, hour, minute, second), arrival point waypoint ID, arrival deadline time (year, month, day, hour, minute, second), and route. This route includes the section sequence number, starting waypoint ID, ending waypoint ID, target destination passage time, and edge ID.

[0149] Upon receiving this exclusive area calculation request, the exclusive area calculation unit 301 calculates the exclusive area of ​​the operation plan (step S30). Prior to the processing in step S30, it is assumed that the trained model described above is stored in the trained model holding unit 103 of the planning unit 300. In other words, if the trained model is given the location (edge ​​(or waypoints at both ends)) where the deviation is to be predicted and the time at which the train is scheduled to operate through that location, it is possible to obtain predicted values ​​of the deviation of the route and the deviation from the time of passing the endpoint that will occur at that time and location. This time of passing the endpoint refers to the time when the end of the edge is passed.

[0150] Step S30 will be explained in detail. The exclusive area calculation unit 301 reads the route shown in Figure 13 one record at a time and uses the trained model in the trained model holding unit 103 to obtain predicted values ​​(predicted route deviation value, predicted time deviation of passing the destination). The exclusive area calculation unit 301 then adds these predicted values ​​(predicted route deviation value, predicted time deviation of passing the destination) to each record.

[0151] This yields information as shown in Figure 31. As shown in Figure 32, the predicted route deviation value is calculated as a predicted value indicating the maximum vertical deviation along the line connecting waypoints (section route), and the predicted endpoint arrival time deviation value is calculated as a predicted value indicating the deviation from the target endpoint arrival time. The concept of deviation is explained with reference to Figure 10. Here, Figure 31 is a diagram showing an example of the data format of the predicted values, and Figure 32 is a schematic diagram to explain the prediction results.

[0152] For each record of the predicted values ​​obtained in this way (predicted path deviation, predicted deviation in arrival time at the endpoint), a safety distance and safety time are added to determine the exclusive area (exclusive area radius, exclusive termination time). The results of determining the exclusive area are shown in Figure 33. Figure 33 is a diagram showing an example of the data format of the calculated exclusive area.

[0153] Next, the exclusive area calculation unit 301 responds to the operation plan formulation unit 304 with the processing result from step S30 shown in Figure 33 (step S31). The operation plan formulation unit 304 receives this response and determines the range for acquiring issued operation plans (issued operation plan acquisition range) using the information shown in Figure 34 as parameters (step S32). Figure 34 is a diagram showing an example of the data format for the issued operation plan acquisition range. These parameters include a time advanced by a safety time relative to the reachable time, a time delayed by a safety time relative to the arrival deadline time, a point (latitude, longitude) extended by a safety range relative to the start point of the requested range, and a point (latitude, longitude) extended by a safety range relative to the end point of the requested range.

[0154] The acquisition range (spatial information acquisition range) determined in this way is the sum of the safety range at the start of the requested range and the safety range at the end of the requested range, as shown in Figure 35, and this is determined as the range of the exclusive area acquisition request. Figure 35 is a schematic diagram showing an example of the spatial information acquisition range determined in Figure 34. Past operation plans will be read out in which either the start or end time overlaps with the above time period and at least a part of the exclusive area falls within this section.

[0155] Next, the operation plan formulation unit 304 requests the issued operation plan management unit 401 to acquire the exclusive area of ​​the issued operation plan, using the exclusive area acquisition request range determined in step S32 as a parameter (step S33).

[0156] The issued operation plan management unit 401 receives this request and calculates the exclusive area of ​​the issued operation plan by searching for and reading the issued operation plan from the operation plans stored in the issued operation plan retention unit 402, whose data format is shown in Figure 16 (step S34). Here, the issued operation plan is searched for and read from the above operation plan that falls within the requested range shown in Figure 35 and that falls within the time between the departure time of the current operation plan candidate, which is advanced by the amount of safety time, and the arrival deadline, which is delayed by the amount of safety time. Note that there may be multiple matching operation plans.

[0157] Here, a list of route issuance numbers, departure waypoint IDs, available departure times, arrival waypoint IDs, arrival deadlines, request management numbers, and their corresponding route and exclusive area lists are read for the route plans that meet the specified criteria.

[0158] Then, the issued operation plan management unit 401 returns the exclusive area of ​​the issued operation plans within the acquisition request range to the operation plan formulation unit 304 as a response (step S35). Here, it responds with a list of operation routes, the waypoint ID of the departure point, the possible departure time, the waypoint ID of the arrival point, the arrival deadline time, the request management number, and the corresponding operation routes and exclusive areas of the operation plans that meet the conditions read in step S34. An example of this list is shown in Figure 36. Figure 36 shows an example of the data format of an operation plan candidate with the exclusive area calculation result of the operation plan candidate added.

[0159] The information shown in Figure 36 is for a candidate route plan, with the result of calculating the exclusive area of ​​the candidate route plan added. The route includes a list of route sections containing the section sequence number, starting waypoint ID, ending waypoint ID, target ending time, and edge ID. The information shown in Figure 36 also includes route deviation prediction values ​​and ending time deviation prediction values ​​as exclusive area information.

[0160] The operation plan formulation unit 304 receives this response and checks for conflicts between the exclusive domain of the candidate operation plan and the exclusive domain of the issued operation plan (step S36). There are two cases in which a conflict is determined at this point.

[0161] The first case is when the combination of starting waypoints and ending waypoints matches between the route records included in the candidate route plan and the route records included in the issued route plan obtained in step S31, and when the exclusive access times overlap. Such a case is shown in Figure 37. Figure 37 is a schematic diagram showing an example of a race condition. Here, the combination of starting waypoints and ending waypoints matching means that the start and end points can be reversed, and it does not matter whether it is an uphill or downhill route. Furthermore, the exclusive access time refers to the period from the end of exclusive access of the record with the section sequence number immediately preceding the section sequence number of the route record of interest to the end of exclusive access of the route record of interest.

[0162] The second case is when the space determined from the exclusive area radius overlaps, even partially, between the route records included in the candidate route plan and the route records included in the issued route plan obtained in step S31, and the exclusive periods also overlap. Such cases are shown in Figures 38 and 39. Figures 38 and 39 are schematic diagrams illustrating other examples of race conditions.

[0163] Refer to Figure 40 for a supplementary explanation of domain competition. Figure 40 is a diagram illustrating domain competition. The space determined from the exclusive domain radius refers to a cylindrical space with hemispherical ends and a radius equal to the exclusive domain radius, surrounding the edges and waypoints of one route section. It is a space where hemispheres of the same radius as the cylinder are attached to both ends of the cylindrical space in Figure 32 (see Figure 10). The exclusive time refers to the period from the end of exclusive access of the record with the section sequence number immediately preceding the section sequence number of the route record of interest to the end of exclusive access of the route record of interest.

[0164] As shown in Figure 40, the system determines whether there is a conflict between the service area candidates by comparing the exclusive area calculation results of the service area candidates with the exclusive area of ​​the issued service area within the requested acquisition range. In determining the conflict, the section occupancy time is calculated from the time obtained by adding the time difference prediction value to the time of adding the target endpoints before and after the section, then safety time is added, and a safety range is added to the predicted route deviation value.

[0165] Then, if any of the above conflict conditions are met, the operation plan formulation unit 304 considers it to be a route conflict or an exclusive area conflict and sets the conflict flag of the operation route record of the operation plan candidate stored in the operation plan candidate temporary storage unit 305 to true.

[0166] On the other hand, if a route plan is obtained in which no conflicts occur, the route plan formulation unit 304 exits the loop processing shown in Figure 20 and proceeds to the next step S37.

[0167] In step S37, the operation plan formulation unit 304 registers as an operation plan candidate to the issued operation plan management unit 401, which does not have an exclusive domain conflict with an issued operation plan. As described above, once an operation plan candidate that does not conflict with an issued operation plan is obtained, the loop processing shown in Figure 20 is exited, and the process of step S37 is initiated.

[0168] In step S37, the operation plan formulation unit 304 requests the issued operation plan management unit 401 to register the information of the items shown in Figure 15 (information of the last operation plan generated at this point among the operation plan candidate stored in the operation plan candidate temporary storage unit 305) as parameters.

[0169] The issued operation plan management unit 401 registers the operation plan by adding the operation plan information (Figure 15) provided in step 37 to the issued operation plan storage unit 402 (step S38). When adding, an operation route issuance number is assigned and added to the operation plan. As a result, the information of the items shown in Figure 16 is stored in the issued operation plan storage unit 402.

[0170] As shown in Figure 41, the Operation Plan Formulation Unit 304 adds the request management number, departure point, and arrival point information received in the operation plan formulation request in step S23. The Operation Plan Formulation Unit 304 also queries the Spatial Information Management Unit 501 by specifying the waypoint ID, and adds the corresponding coordinates to the starting waypoint ID and ending waypoint ID. Then, the Operation Plan Formulation Unit 304 responds to the Operation Plan Issuance Request Management Unit 302 with the data of the items shown in Figure 41 as an operation plan (step S39). Figure 41 is a diagram showing an example of the format of the operation plan data.

[0171] Upon receiving this response, the Operation Plan Issuance Request Management Unit 302 sets an issued flag for the Operation Plan Issuance Request in the Operation Plan Issuance Request Holding Unit 303 (step S40). Here, the operation route issuance number is added to the record (see Figure 12) where the received operation plan and request management number match, and the issued flag is set to true. As a result, the data for the items shown in Figure 42 is generated as an operation plan. Figure 42 shows an example of the format of the operation plan data.

[0172] Next, the operation plan issuance request management unit 302 issues the operation plan by responding to the operation plan issuance requester with the operation plan (step S41). The issued operation plan can have the issued flag removed and may include information on items such as those shown in Figure 43. Figure 43 is a diagram showing an example of the data format of an operation plan.

[0173] As described above, the operation management device 10 according to this embodiment includes a function for learning route deviations, a function for predicting route and passing time deviations (spatial and temporal deviations), and a function for automatically generating operation plans.

[0174] Then, based on these predictions, the operation management device 10 sets an exclusive area around each edge connecting adjacent waypoints included in the candidate operation plan. This improves spatial utilization efficiency compared to cases where no settings are made according to the prediction results, and consequently improves temporal utilization efficiency as well.

[0175] In this case, the trained model corresponding to the trained model 4e described in Embodiment 1 can be generated as follows. Here, we will explain using an example of training with the learning device 4 shown in Figure 4.

[0176] The acquisition unit 4a acquires information in response to a request to generate an operation plan that includes operation request information, including an executed operation plan that has already been executed, and operation results, from among the operation plans that have been generated to include multiple waypoints for the autonomous mobile device included as the target of operation in the operation request information. The operation results are the operation results at the time the executed operation plan was executed. Here, the operation plan is a plan in which an exclusive region is set to surround each edge connecting adjacent waypoints in the multiple waypoints included in the operation plan. Here again, the exclusive region is a region that prohibits other autonomous mobile devices from operating in competition with it.

[0177] The control unit 4b generates a trained model that predicts, based on acquired information, the positional and temporal deviations that may occur when operating according to a candidate route plan that is generated to include multiple waypoints based on new route request information (i.e., other route request information).

[0178] Furthermore, this learning device 4 can perform the learning methods described above. Also, as mentioned above, the learning device 4 can be configured as a computer, in which case the program can be a program that causes the computer to perform the above-described processes.

[0179] Furthermore, the operation management device 10 can set an exclusive period for each candidate operation plan, based on the predicted results, along with an exclusive area, which is a period during which the operation of other autonomous mobile devices for which no other operation plans have been generated is prohibited within the exclusive area. Therefore, compared to setting only an exclusive area, the time utilization efficiency can be further improved. In particular, the length of the exclusive period for each candidate operation plan can be set to be longer in accordance with the length of the time difference predicted for the passage time at each waypoint, thereby further improving the time utilization efficiency.

[0180] Furthermore, in the operation management device 10 according to this embodiment, instead of partitions in space, waypoints are defined at high density, and the operation route is designed by setting the route to pass through these waypoints. In this embodiment, the danger zone (the area that excludes other autonomous mobile devices) is set to surround the operation route, and its size and exclusion period are determined based on the predicted deviation value predicted from learning past route deviation information. As a result, the danger zone can be set to a different size for each section between waypoints in the designed route, based on the prediction results.

[0181] Furthermore, in the operation management device 10 according to this embodiment, non-conflicting candidates are determined to be candidates that operate within the area excluding the pre-configured exclusive region that surrounds each edge included in the generated but incomplete operation plan among the operation plan candidates. Therefore, since this pre-configured exclusive region can also be determined based on the prediction results, the temporal and spatial utilization efficiency can be further improved.

[0182] Regarding the process of formulating a train operation plan, we have given an example where a conflict is determined each time a candidate train operation plan is generated. However, the process is not limited to this example. It is also possible to generate multiple candidate train operation plans and then select the one with the lowest degree of conflict as the train operation plan.

[0183] Furthermore, in this embodiment, non-conflicting candidates are determined to be candidates that operate in areas excluding the pre-configured exclusive area for the pre-configured exclusive area related to the generated but incomplete operation plan among the operation plan candidates. Therefore, since this exclusive period can also be determined based on the prediction results, the temporal and spatial utilization efficiency can be further improved.

[0184] Furthermore, if a non-conflicting candidate is determined to be a candidate that operates within the region of the operation plan candidates, excluding the pre-configured exclusive region related to the generated but incomplete operation plan, the trained model corresponding to the trained model 4e described in Embodiment 1 can be generated as follows. Here, we will explain using an example of training with the learning device 4 in Figure 4.

[0185] The acquisition unit 4a acquires information in response to a request to generate an operation plan that includes operation preference information. This information includes an executed operation plan that has been executed, and the operation record, from among the operation plans generated to include multiple waypoints for the autonomous mobile device included as the target of operation in the operation preference information. The operation record is the operation record at the time the executed operation plan was executed. Here, the operation plan is a plan that can operate in an area excluding the following exclusive area, from among the candidate operation plans that include multiple waypoints generated based on the operation preference information. The exclusive area here refers to a configured exclusive area that surrounds each edge connecting adjacent waypoints in multiple waypoints included in an operation plan that has been generated but is not yet completed for another autonomous mobile device.

[0186] The control unit 4b generates a trained model that predicts the positional and temporal deviations that may occur when operating according to a candidate operation plan that is generated to include multiple waypoints based on other desired operation information, based on the information acquired by the acquisition unit.

[0187] Furthermore, the learning device 4 in this example can perform the learning method described above. Also, as mentioned above, the learning device 4 can be configured as a computer, and in that case, the program can be a program that causes the computer to perform the above-described processing.

[0188] As described above, according to this embodiment, similar to Embodiment 1, it is possible to achieve both collision prevention and improved space utilization efficiency, as well as to realize the effects of each individually provided component. In this embodiment, the problem described as the first issue, that there is no way to automatically perform the series of procedures of designing a route and generating an operation plan coordinated with the routes of other drones 3 simply by setting the start and end points, can be solved and made possible automatically. Furthermore, in this embodiment, the problem described as the second issue, that existing operation route design methods are inefficient in terms of space utilization (both spatially and temporally), making it impossible to operate multiple drones efficiently, can also be solved. In other words, according to this embodiment, in places where the discrepancy between the planned operation route and the actual route is small, the number of drones that can be accommodated per unit of space can be increased, thus improving utilization efficiency both spatially and temporally.

[0189] <Embodiment 3> Embodiment 3 describes various application examples applicable to Embodiments 1 and 2, focusing on the differences from Embodiment 2. Embodiment 3 can appropriately utilize the various examples described in Embodiments 1 and 2.

[0190] Although Embodiment 2 used a drone as an example, the method can also be applied to autonomous mobile devices that travel on the ground without flying. In that case, instead of the 3D space operation plan exemplified in Embodiment 2, a 2D plane operation plan can be applied for the autonomous mobile device traveling on the ground. In a 2D plane operation plan, information such as upper and lower altitude limits, target coordinates, and actual coordinates of altitude does not need to be used for each section. Furthermore, even when applied to a mobile autonomous device, if used in a multi-story structure such as a multi-story parking garage, a large logistics center, or a factory that can accommodate vehicles, the information can be expressed using floor height information in addition to altitude.

[0191] Furthermore, regarding the placement of waypoints, in Figure 9, the waypoints are placed at equal intervals for the sake of simplicity, but they do not necessarily have to be placed at equal intervals. For example, in cases where it is clear that a space is impassable due to mountains or buildings, or where it is clear that operation is not possible due to legal regulations, there may be places where no waypoints are set from the beginning. In other words, the multiple waypoints included in the operation plan or candidates may be selected and determined from pre-defined spatial points.

[0192] Furthermore, waypoints can also be arranged in the following way to represent spaces that are clearly impassable due to obstacles such as mountains or buildings, or spaces where operation is prohibited due to legal restrictions. That is, as shown in Figure 9, waypoints can be set closely together at equal intervals regardless of obstacles, and spaces that cannot be operated through can be represented by setting the "unavailable" flag to true only at locations that intersect with obstacles. The data structure of the edges in this case is illustrated in Figure 27.

[0193] In other words, the multiple waypoints included in the route plan or candidates may be selected and determined from spatial points that have been predefined as grid points of an equally spaced grid, excluding spatial points where obstacles exist and spatial points that are located in areas where operation is not legally permitted.

[0194] As these examples show, multiple waypoints included in a route plan or candidate can be determined such that the edge lengths differ between at least two edges. This allows for route design that passes through smaller (or larger) sections rather than large, fixed-size sections.

[0195] Furthermore, the exclusive region can also be configured to allow the exclusive line to represent the edge itself.

[0196] Furthermore, in Embodiment 2, week number tags, time zone tags, month tags, and seasonal zone tags were used as explanatory variables for learning in the operation performance learning unit 102 and for predicting positional and time deviations in the exclusive area calculation unit 301. However, these items may be increased or replaced with other items. In addition, if the information is expected to correlate with the target variable, learning data can be obtained not only from operation performance but also from other information sources.

[0197] For example, when applied to an autonomous mobile device operating on land, additional information such as numbered categories of ground surface conditions for each section (paved, unpaved, gravel, sand, etc.) may be added. Similarly, when applied to an autonomous mobile device operating at sea, wave height for each section may be added. Even for autonomous mobile devices operating in the air, weather information for each section may be added. Furthermore, the information obtained from operational performance described in Embodiment 2 may have additional items. For example, sensors may be mounted on the autonomous mobile device, and information such as vibration, temperature, fluctuations in the output of the motors or engines used for propulsion, tilt, and fuel consumption may be added as explanatory variables. This can improve prediction accuracy and build a learning model that more closely reflects actual conditions.

[0198] Furthermore, in Embodiment 2, an example was shown where the objective variable was the positional displacement distance when learning in the operation performance learning unit 102 and predicting the positional and time deviations in the exclusive area calculation unit 301. However, the direction of the positional deviation may also be used as the objective variable. In other words, the operation management device 10 can also be configured to perform predictions of distance deviation and directional deviation as predictions of positional deviation. This makes it possible to exclusively isolate space in a specific direction, further increasing the efficiency of space utilization.

[0199] Furthermore, while Embodiment 2 does not mention the machine learning methods used for learning in the operation performance learning unit 102 or for predicting positional and time deviations in the exclusive area calculation unit 301, multiple regression analysis, for example, can be used. However, other machine learning methods, such as deep learning, may be used for the learning method.

[0200] <Embodiment 4> Embodiment 4 will be explained again, focusing on the differences from Embodiment 1, with reference to Figure 1 once more. Various examples described in Embodiments 1 to 3 can be used as appropriate in Embodiment 4.

[0201] The operation plan generation device 1 according to this embodiment includes a control unit 1a. In response to an operation plan generation request that includes operation request information, the control unit 1a generates an operation plan including multiple waypoints for the autonomous mobile device included as the target of operation in the operation request information. In this embodiment, the storage unit 1b does not need to store data such as operation results, and can be used for temporary storage of intermediate processing data and generated data, but it is sufficient to have a storage unit within the control unit 1a.

[0202] Based on the desired operation information, the control unit 1a generates a candidate operation plan that includes multiple waypoints. The control unit 1a sets an exclusive area for the candidate operation plan as an area where other autonomous mobile devices are prohibited from operating in competition. Here, a feature of this embodiment is that the exclusive area is not based on prediction results, but is set to surround each section route connecting adjacent waypoints in the multiple waypoints included in the candidate operation plan. The control unit 1a then selects a candidate from the candidate operation plan that does not conflict with an operation plan that has already been generated but is not yet completed for another autonomous mobile device, and uses that as the operation plan to respond to the operation plan generation request.

[0203] Thus, in this embodiment, unlike Embodiment 1, exclusive areas are not set using prediction results, but exclusive areas can be set to enclose each section of the route. Since it is possible to set exclusive areas in more detail compared to the comparative example, according to this embodiment, it is possible to automatically generate an operation plan that increases the number of autonomous mobile devices that can operate within a certain space in the same time period, thereby achieving both collision prevention and improved space utilization efficiency. Although not explained in this embodiment, such operation plan generation methods and processing programs can also be employed.

[0204] <Other Embodiments> [a] In each embodiment, the functions of the operation plan generation device, the autonomous mobile device, the learning device, etc., have been described. However, the devices are not limited to the example configurations shown, and it is sufficient if each device can realize these functions.

[0205] [b] Each device according to each embodiment may have the following hardware configuration. Figure 44 shows an example of the hardware configuration of the device. The same applies to the other embodiment [a] described above.

[0206] The device 1000 shown in Figure 44 may include a processor 1001, a memory 1002, and an interface 1003. The processor 1001 may be, for example, a microprocessor, an MPU (Micro Processor Unit), or a CPU. The processor 1001 may include multiple processors. The memory 1002 is composed of, for example, a combination of volatile memory and non-volatile memory. The functions of each device described in each embodiment are realized by the processor 1001 reading and executing a program stored in the memory 1002. In this case, information input and output can be performed via the interface 1003, such as a communication interface that communicates with other internal parts or other external devices. For example, if the device 1000 is a route planning device or a learning device, the interface 1003 may include at least a communication interface and may include a user interface that accepts user operations. For example, if the device 1000 is an autonomous mobile device, the interface 1003 may include at least interfaces with each sensor and a communication interface.

[0207] In the examples described above, the program includes a set of instructions (or software code) that, when loaded into a computer, cause the computer to perform one or more of the functions described in the embodiments. The program may be stored on a non-temporary computer-readable medium or a physical storage medium. Examples, but not limited to, include random-access memory (RAM), read-only memory (ROM), flash memory, solid-state drive (SSD) or other memory technologies, CD-ROM, digital versatile disc (DVD), Blu-ray® disc or other optical disc storage, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage devices. The program may be transmitted over a temporary computer-readable medium or a communication medium. Examples, but not limited to, include temporary computer-readable medium or a communication medium that includes electrically, optically, acoustically, or otherwise propagating signals.

[0208] This disclosure is not limited to the embodiments described above, and may be modified as appropriate without departing from its spirit. Furthermore, this disclosure may be implemented by combining the respective embodiments as appropriate.

[0209] Some or all of the above embodiments may also be described as follows, but are not limited to the following: (Note 1) A control unit that, in response to a request to generate a route plan including route preference information, generates a route plan including multiple passing points for an autonomous mobile device included as a target for operation in the route preference information, A storage unit that stores the relationship between an executed operation plan, which is an executed operation plan, and the actual operation results at the time of execution of the said executed operation plan, Equipped with, The control unit, Based on the aforementioned route preference information, a candidate route plan including multiple passing points is generated. Based on the aforementioned relationship, the candidate operation plan is used to predict the positional and temporal deviations that may occur if the operation is carried out according to the candidate operation plan. Based on the predicted results, an exclusive zone is set for the aforementioned candidate operation plan, which is an area where other autonomous mobile devices are prohibited from operating in competition. Among the aforementioned candidate operation plans, the one that does not conflict with an operation plan that has already been generated but is not yet completed for another autonomous mobile device is selected as the operation plan to respond to the operation plan generation request. Train schedule generation device. (Note 2) The control unit, with respect to the candidate operation plan, sets the exclusive area to surround each of the section routes connecting adjacent passing points in the plurality of passing points included in the candidate operation plan, based on the predicted results. The operation plan generation device described in Appendix 1. (Note 3) The control unit, with respect to the candidate operation plan, sets an exclusive period, along with the exclusive region, based on the predicted results, which is a period during which other autonomous mobile devices are prohibited from operating in competition with the exclusive region. The operation plan generation device described in Appendix 2. (Note 4) The control unit determines, as a candidate that does not conflict with the generated and incomplete operation plan, to be the response operation plan, which is an operation plan candidate that operates within a region excluding the pre-configured exclusive region that encloses each of the section routes connecting adjacent passing points in the multiple passing points included in the generated and incomplete operation plan. A train schedule generation device as described in any one of the appendices 1 to 3. (Note 5) The control unit determines, as a candidate that does not conflict with the generated and incomplete operation plan, a candidate from among the operation plan candidates that operates in an area excluding the set exclusive area during the set exclusive period set for the set exclusive area, and assigns this to the response operation plan. The operation plan generation device described in Appendix 4. (Note 6) The size of the plane perpendicular to the direction of travel in the aforementioned exclusive area is set to differ for at least two of the section paths connecting adjacent passing points in the plurality of passing points. A train operation plan generation device as described in any one of the items 1 to 5 of the appendix. (Note 7) The aforementioned exclusive area is set up to allow the section path connecting adjacent passage points in the plurality of passage points to be an exclusive line representing the section path itself. A train schedule generation device as described in any one of the appendices 1 to 6. (Note 8) The aforementioned plurality of passing points are determined such that, with respect to the section paths connecting adjacent passing points, the lengths of at least two of the section paths differ. A train schedule generation device as described in any one of the appendices 1 to 7. (Note 9) The aforementioned multiple passing points are determined by selecting from spatial points predetermined as grid points of an equally spaced grid, excluding spatial points where obstacles exist and spatial points that are included in areas where operation is not legally permitted. A train schedule generation device as described in any one of the appendices 1 to 8. (Note 10) The aforementioned multiple passing points are determined by selecting from a set of spatial points. A train schedule generation device as described in any one of the appendices 1 to 8. (Note 11) The exclusive period, which is a period during which other autonomous mobile devices are prohibited from operating in competition with the exclusive area, is set to be different for at least two of the section routes connecting adjacent passage points in the plurality of passage points. A train schedule generation device as described in Appendix 3 or 5. (Note 12) The control unit sets the length of the exclusion period for the candidate route plan to be longer in accordance with the length of the time difference predicted for the passage time at each passing point. The operation plan generation device described in Appendix 3. (Note 13) The control unit performs predictions of distance deviation and directional deviation as a prediction of positional deviation. A train schedule generation device as described in any one of the appendices 1 to 12. (Note 14) The aforementioned relationship is stored in the memory unit as a trained model that predicts the positional and temporal deviations that may occur when operating according to the candidate operation plan, which is generated by machine learning based on the actual operation plan and the actual operation results. The control unit performs the prediction using the trained model. A train schedule generation device as described in any one of the appendices 1 to 13. (Note 15) In response to a request to generate a route plan that includes route preference information, the system includes a control unit that generates a route plan that includes multiple passing points for autonomous mobile devices included as targets in the route preference information. The control unit, Based on the aforementioned route preference information, a candidate route plan including multiple passing points is generated. With respect to the aforementioned candidate operation plan, an exclusive area is set to enclose each section of the route connecting adjacent passing points in the multiple passing points included in the candidate operation plan, as an area where other autonomous mobile devices are prohibited from operating in competition with it. Of the aforementioned candidate operation plans, the one that does not conflict with an operation plan that has already been generated but is not yet completed for another autonomous mobile device is selected as the operation plan to respond to the operation plan generation request. Train schedule generation device. (Note 16) In response to a request to generate an operation plan that includes operation preference information, an acquisition unit acquires information including an executed operation plan that has been executed among the operation plans generated to include multiple waypoints for the autonomous mobile device included as the target of operation in the operation preference information, and the operation record at the time of execution of the executed operation plan. A control unit generates a trained model that predicts the positional and temporal deviations that may occur when operating according to a candidate operation plan that includes multiple passing points based on other desired operation information, based on the information acquired by the acquisition unit. Equipped with, The aforementioned operation plan is a plan in which, for each section route connecting adjacent passing points in a plurality of passing points included in the operation plan, an exclusive area is set as an area that prohibits other autonomous mobile devices from operating in competition with the section route. Learning device. (Note 17) In response to a request to generate an operation plan that includes operation preference information, an acquisition unit acquires information including an executed operation plan that has been executed among the operation plans generated to include multiple waypoints for the autonomous mobile device included as the target of operation in the operation preference information, and the operation record at the time of execution of the executed operation plan. A control unit generates a trained model that predicts the positional and temporal deviations that may occur when operating according to a candidate operation plan that includes multiple passing points based on other desired operation information, based on the information acquired by the acquisition unit. Equipped with, The aforementioned operation plan is, Among the candidate route plans, which include multiple passing points, generated based on the aforementioned route preference information, This plan allows other autonomous mobile devices to operate within an area excluding a pre-configured exclusive area that encloses each of the section routes connecting adjacent passage points at multiple passage points included in the generated and incomplete operation plans. Learning device. (Note 18) The system comprises a flight plan generation device as described in Appendix 14, a learning device, and an autonomous mobile device equipped with a group of sensors that collect information to be included in the flight performance and that can communicate with the flight plan generation device. The learning device is An acquisition unit that acquires information including the aforementioned operational plan and the aforementioned operational results, A learning device-side control unit generates the trained model based on the information acquired by the acquisition unit, Equipped with, Autonomous mobile system. (Note 19) An autonomous mobile system comprising a route plan generation device described in any one of the appendices 1 to 14, and an autonomous mobile device equipped with a group of sensors that collect information to be included in the route performance and that can communicate with the route plan generation device. (Note 20) In response to a request to generate a route plan that includes route preference information, the system includes a process for generating a route plan that includes multiple passing points for autonomous mobile devices included as targets in the route preference information. The aforementioned process is, Based on the aforementioned route preference information, a candidate route plan including multiple passing points is generated. With respect to the aforementioned candidate operation plan, based on the relationship between the executed operation plan (an already executed operation plan) and the actual operation results at the time of execution of the executed operation plan, a prediction of the positional and temporal deviations that may occur if the operation is carried out according to the candidate operation plan is performed. Based on the predicted results, an exclusive zone is set for the aforementioned candidate operation plan, which is an area where other autonomous mobile devices are prohibited from operating in competition. Of the aforementioned candidate operation plans, the one that does not conflict with an operation plan that has already been generated but is not yet completed for another autonomous mobile device is selected as the operation plan to respond to the operation plan generation request. A method for generating a train schedule. (Note 21) In response to a request to generate a route plan that includes route preference information, the system includes a process for generating a route plan that includes multiple passing points for autonomous mobile devices included as targets in the route preference information. The aforementioned process is, Based on the aforementioned route preference information, a candidate route plan including multiple passing points is generated. With respect to the aforementioned candidate operation plan, an exclusive area is set to enclose each section route connecting adjacent passing points in the multiple passing points included in the candidate operation plan, as an area where other autonomous mobile devices are prohibited from operating in competition with it. Among the aforementioned candidate operation plans, the one that does not conflict with an operation plan that has already been generated but is not yet completed for another autonomous mobile device is selected as the operation plan to respond to the operation plan generation request. A method for generating a train schedule. (Note 22) In response to a request to generate an operation plan that includes operation preference information, the system acquires information including the executed operation plan, which is generated to include multiple waypoints for the autonomous mobile device included as the target of operation in the operation preference information, and the operation record at the time of execution of the executed operation plan. A trained model is generated that predicts the positional and temporal deviations that may occur when operating according to a candidate route plan that includes multiple passing points based on other route preference information, based on the acquired information. The aforementioned operation plan is a plan in which, for each section route connecting adjacent passing points in a plurality of passing points included in the operation plan, an exclusive area is set as an area that prohibits other autonomous mobile devices from operating in competition with the said section route. Learning methods. (Note 23) In response to a request to generate an operation plan that includes operation preference information, the system acquires information including the executed operation plan, which is generated to include multiple waypoints for the autonomous mobile device included as the target of operation in the operation preference information, and the operation record at the time of execution of the executed operation plan. A trained model is generated that predicts the positional and temporal deviations that may occur when operating according to a candidate route plan that includes multiple passing points based on other route preference information, based on the acquired information. The aforementioned operation plan is, Among the candidate route plans, which include multiple passing points, generated based on the aforementioned route preference information, This plan allows other autonomous mobile devices to operate within an area excluding a pre-configured exclusive area that encloses each of the section routes connecting adjacent passage points at multiple passage points included in the generated and incomplete operation plans. Learning methods. (Note 24) On the computer, A process that, in response to a request to generate a route plan including route preference information, generates a route plan including multiple passing points for an autonomous mobile device included as a target for operation in the route preference information, Based on the aforementioned route preference information, a candidate route plan including multiple passing points is generated. With respect to the aforementioned candidate operation plan, based on the relationship between the executed operation plan (an already executed operation plan) and the actual operation results at the time of execution of the executed operation plan, a prediction of the positional and temporal deviations that may occur if the operation is carried out according to the candidate operation plan is performed. Based on the predicted results, an exclusive zone is set for the aforementioned candidate operation plan, which is an area where other autonomous mobile devices are prohibited from operating in competition. Of the aforementioned candidate operation plans, the one that does not conflict with an operation plan that has already been generated but is not yet completed for another autonomous mobile device is selected as the operation plan to respond to the operation plan generation request. A program that executes a process. (Note 25) On the computer, A process that, in response to a request to generate a route plan including route preference information, generates a route plan including multiple passing points for an autonomous mobile device included as a target for operation in the route preference information, Based on the aforementioned route preference information, a candidate route plan including multiple passing points is generated. With respect to the aforementioned candidate operation plan, an exclusive area is set to enclose each section of the route connecting adjacent passing points in the multiple passing points included in the candidate operation plan, as an area where other autonomous mobile devices are prohibited from operating in competition with it. Of the aforementioned candidate operation plans, the one that does not conflict with an operation plan that has already been generated but is not yet completed for another autonomous mobile device is selected as the operation plan to respond to the operation plan generation request. A program that executes a process. (Note 26) On the computer, In response to a request to generate an operation plan that includes operation preference information, the system acquires information including the executed operation plan, which is generated to include multiple waypoints for the autonomous mobile device included as the target of operation in the operation preference information, and the operation record at the time of execution of the executed operation plan. This system generates a trained model that predicts potential positional and temporal deviations that may occur when operating according to a candidate route plan that includes multiple waypoints based on other requested route information, using acquired data. Execute the process, The aforementioned operation plan is a plan in which, for each section route connecting adjacent passing points in a plurality of passing points included in the operation plan, an exclusive area is set as an area that prohibits other autonomous mobile devices from operating in competition with the section route. program. (Note 27) On the computer, In response to a request to generate an operation plan that includes operation preference information, the system acquires information including the executed operation plan, which is generated to include multiple waypoints for the autonomous mobile device included as the target of operation in the operation preference information, and the operation record at the time of execution of the executed operation plan. This system generates a trained model that predicts potential positional and temporal deviations that may occur when operating according to a candidate route plan that includes multiple waypoints based on other requested route information, using acquired data. Execute the process, The aforementioned operation plan is, Among the candidate route plans, which include multiple passing points, generated based on the aforementioned route preference information, A plan that can operate in an area excluding a set exclusive area that is set to surround the section path connecting between adjacent passing points at a plurality of passing points included in the gener - ated and unfinished operation plan for other autonomous mobile devices. Program.

Explanation of Signs

[0210] 1 Operation plan generation device 1a Control unit 1b Storage unit 2 Autonomous mobile device 2a Sensor group 2b Communication unit 2c Movement control unit 2d Driving unit 3 Drone 4 Learning device 4a Acquisition unit 4b Control unit 4c Storage unit 4d Unlearned model 4e Learned model 10 Operation management device 101 Learned model management unit 102 Operation performance learning unit 103 Learned model holding unit 200 Performance management unit 201 Operation performance management unit 202 Operation performance holding unit 300 Planning unit 301 Exclusive area calculation unit 302 Operation plan issuance request management unit 303 Operation plan issuance request holding unit 304 Operation plan planning unit 305 Operation plan candidate temporary holding unit 400 Plan management unit 401 Issued operation plan management unit 402 Issued operation plan holding unit 500 Space management unit 501 Space information management unit 502 Waypoint (WP) holding unit 503 Edge information holding unit 1000 devices 1001 Processor 1002 memory 1003 Interface

Claims

1. In response to a request to generate an operation plan that includes operation preference information, an acquisition unit acquires information including an executed operation plan that has been executed among the operation plans generated to include multiple waypoints for the autonomous mobile device included as the target of operation in the operation preference information, and the operation record at the time of execution of the executed operation plan. A control unit that generates a trained model that predicts the positional and temporal deviations that may occur when operating according to a candidate operation plan that is generated to include multiple passing points based on other operation preference information, based on the information acquired by the acquisition unit, Equipped with, The aforementioned operation plan is, Among the candidate route plans, which include multiple passing points, generated based on the aforementioned route preference information, This plan allows other autonomous mobile devices to operate within an area excluding a pre-configured exclusive area that encloses each of the section routes connecting adjacent passage points in multiple passage points included in the generated and incomplete operation plans. Learning device.

2. In response to a request to generate an operation plan that includes operation preference information, the system acquires information including the executed operation plan, which is generated to include multiple waypoints for the autonomous mobile device included as the target of operation in the operation preference information, and the operation record at the time of execution of the executed operation plan. A trained model is generated that predicts the positional and temporal deviations that may occur when operating according to a candidate route plan that includes multiple passing points based on other route preference information, based on the acquired information. The aforementioned operation plan is, Among the candidate route plans, which include multiple passing points, generated based on the aforementioned route preference information, This plan allows other autonomous mobile devices to operate within an area excluding a pre-configured exclusive area that encloses each of the section routes connecting adjacent passage points in multiple passage points included in the generated and incomplete operation plans. Learning methods.

3. On the computer, In response to a request to generate an operation plan that includes operation preference information, the system acquires information including the executed operation plan, which is generated to include multiple waypoints for the autonomous mobile device included as the target of operation in the operation preference information, and the operation record at the time of execution of the executed operation plan. This system generates a trained model that predicts potential positional and temporal deviations that may occur when an operation is carried out according to a candidate route plan that includes multiple waypoints based on other requested route information, using acquired data. Execute the process, The aforementioned operation plan is, Among the candidate route plans, which include multiple passing points, generated based on the aforementioned route preference information, This plan allows other autonomous mobile devices to operate within an area excluding a pre-configured exclusive area that encloses each of the section routes connecting adjacent passage points in multiple passage points included in the generated and incomplete operation plans. program.

Images