Control system, control method, and computer-readable recording medium

Abstract

This control system comprises: a determination means for determining, prior to the start of operation, whether or not a target task can be accomplished, on the basis of plan information for work in which an action is performed on a work object using a work machine, object information about the work object, and information about the work machine; and a modification means for outputting proposal information representing a modified target task obtained by modifying the original target task if it is determined that the original target task cannot be accomplished. On the basis of the proposal information, the control system creates revised plan information representing a plan for controlling the operation of the work machine, and controls the work machine according to the created revised plan information.

Description

Control system, control method, and computer-readable recording medium

[0001] This disclosure relates to a control system, a control method, and a computer-readable recording medium.

[0002] Technologies are known for giving work commands (hereinafter referred to as tasks) to autonomous work machines (hereinafter simply referred to as work machines) and generating or controlling action plans to achieve predetermined goals (hereinafter referred to as target tasks) or actions. However, if the changes or responses to the actions (hereinafter referred to as actions) performed by the work machine on the task object are irreversible, specifically when the task object is an irregularly shaped or deformable object, it is difficult to redo the action if it is inappropriate or the result of the task is inappropriate. In other words, it becomes difficult to continue the work or achieve the goal.

[0003] As a related technology, Patent Document 1 describes sheet metal processing, particularly the bending plan of a workpiece using a bending device. According to the description, the selection and arrangement of tools used in the bending device are determined to achieve the design objectives, and an operation plan is generated.

[0004] Japanese Patent Publication No. 2007-50451

[0005] The system described in Patent Document 1 includes a determination means for determining the time required for manufacturing and feasibility based on the generated operation plan. The feasibility determination is used to exclude operation conditions that are deemed unfeasible or to redesign the workpiece. However, in that system, achieving the goal becomes difficult when the operation is deemed unfeasible.

[0006] One of the objectives of the present invention is to achieve the target task even in tasks where the changes or responses to actions by the work machine are irreversible.

[0007] According to one aspect of this disclosure, the control system includes: a determination means for determining whether the objective task of the work can be achieved before the start of operation, based on planning information for the work performed on a work object by a work machine, object information for the work object, and information about the work machine; and a modification means for outputting proposed information representing a modified objective task in which the objective task has been changed, if it is determined that the objective task cannot be achieved. Based on the proposed information, the system creates modified plan information representing a plan for controlling the operation of the work machine, and controls the work machine according to the created modified plan information.

[0008] According to one aspect of this disclosure, the control method involves a computer determining, before starting operation, whether the objective task of the work can be achieved based on plan information for the work performed on the work object by the work machine, object information for the work object, and information about the work machine. If it is determined that the objective task cannot be achieved, the computer outputs proposed information representing a modified objective task in which the objective task has been changed. Based on the proposed information, the computer creates modified plan information representing a plan for controlling the operation of the work machine, and controls the work machine according to the created modified plan information.

[0009] According to one aspect of this disclosure, a computer-readable recording medium records a program that causes a computer to perform the following actions: determine whether the objective task of the work can be achieved before the start of operation, based on planning information for the work to be performed on a work object by a work machine, object information for the work object, and information about the work machine; output proposed information representing a modified objective task in which the objective task has been changed if it is determined that it cannot be achieved; create modified plan information representing a plan to control the operation of the work machine based on the proposed information; and execute a process to control the work machine in accordance with the created modified plan information.

[0010] The control system, control method, and computer-readable recording medium described herein enable the achievement of the target task even when the changes or responses to actions by the work machine are irreversible.

[0011] It is a diagram for explaining an example of the configuration of the control system according to Embodiment 1. It is a diagram for explaining an example of the data structure stored in the storage device in the control system according to Embodiment 1. It is a diagram for explaining an example of the functional block of the feasibility determination unit in the control system according to Embodiment 1. It is a diagram for explaining an example of the functional block of the plan modification unit in the control system according to Embodiment 1. It is a diagram for explaining an example of the operation of the control system according to Embodiment 1. It is a diagram for explaining an example of the configuration of the control system according to Embodiment 2. It is a diagram for explaining an example of the operation of the control system according to Embodiment 2. It is a diagram for explaining an example of the configuration of the control system according to Embodiment 3. It is a diagram for explaining an example of the functional block of the feasibility determination model in the control system according to Embodiment 3. It is a diagram for explaining an example of the operation of the control system according to Embodiment 3. It is a diagram for explaining an example of the configuration of the control system according to Embodiment 4. It is a diagram for explaining an example of the configuration of the first plan adjustment unit in the control system according to Embodiment 4. It is a diagram for explaining an example of the operation of the control system according to Embodiment 4. It is a diagram for explaining an example of the configuration of the control system according to Application Example 1. It is a diagram for explaining an example of the operation of the control system according to Application Example 1. It is a diagram for explaining an example of the operation of the control system according to Application Example 1. It is a diagram for explaining an example of the operation of the control system according to Application Example 2. It is a diagram showing the configuration of the control system according to Embodiment 5. It is a flowchart showing an example of the operation of the control system according to Embodiment 5.

[0012] Hereinafter, embodiments will be described with reference to the drawings. In the drawings described below, elements having the same function or corresponding functions are denoted by the same reference numerals, and repeated description thereof may be omitted.

[0013] <Embodiment 1> (Description of Configuration) FIG. 1 is a block diagram showing a configuration example of one embodiment of a control system (control device). The components shown in FIG. 1 (planning management device 1, observation device 2, storage device 3, plan adjustment device 4, control device 5, working machine 6) can perform necessary exchanges by wired or wireless communication means. This also applies to other block diagrams.

[0014] The control system 100 shown in FIG. 1 includes a plan management device 1, an observation device 2, a storage device 3, a plan adjustment device 4, a control device 5, and a working machine 6. The plan adjustment device 4 includes a plan information conversion unit 40, a feasibility determination unit 41, and a plan modification unit 42.

[0015] The control system 100 executes a target task for the work target by controlling the working machine 6 by the control device 5 based on the initial plan information input by the plan management device 1 and the information stored in the storage device 3. Information about the work target is acquired by the observation device 2 and input to the plan adjustment device 4. The block diagram shown in FIG. 1 does not limit the physical arrangement. For example, the observation device 2 and the control device 5 may be mounted on the working machine 6. The plan management device 1, the storage device 3, and the plan adjustment device 4 may also be mounted on the working machine 6, or may be provided at separate independent locations as separate entities from the working machine 6.

[0016] The plan management device 1 outputs initial plan information for causing the working machine 6 to execute the target task based on information about the target task and the work target, and information about the working machine 6. The target task is a work command (task) having a predetermined goal or objective to be executed by the working machine 6, and may include a quantitative target value per unit time, or a specific period, per specific area. In the present embodiment, it is assumed that the target task is set at the time of generating the initial plan information, and the input means and input processing are not questioned. However, it is assumed that the information about the target task is included in the initial plan information. The information about the target task is, for example, excavation of a predetermined volume within a designated area, loading, embankment, or shaping into a predetermined shape when the working machine 6 is a construction machine and the work target is earth and sand. Here, since the work target is amorphous, the target task includes continuous real values such as a predetermined volume and a predetermined shape. The information about the work target and the working machine 6 is the information acquired at the time when the plan management device 1 generates the initial plan information, and the initial plan information includes an operation plan in which the working machine 6 can achieve the target task at this time. In the present embodiment, it is assumed that the initial plan information has been generated, and the process of the plan management device 1 generating the initial plan information is not questioned.

[0017] The observation device 2 acquires information about the work object targeted by the objective task, that is, the work object handled by the work machine 6. In this embodiment, the work object may be irregular in shape or a deformable object. Changes or responses to physical actions (hereinafter referred to as "actions") by the work machine 6 may be irreversible. Here, an irreversible response to an action means that when the state of the work object changes due to a particular action of the work machine 6, it is not possible or difficult to return it to its previous state. Changes in the state of the work object include, for example, changes in position or orientation, changes in shape, and changes in properties. However, in this embodiment, "irreversible" does not mean that it is not possible to strictly return it to its previous state, but also includes cases where it is difficult to return it to its previous state by an action of the work machine 6 after the change, or where a predetermined difference occurs. Details and specific examples of work objects will be described later.

[0018] Observation device 2 is an imaging device such as an RGB-D camera or a Time of Flight (ToF) camera, which can acquire three-dimensional information about the work object by combining a depth sensor with a monocular, compound-lens, monochrome, or RGB camera. Alternatively, it may be a device that optically measures distance information to the work object in two or three dimensions, horizontally or vertically to the distance direction. For example, it may be a 3D LiDAR (3D-Light Detection and Ranging) or a radar (Radio Detection and Ranging). Observation device 2 may be a single device as described above or a combination of multiple devices. The installation position, installation direction (angle), and number of observation devices 2 are appropriately determined according to the unique specifications and performance of observation device 2 (e.g., field of view, measurable distance, etc.), the type of work machine 6, and the intended task.

[0019] The format of the information acquired by the observation device 2 may include at least three-dimensional information about the work object, such as a combination of RGB pixel data and depth values ​​or distance information, or a set of three-dimensional position information. The set of three-dimensional position information may be, for example, point cloud data.

[0020] The storage device 3 stores information that is input to the plan management device 1 and the plan adjustment device 4. The storage device 3 is, for example, a device such as a hard disk or a storage medium such as flash memory. The storage device 3 may be built into the plan management device 1 or the plan adjustment device 4. Alternatively, the storage device 3 may be a server device installed in a different location or a storage device on the cloud connected via a network. The storage device 3 may be distributed in multiple locations. For example, some parts may be built-in devices as described above, and others may be external devices. The storage device 3 has an interface for electrical connection with the plan management device 1 and the plan adjustment device 4, and can exchange necessary information. Details and specific examples of the stored information will be described later.

[0021] The planning adjustment device 4 comprises a planning information conversion unit 40, a feasibility determination unit 41, and a planning modification unit 42. Based on the initial planning information output by the planning management device 1, the information about the work object acquired by the observation device 2, the state information about the work machine 6, and the information stored in the storage device 3, the planning adjustment device 4 controls the work machine 6 with the control device 5 and outputs feasible planning information for executing the target task for the work object. The information about the work object is three-dimensional information representing the shape of the work object. For example, if the work object is an irregularly shaped object such as soil, it is height information of the soil surface at a certain location. If the work object is a deformable object such as sheet metal, it is a set of three-dimensional position information. The state information about the work machine 6 is information about the arrangement of the work machine 6, i.e., position and angle information, and information about the control unit of the work machine 6, for example, information such as the rotation angle and length of the movable part. The feasible planning information includes action planning information for causing the work machine 6 to perform a specific action, and may be symbolic information including time-series data or information about sequence. Details and specific examples of feasible plan information will be provided later.

[0022] The control device 5 outputs a control command (or control signal) to control the position and orientation of the control unit of the work machine 6 to achieve a specific operation, based on the feasible plan information output by the planning adjustment device 4. The control unit of the work machine 6 consists of movable parts (actuators) that move according to this control command. If the actuator is controlled by an electrical signal, the control command is an electrical signal, and its value may be output directly to the actuator. If the actuator cannot be controlled electrically, for example, if it is controlled by hydraulics, the control command is a signal from the control unit that controls the hydraulics, or a signal that controls a remote control device such as a surrogate attached to an operating lever that controls the control unit.

[0023] The work machine 6 includes, for example, construction machinery and heavy equipment (hereinafter referred to as construction machinery), robots, and transport vehicles (AGV: automatic guides vehicle, AMR: autonomous mobile robot). However, it is not limited to these. Construction machinery includes, for example, power shovels, backhoes, cranes, and forklifts. Robots include articulated robot arms, robots with multiple arms, and arms mounted on transport vehicles. The work machine 6 has movable parts for performing a target task on the work object, and these movable parts are controlled by external control signals. Therefore, hereinafter, the movable parts will also be referred to as the controlled parts.

[0024] Examples of target tasks performed by the work machine 6 are described below. For example, if the work machine 6 is construction equipment, especially an excavator or backhoe, it may perform tasks such as excavating soil, gravel, crushed stone, wood, waste materials, or natural ground and moving them to another location or loading them onto a dump truck. In this case, the objectives would be, for example, the amount of material processed per unit time or the area processed. If the work machine 6 is a robot, especially one that includes a robotic arm, it may perform tasks such as picking up objects and moving them to another location, a task known as pick and place. In this case, the objectives would be, for example, the amount of material processed per unit time.

[0025] Next, we will describe the components of the plan adjustment device 4: the plan information conversion unit 40, the feasibility determination unit 41, and the plan modification unit 42.

[0026] The planning information conversion unit 40 acquires the initial planning information output by the planning management device 1 and outputs planning information that causes the work machine 6 to execute a series of actions. The planning information may be time-series data or symbolic information including information about the sequence, and preferably it is represented as abstract state information for the work machine 6. Here, the abstract state of the work machine 6 represents each state of the work machine 6 in an abstract model of the work machine 6. The abstract model of the work machine 6 is a model that mathematically represents the state of the actual work machine 6 using geometric information, components, movable parts, etc. Therefore, the abstract state of the work machine 6 specifically includes the position and orientation of the work machine 6, the displacement of each movable part, etc. These abstract states may be set as the initial solution to the optimization problem described later.

[0027] The feasibility determination unit 41 acquires the plan information output by the plan information conversion unit 40, the initial plan information output by the plan management device 1, the information about the work object acquired by the observation device 2, the state information about the work machine 6, and the information stored in the storage device 3, and outputs a determination value representing feasibility and, if feasible, feasible plan information. The information about the work object is represented as an abstracted state (abstract state) based on the information acquired from the observation device 2. Abstraction of the work object means that the properties that determine the behavior and characteristics of the work object are extracted. Specifically, it is represented by information such as the position, posture, shape, size, and contour of the work object. Details and specific examples will be described later. The state information about the work machine 6 is represented as an abstract state based on information about the work machine 6 acquired directly from the work machine 6 or via the control device 5. The determination value representing feasibility may be output as, for example, Feasible (logical value: 1) if feasible, or Infeasible (logical value: 0) if not feasible. Alternatively, the output may be a continuous value rather than a logical value, and the result may be compared with a pre-set threshold for determination. If it is determined to be feasible, feasibility planning information is output to the control device 5. The feasibility determination process is based on the solution result of the optimization problem described later. Feasibility planning information is information in which a part of the initial planning information output by the planning management device 1 has been modified, and is information for enabling the work machine 6 to realize the target task. Specifically, it is time-series information for controlling the work machine 6 by the control device 5 to execute a series of operations, and includes logical information representing the specification and order of operations, and continuous value information representing the content of the operations.

[0028] The plan modification unit 42 operates when the feasibility determination unit 41 determines that the plan is not feasible. A detailed operation flow will be described later. Based on the solution result of the feasibility determination unit 41 and the plan information output by the plan information conversion unit 40, the plan modification unit 42 modifies the initial solution or the settings of the optimization problem that are allowed to be changed. After that, the feasibility determination unit 41 returns to the process of determining feasibility, i.e., the process of solving the optimization problem. At this time, the initial solution or the settings of the optimization problem that were modified by the plan modification unit 42 are applied. Specific examples of the initial solution and the settings of the optimization problem that are allowed to be changed, and how to change them, will be described later.

[0029] Next, the structure of the information stored in the storage device 3 will be explained. Figure 2 shows an example of the structure of the information (data) stored in the storage device 3. The stored information includes at least model information 3a and task information 3b. The model information includes, for example, work machine model information Ia1, target model information Ia2, response model information Ia3, etc. The task information includes, for example, constraint information Ib1, subtask information Ib2, action information Ib3, etc. This information may be stored in advance or updated later.

[0030] (Model Information) The model information 3a stored in the storage device 3 is referenced in the planning and adjustment device 4 when setting up abstract models for the work machine 6 and the work object. A specific example is described below.

[0031] The work machine model information Ia1 is information about the work machine 6. The work machine 6 can control a controlled unit by a control signal to achieve a desired task. The work machine model information Ia1 may include, for example, geometric configuration information of the work machine 6, such as the number, configuration, length, and angle of movable parts. The work machine model information Ia1 may also include information that allows calculation of the relationship between the control signal and the controlled unit, that is, the movement of the controlled unit when a control signal is input. In other words, this information is model information that abstracts the actual work machine 6. From the geometric information and the information on the movement of the controlled unit in response to the control signal, the expected movement (dynamics) of the work machine 6 can be calculated, for example, by forward kinematics. That is, the operation can be simulated without actually operating the work machine 6. The model of the work machine 6 does not need to include detailed shape, properties such as color and material, or the structure of movable parts (actuators), and may be simplified or simplified depending on the type of work machine 6, the task, and the work object. For example, only the movable parts involved in the task may be modeled, and information about other parts may be omitted.

[0032] The target model information Ia2 is model information that abstracts the work object, that is, a model that mathematically represents the state of the work object. It may be modeled based on information observed by the observation device 2. Specifically, it may include candidate mathematical formulas or functions that approximately represent the surface shape of the work object, candidate algorithms for approximation and interpolation, and information about trained models represented by neural networks, etc. The target model information Ia2 stores candidate information necessary to abstractly represent the work object, and may be selected from among this information when used.

[0033] Response model information Ia3 is information about the response when the work machine 6 acts on the work object. The work machine 6 acts on the work object based on feasible planning information output by the planning adjustment device 4 in order to achieve the desired task. The specific action may involve moving the position of the work object or changing its shape, but this depends on the type of work machine 6 and the task. Response model information Ia3 is used by the planning adjustment device 4 when determining feasibility. In other words, the response of the work object can be simulated without actually acting on the work object with the work machine 6. Therefore, even if the work object exhibits an irreversible response to the action of the work machine 6, the response can be calculated using the response model without actually changing the work object. The response model only needs to be abstracted to the necessary extent and accuracy according to the properties of the work object, such as its size, shape, and hardness, and the action applied by the work machine 6.

[0034] (Task Information) The task information 3b stored in the memory device 3 is referenced by the planning adjustment device 4 when determining feasibility. Specifically, it is referenced in the optimization calculation described later. A specific example is given below.

[0035] Constraint information Ib1 is information that indicates the conditions that must be met when the work machine 6 performs a task. For example, for the work machine 6, it may include conditions that define the range of motion and operating speed, and conditions that ensure safety by preventing the work machine 6 from colliding with other devices or structures. Regarding the relationship between the work machine 6 and the work object, it may include conditions that define the magnitude and range of the effect on the work object. Furthermore, constraint information Ib1 may also include conditions that represent the properties of the work object, such as the range of size, shape, and hardness. This constraint information Ib1 may be defined as numerical data (absolute / relative values) or as mathematical formulas (inequalities and equations). The conditions indicated by this information may also be defined as propositions (a form in which the truth value of a statement or expression can be determined).

[0036] Subtask information Ib2 includes information for determining the types of subtasks necessary to achieve a goal for a given task, as well as information defining the subtasks. In other words, a subtask refers to a predetermined action obtained by breaking down a task, and is defined as subtask information Ib2. Subtask information Ib2 is referenced by the planning adjustment device 4 when generating feasible planning information. Specifically, it is used when selecting and determining the order of subtasks, and determining the start and end times. The information defining a subtask may include information for determining the types of actions necessary to perform a given subtask, as well as information for determining multiple combinations. For example, for a given subtask, this could include data indicating candidate action types, information regarding rules and constraints on the order between actions, and information such as parameters for specifying actions. This subtask information Ib2 may be stored in the form of data representing the relationship between tasks, subtasks, and actions, such as table data or directed / undirected graphs.

[0037] Action information Ib3 includes information for determining the type of motion required to execute an action, multiple combinations thereof, and information defining the motion. Action information Ib3 is referenced by the planning adjustment device 4 when generating feasible planning information. Specifically, it is used when selecting actions, determining their order, and determining parameters. The information defining the motion is information regarding the association between motion and control signals for executing each motion. Therefore, action information Ib3 may also be used by the control device 5 when generating control commands for the work machine 6 to execute each action. This action information Ib3 may be stored in the form of data representing the relationship between actions, motions, and control signals, such as table data or directed / undirected graphs.

[0038] (Explanation of Function) Next, the function of the planning adjustment device 4 will be explained.

[0039] The planning and adjustment device 4 implements the function of setting abstract models of the work machine 6 and the work object based on the model information stored in the memory device 3. For example, the planning and adjustment device 4 sets an abstract model of the work object (target model) based on the target information acquired by the observation device 2, an abstract model representing the relationship between the control commands input to the work machine 6 and the operation of the work machine 6 (machine model), and an abstract model representing the effect that the operation of the work machine 6 has on the work object (response model). These models may be expressed in the form of mathematical formulas (functions), or in the form of trained models using deep learning or machine learning such as neural networks. The above distinction between the target model, machine model, and response model is an example representing a functional distinction, and the form expressed as mathematical formulas or trained models does not necessarily have to be these distinctions, i.e., three independent models.

[0040] First, we will explain an example of setting a target model for the work object based on the information acquired by the observation device 2. We will use the example where the work machine 6 is a construction machine, particularly a power shovel or backhoe, and performs a task involving excavation of soil and sand. First, the observation device 2 acquires the state of the soil and sand, which are the work object, as three-dimensional data. At this time, the target model for the soil and sand can be set, for example, as follows. In the following example, the surface on which the work machine 6 is grounded is assumed to be the XY plane, and the three-dimensional data acquired by the observation device 2 is assumed to be the height information of the soil and sand surface within this XY plane, i.e., the Z value. Time is assumed to be represented by discrete time phases (i = 0, 1, ...) under a certain reference time and time width. At this time, the target model is assumed to be the surface shape of the soil and sand, i.e., the two-dimensional position (x) at a certain time phase i. ｉ , y ｉ ) Height of soil z ｉ For example, function f ｉ Using

[0041]

[0042] It can be expressed as follows. Here, the function f represents the height z of the earthwork surface at the two-dimensional position (x, y), that is, the function representing the earthwork surface shape. The function f can be obtained, for example, by function approximation from the three-dimensional data of the earthwork surface, that is, the set data of the height z and the two-dimensional position (x, y). As a method of function approximation, general methods for approximating the earthwork surface, that is, a curved surface, such as polynomial approximation, Gaussian function, radial basis function (RBF), or Kriging method, may be applied. The function f may be represented by a learned model, that is, a neural network. In equation (1), subscripts of i are attached to represent the relationship in time phase i. The function f is stored as the target model information Ia2 stored in the storage device 3, and may be selected according to the work target and task from among the plurality of stored functions and expression forms.

[0043] Next, an example of a machine model representing the relationship between the control command input to the working machine 6 and the operation of the working machine 6 is shown. As the working machine 6, a backhoe is similarly taken as an example. It is assumed that time is represented by discrete time steps (k = 0, 1,...) under a certain reference time and time width. At this time, the state vector X represents the abstracted state of the backhoe at time step k during time phase i. ｉ，ｋ The abstracted state of the backhoe means, for example, the position and orientation of the backhoe in a certain coordinate system, and the displacements of each movable part, specifically, the swing angle of the upper swing body and the angles of each arm part. The state vector is a column vector obtained by arranging these states. Since the dynamics of the backhoe can be expressed by the values of this state vector, the state vector represents an abstracted state (abstract state). The state vector X at the next time step k + 1 is, for example, ｉ，ｋ＋１ It can be expressed as follows. Here, U

[0044]

[0045] It can be expressed as follows. Here, U ｉ，ｋ is a vector representing the control input at time step k during time phase i, (t ｉ，ｋ＋１ ​​) is the time step width at time step k during time phase i. Equation (2) represents the machine model because it describes the update of the abstract state of the backhoe from time step k to k+1 using the control input included in the control command information. Equation (2) above assumes that time phase i, which represents the change in the soil being worked on, is longer than time step k, which represents the change in the state of the work machine 6. That is, there are multiple time steps k within time phase i. This setting is valid when the state change of the work machine 6 is relatively faster than the state change of the work object, but in practice it is not limited to this and can be changed as appropriate depending on the type of work machine 6 and the purpose task. The machine model exemplified in equation (2) is stored in the work machine model information Ia1 stored in the memory device 3, and may be selected from among the multiple functions and representation formats stored depending on the work object and task.

[0046] Next, we present an example of an abstract model of the backhoe, which is the work machine 6, when it excavates soil, that is, a response model that represents the effect that the work machine 6 has on the work object. Depending on the operation of the backhoe, such as excavation, in a time step k of a certain time phase i, the function f of equation (1) ｉ The soil surface shape represented by the function f represents the soil surface shape in time phase i+1. ｋ＋１ If it changes in this way, this change in the surface shape of the soil can be described, for example, using the function F,

[0047]

[0048] It can be expressed as follows: Here, θ ｆ This parameter expresses the properties of the soil and changes in the soil surface shape, such as the viscosity and degree of diffusion of the soil. This parameter may also include parameters that characterize the conditions and trajectory of the backhoe during excavation. Among these parameters, those that depend on the site environment are specifically called site parameters and may be adjusted as appropriate according to the site environment. Specific methods for adjusting site parameters will be described later. X in equation (3) ｉ，ｋThis represents the state vector of the backhoe and expresses a response model that shows the effect the work machine 6 has on the work object. Information about the response model and parameters exemplified in equation (3) is stored in the response model information Ia3 of the storage device 3, and may be selected from among the multiple functions and representation formats stored, depending on the work object and task.

[0049] The above examples of abstract models (target model, machine model, and response model) set by the planning adjustment device 4 are illustrated in equations (1) to (3) using a backhoe as the work machine 6, soil as the work object, and a task involving excavation. However, each of these equations is a simplified example, and in practice, the models are not limited to these and can be set appropriately depending on the type of work object, the type of work machine 6, and the objective task.

[0050] Preferably, the settings for each of the abstract models described above are fixedly set by the planning and adjustment device 4 according to the type of work machine 6, the work object, and the content of the task, except for the field parameters. In other words, they can be pre-set in the planning and adjustment device 4 and do not need to be set in each operation flow. They may also be used in common in other functional blocks of the planning and adjustment device 4.

[0051] Next, the function of the feasibility determination unit 41 of the plan adjustment device 4 will be explained.

[0052] The feasibility determination unit 41 implements the function of outputting a feasibility determination value and feasible plan information if feasible, based on the initial plan information output by the plan management device 1, the plan information output by the plan information conversion unit 40, the information about the work target acquired by the observation device 2, the status information about the work machine 6, and the information stored in the storage device 3. Figure 3 shows an example of the functional blocks of the feasibility determination unit 41. The feasibility determination unit 41 includes a target logic formula generation unit 43 and a plan determination unit 44.

[0053] The target logic formula generation unit 43 generates a target logic formula based on the initial planning information output by the planning management device 1 and the abstract model set by the planning adjustment device 4. The target logic formula is expressed as a logical formula or proposition that represents the target achievement state of the task. In practice, it is expressed as an equality or inequality using variables. The conditions for achieving the objective task and the constraint information Ib1 stored in the storage device 3 may be combined into a single logical formula and expressed as the target logic formula. The target achievement state of the task is defined as the objective task and is included in the initial planning information output by the planning management device 1.

[0054] In the following explanation, similar to the description of the function of the planning adjustment device 4 above, we will use the example of a case where the work machine 6 is a construction machine, particularly a backhoe, and performs a task involving the excavation of soil and sand. In this case, the objective task is to "repeatedly perform the operation of excavating the soil to be worked on, and when the excavated volume is a positive value, releasing the excavated material into a predetermined release area, thereby satisfying the total excavated amount VT in a certain unit time Tu."

[0055] The above objective task can be expressed, for example, by the following three propositions: • Proposition φ １ "The volume excavated by the backhoe is positive." - Proposition φ ２ "Ultimately, the backhoe bucket reaches the area where the soil is being discharged." - Proposition φ ３ "Total excavation volume V per unit time Tu" Ｔ Since these three propositions must be satisfied simultaneously, when expressed using modal logic: ∧ (and) and ∨ (or), the propositions that must be satisfied are "φ １ ∧φ ２ ∧φ ３ It can be expressed as ".

[0056] proposition φ １ This is the condition that "the volume excavated by the backhoe is positive." This condition is an example of how to express it based on the fact that excavation is the action of scooping up soil and sand, and that it does not constitute excavation unless a finite volume is scooped up. If V[i] is the volume excavated in time phase i (where i is an integer greater than or equal to 1), then proposition φ １ teeth,

[0057]

[0058] This can be expressed as follows. Here, we assume that excavation is performed for each discretized time phase i and that a volume V[i] is obtained, so the time phase i corresponds to the number of excavations. Since the excavation operation causes a change in the soil shape, the number of excavations corresponds to the update of the soil shape expressed in equation (3). The reason why the excavated volume is V[i-1] in time phase i-1 is because of proposition φ ２ This is to take into account the order of these elements.

[0059] proposition φ ２ The condition is that "the backhoe bucket eventually reaches the soil discharge area." The soil discharge area is the area where the soil scooped up by the bucket is dropped. For example, the backhoe bucket reaching the soil discharge area is when the backhoe bucket position P ｂｕｃ and the release position P ｌоａｄ The difference in distance from is a constant value r ｌоａｄ It can be expressed under the following conditions. Here, proposition φ １ Consider the order relationship with Proposition φ. １ Therefore, in phase i-1, the backhoe will have scooped up a finite amount of soil. ２ This indicates that in the next phase i, the bucket position will eventually reach the release position. This is based on the order of the propositions, which states that the soil must be scooped up before the bucket position reaches the release position. That is, proposition φ in equation (4) １ is the proposition φ ２ It must be filled before this. Discharge position P ｌоａｄ and a constant value r of the difference in distance ｌоａｄ Assume that this is given in advance.

[0060] Here, we will explain how to express the proposition "ultimately" mathematically. The backhoe bucket only needs to reach the soil discharge area during the time phase i following the scooping up a finite amount of soil, during which time the backhoe operates in time steps k represented by equation (2). Therefore, if time phase i is k = 1 to n (where n is a natural number greater than or equal to 2), that is, if k = n represents the last time step of each time phase i, then the backhoe bucket only needs to reach the soil discharge area by the last time step n.

[0061] As a way to mathematically describe such time constraints, one might introduce a representation using Signal Temporal Logic (STL). Signal Temporal Logic (hereinafter referred to as STL) is a logical system that uses temporal logic operators such as F (Eventually, eventually) and G (Always). Using this representation, it is possible to describe continuous real-valued constraints. For example, the above condition, "by the last time step n," can be expressed using the F (Eventually, eventually) operator:

[0062]

[0063] This can be expressed as follows. In equation (5), the distance is expressed in terms of the square (norm), and the time of the time phase i and the time step k of the backhoe are represented as ti and k. Pbuc,k represents the bucket position at time step k. This position can be calculated using forward kinematics from the backhoe state vector Xi,k calculated from equation (2) and the work machine model information Ia1 stored in the memory device 3.

[0064] To determine whether the constraints of equation (5) are met, first consider the excavation position P. ｌоａｄ and a constant value r of the difference in distance ｌоａｄ Given the values, we calculate the norm. Next, we describe how to determine the F operator. In STL representation, the signal ζ is at time t ｋ We introduce a function ρ to determine whether the proposition φ is satisfied, and if its value is 0 or greater, that is,

[0065]

[0066] Therefore, it is determined that proposition φ is satisfied. Based on this equation (6), time t ｋ＋a ~t ｋ＋ｂ Within the range, the F operator that ultimately satisfies proposition φ is,

[0067]

[0068] This can be written as follows. Using this equation (7), equation (5) can be determined.

[0069] Next, the proposition φ １ and φ ２ "φ" that satisfies both conditions simultaneously １ ∧φ ２ Let's explain how to determine this. The operator ∧ representing (and) is also, based on formula (6),

[0070]

[0071] Therefore, using this equation (8), we can write the proposition "φ" that represents the target task. １ ∧φ ２ It is possible to determine that "

[0072] Finally, as the objective logical formula representing the target task, "the total excavation amount V in a certain unit time Tu" Ｔ The proposition φ that satisfies the condition ３ An example illustrating this is given. If NT is the number of drilling operations during a certain unit time Tu, then proposition φ ３ For example, using the drilling volume V[i] in phase i,

[0073]

[0074] It can be expressed as follows: Here δ Ｖ The target total excavation volume V Ｔ This represents the allowable error from the number of drilling cycles N. That is, equation (9) is given by Ｔ Total drilling volume equals target drilling volume V Ｔ ±δ Ｖ This is a constraint that indicates that the following conditions must be met.

[0075] The above illustrates the function of the target logic formula generation unit 43, including generating a target task as a target logic formula, describing it using STL representation, and demonstrating a method for making a determination. However, the above generation of the target logic formula and STL representation are merely examples and are not limited to these methods of description.

[0076] Constraints other than the objective task, such as constraint information Ib1 stored in memory device 3, may also be included in the target logical expression by generating a target logical expression, for example, using the ∧ (and) operator in expression (8). Specific examples of constraints other than the objective task include, but are not limited to, constraints that limit the volume that the backhoe can excavate in one pass, constraints that prevent collisions between the backhoe and other heavy machinery, and constraints that specify no-entry zones.

[0077] Next, we will explain the function of the planning determination unit 44, one of the functional blocks of the feasibility determination unit 41 shown in Figure 3. The planning determination unit 44 acquires planning information output by the planning information conversion unit 40, information on the current work target acquired by the observation device 2, and information on the current state of the work machine 6, based on the abstract model set by the planning adjustment device 4 and the target logic formula generated by the target logic formula generation unit 43, and implements the function of outputting a feasibility determination value and feasible planning information if feasible. Hereinafter, the abstract model set by the planning adjustment device 4, which is equation (2) representing the dynamics of the backhoe and equation (3) representing the dynamics of the soil, will be referred to as the abstract model Σ. The target logic formula generated by the target logic formula generation unit 43 is the proposition φ exemplified above. １ ~φ ３ In addition, assuming other constraints are also included, the proposition φ of total Nφ ｊ (j=1 to N) φ Assume that the following is set. The planning determination unit 44 uses these abstract models and target logic formulas as constraints to determine at least the state vector X representing the state of the backhoe. ｉ，ｋ And, vector U representing the control input ｉ，ｋ We construct an optimization problem for the optimization variable Z that includes and and find the optimal solution. For example, assuming that the time phase i is i = 1 to m with m (a natural number greater than or equal to 2) as the upper limit, and that each time step k in time phase i is k = 1 to n, then the optimization problem for all time steps k = 1 to nm is, for example,

[0078]

[0079] It can be expressed as follows: Here, J(Z) is the evaluation function of the optimization problem, for example,

[0080]

[0081] This can be expressed as follows: Here, Q is a matrix that determines the weights of the control input U, and tk is the time step. In other words, equation (10) represents an optimization problem that minimizes the evaluation function J in equation (11), and specifically, it seeks to find the optimization variable Z such that the control input U and the time step are minimized. The optimization variable Z in equation (10) includes volume V, which is because constraints on volume V are set as in equations (4) and (9). The reason that the optimization variable Z includes the time step t is because the evaluation function J includes the time step t. This is aimed at improving work efficiency by finding a solution such that the value of the time step is minimized, i.e., the work time is minimized.

[0082] Here, when solving the optimization problem in equation (10), a portion of the initial solution (initial value) of the optimization variable Z is set based on the output of the planning information conversion unit 40. As described in the explanation of the planning information conversion unit 40, the planning information conversion unit 40 receives the initial planning information output by the planning management device 1 and outputs time-series values ​​(planning information) for each state in the abstract model of the work machine 6. These values ​​correspond to the state vector X representing the state of the backhoe, which is included in the optimization variable Z expressed in equation (10). Therefore, the values ​​of the state vector X at each time step before starting optimization can be set from the output of the planning information conversion unit 40. Of the variables included in the optimization variable Z, those not included in the information output by the planning information conversion unit 40 are set to other values ​​in advance.

[0083] Based on the above, the plan determination unit 44 solves the optimization problem of equation (10) using the information output by the plan information conversion unit 40, that is, the plan information based on the initial plan information output by the plan management device 1, as the initial solution. At this time, the method and means of solving (solver, library, etc.) are not specified in this embodiment and can be appropriately selected depending on the problem. If a feasible solution Z is obtained as a result of the solution, the plan determination unit 44 determines that it is feasible and outputs feasible plan information based on the obtained feasible solution. On the other hand, if a feasible solution is not obtained, it determines that it is not feasible.

[0084] Here, the feasible solution obtained through optimization, i.e., the feasible plan information generated based on the optimization variable Z, may be in the same format or have the same amount of information as the initial plan information output by the planning management device 1. In other words, the feasible plan information may be information in which some of the information of the initial plan information has been changed. When generating the feasible plan information, the subtask information Ib2 and action information Ib3 stored in the storage device 3 may be referred to.

[0085] This section describes an example of generating feasible plan information from a feasible solution Z, where the objective task is excavation of a predetermined volume using a backhoe. The feasible solution Z includes the backhoe's state vector X and control input vector U at each time step. For example, a time step in the control input vector U where the value related to the backhoe's movement (travel) is a finite value and the other values ​​are zero can be considered a movement subtask. On the other hand, a time step where the values ​​related to each arm of the backhoe are finite values ​​and the value related to movement is zero can be considered an excavation subtask. The target value (target coordinate), which is one of the specifications defining the movement subtask, can be considered the value of the state vector X at the final time step of the movement subtask. The excavation point, which is one of the specifications defining the excavation subtask, can be considered the coordinate where the backhoe's bucket touches the soil surface, which is one of the values ​​of the state vector X. The definitions of these subtasks may also refer to the subtask information Ib2 stored in the storage device 3. Similarly, the actions required to perform each subtask can also be determined by referring to the action information Ib3 stored in the memory device 3, and using the values ​​of the state vector X and the control input vector U.

[0086] The above describes an example in which the state vector X and control input vector U of the backhoe are set as the optimization variable Z, and feasible planning information is generated from the obtained solution. However, this problem setting depends on the abstract model set in the planning adjustment device 4. For example, a different abstract model may be set by adding a logical variable (or switching variable) that represents the selection of a subtask to the backhoe dynamics expressed in equation (2) and the response between the backhoe and soil expressed in equation (3). For example, a logical variable that takes a logical value of 1 when the excavation subtask is performed and a logical value of 0 in other cases, i.e., the movement subtask, can be introduced. By including this logical variable in the optimization variable Z, the logical variable value as the optimal solution can be obtained. Therefore, the movement and excavation subtasks can be determined from the logical value 0 / 1 of this obtained logical variable. In this way, there is a degree of freedom in setting the abstract model and the optimization problem, so it is not limited to the above example, and can be set appropriately according to the target task and the type of work machine 6.

[0087] Next, the functions of the plan modification unit 42 will be explained. Figure 4 shows an example of the functional blocks of the plan modification unit 42. The plan modification unit 42 has a search unit 45 and a proposal unit 46, and when the feasibility determination unit 41 determines that it is not feasible, it modifies the initial solution and the settings of the optimization problem that can be changed, and implements the function of re-executing the solution-finding process by the plan determination unit 44 of the feasibility determination unit 41. When it is determined that a solution cannot be found under predetermined conditions, the plan modification unit 42 implements the function of generating proposal information and outputting it to the plan management device 1.

[0088] The search unit 45 of the plan modification unit 42 implements a function that, when the feasibility determination unit 41 determines that it is not feasible, modifies the initial solution and the settings of the optimization problem that are permitted to be changed, and re-executes the solution-finding process by the plan determination unit 44 of the feasibility determination unit 41. The search unit 45 receives the initial solution output by the plan information conversion unit 40 and the solution-finding result of the feasibility determination unit 41 as input. Here, "permitted to be changed" is defined as "not affecting the target task." The applicable values ​​and ranges are predetermined according to the target task, work machine 6, and work object, and may be included in the information stored in the storage device 3. Specific examples will be described later. The values ​​to be changed and the degree (range) of change may also be predetermined according to the target task, work machine 6, and work object. Alternatively, the values ​​may be changed experimentally to investigate the solution rate of the optimization problem, i.e., to determine them using sensitivity analysis or the like. Methods for changing the values ​​include, but are not limited to, introducing random disturbances using random numbers, providing values ​​sampled from a normal distribution with defined mean and variance, or using exploratory methods. You can change a single value or multiple values ​​simultaneously.

[0089] The number of times the problem-solving process is re-executed (number of iterations) may be predetermined. Alternatively, the decision of whether or not to re-execute may be made based on the solution result, for example, by setting a threshold value for the evaluation function.

[0090] Here, we will explain the differences in the re-execution (repetition) of the solution-finding process. First, as described above, the search unit 45 changes the initial solution and the settings of the optimization problem each time the solution-finding process is performed. It is also possible to acquire information about the work object and the work machine 6 each time the solution-finding process is performed and reflect it in the initial solution. In other words, the timing of acquiring information from the observation device 2 and the work machine 6 is not limited to the first solution-finding process. As a result, by iteratively solving the optimization problem set by the feasibility determination unit 41 under different initial solutions and settings, the possibility of obtaining a feasible solution can be increased.

[0091] The following are specific examples of values ​​that the search unit 45 changes when the target task is excavation of a predetermined volume using a backhoe. For example, the initial position and orientation defined by the backhoe's state vector X are changed. The change is not made significantly from the previously acquired backhoe position and orientation, or it is assumed that the acquired values ​​contained errors. Other examples include parameters that adjust the excavation trajectory, specifically the bucket insertion depth and pull-in amount. The initial values ​​of the logical variables that determine the selection and order of subtasks may also be changed. If these parameters or logical variables are not included in the optimization variable Z, they are added as appropriate.

[0092] As an example of the settings for an optimization problem, if the unit time Tu expressed in equation (9) can be varied within a certain range from the proposition given as the objective task, then the number of drilling iterations, i.e., the number of optimization time steps, may be changed. Alternatively, the allowable error δ from the target drilling amount in equation (9) may be changed. Ｖ You may change it if it is permissible to vary it within a certain range of values.

[0093] As described above, if the search unit 45 has successfully solved the optimization problem by changing the initial solution or settings, that is, if a feasible solution has been obtained, the plan determination unit 44 generates feasible plan information and instructs the work machine 6 to execute it.

[0094] On the other hand, if a feasible solution cannot be obtained, or if the number of iterations reaches the upper limit, it is determined that the solution is not feasible. In this case, the work performed by the work machine 6 may be interrupted.

[0095] The proposal unit 46 of the plan modification unit 42 implements a function to generate proposal information and output it to the plan management device 1 when it is determined that the problem cannot be solved under predetermined conditions. The proposal unit 46 obtains information on the variables and set values ​​in the optimization problem, including the values ​​before change and the history of changes by the search unit 45, generates proposal information that includes at least the values ​​that are determined to need to be changed, and outputs it to the plan management device 1. The modified objective task represented by the proposal information is also called the modified objective task.

[0096] The information acquired by the proposal unit 46 includes variables or settings that were not permitted to be changed by the search unit 45. It is also possible to determine whether the change contributed to the solution based on the pre-change value and change history information from the search unit 45. The information acquired by the proposal unit 46 may also include the value of the final evaluation function J when the optimization problem was solved. From this evaluation function value, the contribution of the change to the solution can be quantitatively evaluated. Based on this acquired information, values ​​or settings that were not permitted to be changed but need to be changed are identified. Alternatively, as a result of determining the contribution to the solution, for example, values ​​that should be set to values ​​outside the range changed by the search unit 45, and their corresponding settings, may be identified. This information represents information that needs to be changed because, given the current state of the work object and work machine 6, the initial planning information output by the planning management device 1 determined that the target task could not be achieved. This factor could be, for example, the difference between the state of the work object or work machine 6 at the time the planning management device 1 acquired the initial planning information and the current state of the work object or work machine 6.

[0097] A specific example is given where the objective task is excavation of a predetermined volume using a backhoe. A value that was not permitted to be changed in the search unit 45 is, for example, the target total excavation volume V, expressed by equation (9). Ｔ There is a value related to the objective task, and changing it may prevent the objective task from being achieved. For this reason, the search unit 45 did not allow any changes. There are cases where it is necessary to change the unit time Tu, which was defined as the objective task, beyond the range that the search unit 45 allowed changes to. Another example is when it is necessary to change the initial value of the state vector of the work machine 6, that is, the current position of the work machine 6.

[0098] The proposed information output to the planning management device 1 includes at least one changeable value or set value. If there are multiple candidates that require modification, multiple values ​​may be included, or a selection may be made as appropriate depending on the target task and the type of work machine 6. Preferably, the planning management device 1 performs replanning based on the proposed information. However, automatic replanning may not be performed, or control may be interrupted, and the response of the planning management device 1 is not limited in this embodiment. The planning information generated based on the proposed information is also referred to as revised planning information.

[0099] Here, we will explain the distinction between "feasible" and "realizable" in this embodiment. Generally, the solution result of an optimization problem is output as either "feasible" or "realizable" depending on the solution means (solver, library, etc.). Here, "feasible" means that the given constraints are satisfied and the optimization variables obtained are such that the evaluation function value converges to a predetermined value. However, if the number of constraints is large or complex (for example, if there is nonlinearity), it may be feasible even if not all constraints are sufficiently satisfied. For example, the constraint expressed in equation (9) is an inequality that represents a condition on the excavation volume, but the value of the excavation volume does not directly affect the state variable X related to the control of the work machine 6. In other words, even if the condition in equation (9) is not satisfied, it is possible to execute control that satisfies other objective tasks, namely equation (8). On the other hand, in order to achieve the given objective task, it is necessary to satisfy equation (9). Therefore, in this embodiment, if the optimization result is feasible and satisfies the constraints for achieving the given objective task, it is described as realizable.

[0100] (Explanation of Operation) Next, the processes performed by the control system 100 will be explained with reference to Figure 5. Figure 5 is a flowchart showing an example of the operation of the control system 100.

[0101] The planning adjustment device 4 acquires initial planning information from the planning management device 1, information about the work object (target information) from the observation device 2, and status information from the work machine 6 (step S101). The status information of the work machine 6 may also be obtained via the control device 5.

[0102] Next, the planning adjustment device 4 sets an abstract model and a target logic formula based on the initial planning information and the model information and task information stored in the storage device 3 (step S102). In setting the abstract model, the model information, namely the work machine model information Ia1, the target model information Ia2, and the response model information Ia3, may be used. In setting the target logic formula, the task information, namely the constraint condition information Ib1, may be used, or the description method of the signal time-phase logic formula (STL) may be used.

[0103] Furthermore, the planning adjustment device 4 sets an initial solution for the optimization variables based on the planning information converted from the initial planning information by the planning information conversion unit 40, the target information, and the state information of the work machine 6 (step S103). The initial solution is a value over multiple time steps for the state variables of the abstract model.

[0104] Then, the feasibility determination unit 41 of the planning adjustment device 4 sets up an optimization problem with the abstract model and target logic formula as constraints, and solves it by providing an initial solution (step S104). The optimization variables include state variables for at least multiple time steps. The evaluation function includes control inputs for at least multiple time steps.

[0105] If a feasible solution is obtained as a result of the solution search (step S105; YES), the feasibility determination unit 41 generates feasible plan information from the feasible solution (step S106). The generation of feasible plan information may refer to task information, such as subtask information Ib2 and action information Ib3. The format of the feasible plan information may be the same as that of the initial plan information.

[0106] On the other hand, if a feasible solution cannot be obtained (step S105; NO), and the number of iterations is less than or equal to a predetermined value (step S107; YES), the plan modification unit 42 changes the initial solution that is allowed to be changed, or the setting values ​​of the optimization problem (step S108), and returns to the process of finding the solution (step S104). At this time, the initial solution may be updated based on the target information obtained from the observation device 2 and the state information obtained from the work machine 6. Here, "allowed to be changed" means "without changing the objective task," and the initial solution is the initial value of the optimization variables before optimization. For example, the initial value of the state information of the work machine 6.

[0107] If feasible plan information is obtained as described above, the control device 5 controls the work machine 6 (step S109), and when the planned work is completed, the operation is terminated.

[0108] If, ultimately, a feasible solution cannot be obtained (step S105; NO), and the number of iterations is not less than or equal to a predetermined value (step S107; NO), the plan modification unit 42 generates proposed information and outputs it to the plan management device 1 (step S110). The plan management device 1 may replan based on the proposed information. In this case, control of the work machine 6 may be interrupted.

[0109] (Effects of Embodiment 1) One of the factors that poses a challenge to the work machine 6 in achieving its target task is the difference in the on-site environment between the time the planning management device 1 generates the initial planning information and the time the work machine 6 actually performs the work. Since the time when the work machine 6 actually performs the work is later than the time when the planning management device 1 generates the initial planning information, this time difference can cause differences in the on-site environment. Differences in the on-site environment include, for example, the state of the work object and the work machine 6. Therefore, in a system that does not have a mechanism to determine whether the target task can be achieved when the planned operation is applied to the current work equipment and work object, it is possible that the target task cannot be achieved. The second challenge is that, since it is not possible to determine this before the start of operation, the only way is to actually start the work and check the results. In particular, for work objects where changes or responses to the actions of the work machine are irreversible, it is difficult to revert the changes once they have occurred, or the work efficiency is poor. Therefore, even if there is a discrepancy between the plan and the actual results, and it is anticipated that achieving the target task will be difficult, it will be difficult to redo the work, and as a result, achieving the target task will be difficult.

[0110] To address the above issues, the control system 100 of this embodiment can determine the feasibility of the target task based on the current state of the work machine 6 and the work object before starting operation. If the initial plan output by the planning management device 1 is determined to be unfeasible, the system can search for a feasible solution within the scope that does not affect the target task. On the other hand, if the target task is ultimately determined to be unfeasible, the system can request a replanning from the planning management device 1 without starting operation. Therefore, since the system only takes action on the work object when the target task is feasible, it has the effect of preventing discrepancies between the plan and the actual results, i.e., preventing the rework of the work. For this reason, the work machine can autonomously execute control to achieve the target task for work objects where changes and responses to actions are irreversible.

[0111] <Embodiment 2> (Description of Configuration) Figure 6 is a block diagram showing an example of the configuration of another embodiment of the control system. The configuration of the control system 200 shown in Figure 6 is such that the planning adjustment device 4 of the control system 100 in Embodiment 1 is replaced with a planning adjustment device 204. The planning adjustment device 204 is configured such that a performance evaluation unit 47 is added to the planning adjustment device 4 of Embodiment 1. The other configurations of the control system 200 are the same as those of the control system 100, so their description is omitted.

[0112] The performance evaluation unit 47 acquires performance information after the operation of the work machine 6 based on feasible plan information is completed, and implements a function to change the values ​​of field parameters included in the abstract model. The performance information consists of the state information of the work machine 6 after the completion of operation and the state information of the work target. This performance information is preferably acquired from the observation device 2 and the work machine 6, respectively, but it may also be acquired via other observation means, other work machines, control devices, etc., and is not limited to this embodiment. The type and specifications of the performance information to be acquired can be appropriately selected according to the target task and the type of work machine 6.

[0113] If the objective task is excavation of a predetermined volume using a backhoe, the field parameter is θ included in the response model of equation (3). ｆ Specifically, these include properties of the soil such as viscosity and diffusion coefficient, or parameters that characterize the excavation trajectory of the backhoe. The parameters set as field parameters and modified by the performance evaluation unit 47 can be appropriately selected according to the target task and the type of work machine 6.

[0114] The performance evaluation unit 47 illustrates how to change on-site parameters based on performance information. First, the performance evaluation unit 47 generates ideal state information for the work machine 6 after completion of operation and state information for the work target, based on feasible plan information. This process can be achieved by setting values ​​based on feasible plan information in an abstract model, that is, by predicting (simulating) the operation results based on feasible plan information using an abstract model. Next, the acquired performance information is compared with the calculated ideal state information. For example, the state of the soil that is the work target, i.e., the surface shape of the soil, is compared, and if there is a difference, it can be considered that the parameters representing the properties of the soil in the abstract model were inappropriate. Therefore, the parameters representing the properties of the soil are changed. The predicted value of the excavation trajectory based on the feasible plan information of the work machine 6 is compared with the actual value of the state information of the work machine 6, and if there is a difference, it can be considered that the parameters characterizing the excavation trajectory were inappropriate and can be changed.

[0115] The above-mentioned changes to field parameters can also be formulated as an estimation problem with the target parameter as the variable. For example, if the target parameter is defined as the variable and the difference between the predicted value from the abstract model and the acquired actual value is defined as the evaluation function, it can be formulated as an optimization problem to find the parameter that minimizes the value of the evaluation function. Alternatively, if the target parameter is assumed to be a random variable, for example, a Gaussian distribution defined by mean and variance, and the observation error of the actual value is also assumed to be a normal distribution, the parameter that maximizes the likelihood obtained from the predicted and actual values ​​can be estimated probabilistically. Therefore, the method for estimating field parameters can be appropriately selected depending on the target parameter and the actual value.

[0116] (Description of operation) Figure 7 is a diagram illustrating an example of the operation of the control system 200 according to Embodiment 2. The processing of the control system 200 (steps S101 to S110) is the same as the processing of the control system 100 according to Embodiment 1 described in Figure 5, so the explanation is omitted.

[0117] The performance evaluation unit 47 acquires status information of the work machine 6 and the work object after the work by the work machine 6 is completed (step S109), and updates the field parameters included in the abstract model (step S200). The status information of the work object may be acquired by the observation device 2, or via other observation means or devices. The updated field parameters are reflected in the subsequent solution of the optimization problem (step S104).

[0118] (Effects of Embodiment 2) The difference between the control system 200 and the control system 100 is that the plan adjustment device 204 is equipped with a performance evaluation unit 47. The effects of this will be explained. Another example of a factor that poses a challenge to the work machine 6 in achieving its target task is the dynamic difference between the plan and the actual result, which arises from the difference between the real environment and the abstract model. Since the abstract model is a mathematical representation of the real work machine 6 and the work object, it does not match reality, meaning that differences (errors) may occur in the state and behavior. Therefore, in existing systems and configurations without a performance evaluation unit 47, differences (errors) between the plan and the actual result may occur in each action taken by the work machine 6, making it impossible to achieve the target task. Even if the difference is recognized, there is no way to provide feedback and correct it, so as the work continues, errors accumulate, making it difficult to achieve the final target task. This difference can be seen as solving the dynamic differences that arise with each action taken by the work machine 6, whereas the control system 100 solves the fixed (static) differences in the on-site environment between the time the planning management device 1 generates the initial planning information and the time the work machine 6 actually performs the work.

[0119] In response to this challenge, in this embodiment, parameters included in the abstract model that are highly dependent on the field environment, or in other words, parameters that need to be adjusted (tuned) according to the field environment, are called field parameters, and the performance evaluation unit 47 is responsible for updating them. As a result, even if there is a difference between the plan and the actual results after the completion of a predetermined task based on feasible plan information, it is expected that the difference will decrease in subsequent tasks due to the updating of the field parameters. In other words, the effect of this embodiment is that it enables the realization of the final objective task without accumulating errors between the plan and the actual results through feedback based on the difference between the plan and the actual results. In addition to the effect of the control system 100 on static differences in the field environment, this embodiment can also respond to dynamic differences.

[0120] <Embodiment 3> (Description of Configuration) Figure 8 is a block diagram showing an example of the configuration of another embodiment of the control system. The configuration of the control system 300 shown in Figure 8 is such that the planning adjustment device 4 of the control system 100 in Embodiment 1 is replaced with a planning adjustment device 304. The planning adjustment device 304 includes a planning information conversion unit 340, a feasibility determination model 341, and a planning correction unit 342. The other configurations of the control system 300 are the same as those of the control system 100, so their description is omitted.

[0121] The functions implemented by the planning adjustment device 304 are equivalent to those of the planning adjustment device 4 of the control system 100. That is, the planning adjustment device 304 controls the work machine 6 via the control device 5 and outputs feasible planning information for executing the target task for the work object, based on the initial planning information output by the planning management device 1, the information about the work object acquired by the observation device 2, the state information about the work machine 6, and the information stored in the storage device 3. However, while the planning adjustment device 4 of the control system 100 solved an optimization problem, the planning adjustment device 304 of this embodiment uses inference based on a learning model, which is a difference in method.

[0122] The planning information conversion unit 340 of the planning adjustment device 304 implements the same functions as the planning information conversion unit 40 of the control system 100 in Embodiment 1. That is, the planning information conversion unit 340 acquires the initial planning information output by the planning management device 1 and outputs planning information that causes the work machine 6 to execute a series of actions. Based on this planning information, the planning information conversion unit 40 of the control system 100 set the initial solution to the optimization problem, but in the planning information conversion unit 340 of this embodiment, input variables to be input into the learning model are set.

[0123] The feasibility determination model 341 of the plan adjustment device 304 is a model that obtains input variables configured based on the initial plan information output by the plan management device 1, the target information acquired by the observation device 2, and the state information of the work machine 6, and outputs work performance values. The input variables include the plan information output by the plan information conversion unit 340, the target information such as soil shape data, and the state information of the work machine 6 such as backhoe state information data. The work performance values ​​are determined from quantities that define the target task. The work performance values ​​may be a single value or multiple values, i.e., multidimensional. The model that outputs work performance values ​​from the input variables (hereinafter referred to as the learning model) can be a model learned by any learning method, such as deep learning methods such as neural networks or machine learning. In the case of a neural network model, the dimensions of the input layer and the output layer are set according to the dimensions of the input variables and the dimensions of the work performance values, respectively.

[0124] Figure 9 shows an example of a functional block of the feasibility determination model 341. The feasibility determination model 341 has a learning unit 343 and an inference unit 344.

[0125] The learning unit 343 shown in Figure 9 implements the function of learning a model that outputs work performance values ​​from input variables. Therefore, the learning unit 343 operates for learning when the model has not been learned. The learning dataset consists of a set of multiple input variables and work performance values. Preferably, a dataset is prepared according to the target task, work object, and work device, and it is preferable to have a large number of data points. However, the number of data points may be increased from a specific dataset by methods such as data augmentation. Alternatively, data with intentional disturbances (noise) may be generated, and the generation of learning data and the learning method are not limited in this embodiment.

[0126] The inference unit 344 shown in Figure 9 takes input variables into the learning model learned by the learning unit 343 and implements the function of inferring feasible plan information based on the determination of the output work performance values. Therefore, the inference unit 344 operates when feasible plan information is to be output. The work performance values ​​output by the learning model are compared with information defined in the target task to determine whether the target task is feasible. For example, if the difference between the target value defined in the target task and the work performance value inferred by the learning model is less than or equal to a predetermined value, the target task is determined to be feasible. However, the determination is not limited to the above. If the target task is determined to be feasible, feasible plan information is generated based on the input variables input to the learning model.

[0127] A method for generating feasible plan information from input variables is described below. The input variables include information converted from the initial plan information output by the plan management device 1 by the plan information conversion unit 340. Therefore, by performing a certain reverse conversion on the input variables, information equivalent to the initial plan information can be obtained. Since the feasible plan information has the same format as the initial plan information, similar to Embodiment 1, the information obtained by this reverse conversion can be considered as feasible plan information. Similar to Embodiment 1, the subtask information Ib2 and action information Ib3 stored in the storage device 3 may be referenced to generate the feasible plan information.

[0128] Next, the plan modification unit 342 of the plan adjustment device 304 implements the same functionality as the plan modification unit 42 of the control system 100 in Embodiment 1. That is, the plan modification unit 342 operates when the feasibility determination model 341 determines that it is not feasible. The detailed operation flow of the plan modification unit 342 will be described later. Based on the output of the feasibility determination model 341 and the output of the plan information conversion unit 340, the plan modification unit 342 changes the values ​​of the input variables that are allowed to be changed. After that, the feasibility determination model 341 returns to the process of determining feasibility, i.e., the inference process by the learning model. At this time, the input variables changed by the plan modification unit 342 are applied. In Embodiment 1, the plan modification unit 42 changed the initial solution or the setting value of the optimization problem, but in this embodiment, the plan modification unit 342 changes the values ​​of the input variables. The selection of variables to be changed and the method of changing them are the same as in Embodiment 1, so the description is omitted.

[0129] The plan modification unit 342, similar to the control system 100 of Embodiment 1, generates proposed information and outputs it to the plan management device 1 when it is determined that the plan is not feasible under predetermined conditions. The generation of proposed information is also the same as in Embodiment 1, so the description is omitted.

[0130] An example is given where the target task is excavation of a predetermined volume using a backhoe. In this case, the learning model is trained to receive input variables generated based on the output of the planning information conversion unit 340, the target information (soil shape data), and the backhoe status information data, so that, for example, the total excavation amount per unit time is output as the work performance value. The total excavation amount of the work performance value is an example; the excavation amount for each excavation may also be output for each excavation, and this can be set appropriately in accordance with the design of the learning model. Preferably, the soil shape data is the soil shape of the environment in which the backhoe is actually working, but this is not limited to this. Soil shape data acquired by means other than the observation device 2 may also be used.

[0131] Once the model has been trained, in an environment where backhoe work is actually performed, the feasibility determination model 341 receives input variables generated based on the information output by the planning information conversion unit 340 based on the initial planning information output by the planning management device 1, the target information such as soil shape data acquired by the observation device 2, and the backhoe status information data, and outputs the total excavation amount, which is the actual work value. If this total excavation amount satisfies the objective task, the initial planning information output by the planning management device 1 is determined to be feasible. The determination that the objective task is satisfied can be made, for example, if the difference between the inferred total excavation amount and the excavation amount defined in the objective task is less than or equal to a specified value. Then, by generating feasible planning information from the input variables at this time, backhoe work can be performed.

[0132] If the inferred total excavation volume is determined not to satisfy the objective task, for example, if the difference between the inferred value and the value specified in the objective task is greater than or equal to a specified value, the plan modification unit 342 modifies the input variables. The selection of variables to be modified and the method of modification may be the same as in Embodiment 1. Alternatively, a probabilistic method specific to neural network models may be applied. For example, a Bayesian neural network that enables probabilistic inference by treating weight parameters as random variables may be used. This makes it possible to probabilistically estimate the actual work value, assuming uncertainty in the model and input data. In this case, feasibility can be determined, for example, by whether the expected value of the probability distribution satisfies the objective task.

[0133] Although this embodiment has been described as a configuration in which the planning adjustment device 4 of the control system 100 of Embodiment 1 is replaced, it may also be a configuration in which the planning adjustment device 204 of the control system 200 of Embodiment 2 is replaced. That is, the planning adjustment device 304 may include a configuration similar to that of the performance evaluation unit 47 of Embodiment 2. In that case, the field parameters updated by a configuration similar to that of the performance evaluation unit 47 are included in the input variables input to the learning model. Furthermore, it is also possible to configure the planning adjustment device 4 of the control system 100 of Embodiment 1 and the planning adjustment device 204 of the control system 200 of Embodiment 2 to further include a feasibility determination model 341 and a plan modification unit 342.

[0134] (Description of Operation) The processes performed by the control system 300 will be described with reference to Figure 10. Figure 10 is a flowchart showing an example of the operation of the control system 300 (steps S301 to S311). Operations, supplementary information, and conditions that are the same as those of the control system 100 according to Embodiment 1 described in Figure 5 will be omitted from this description.

[0135] If the feasibility determination model 341 has not been trained (step S301; NO), a model for inferring input variables and work performance values ​​is trained using a dataset of work performance values ​​for input variables that include at least initial planning information, target information about the work object, and state information about the work machine 6 (step S302). Preferably, the above dataset contains data about the environment in which the work machine 6 is actually performed, but is not limited to that, and may also include data obtained in other environments or by other means, or artificially generated data such as data augmentation. The model training method can be a method that trains the model so as to minimize the error between the inference result for the input variables and the given work performance value, a so-called supervised learning method, but the training method is not limited in this embodiment.

[0136] If the feasibility determination model 341 has been trained (step S301; YES), the planning adjustment device 4 acquires initial planning information from the planning management device 1, target information about the work object from the observation device 2, and status information from the work machine 6 (step S303).

[0137] Next, the planning adjustment device 4 sets input variables based on the information output by the planning information conversion unit 340, the target information, and the status information of the work machine 6 (step S304). The dimensions of the input variables are set to correspond to the learning model.

[0138] Then, the input variables are input to the feasibility determination model 341, and inference is performed by the learning model to obtain the work performance value as output (step S305).

[0139] If it is determined to be feasible based on the obtained work performance values ​​and the target task (step S306; YES), feasibility plan information is generated from the input variables (step S307).

[0140] On the other hand, if it is determined that it is not feasible (step S306; NO) and the number of iterations is less than or equal to a predetermined number (step S308; YES), the plan modification unit 342 changes the values ​​of the input variables that are allowed to be changed (step S309) and returns to the inference process (step S305). The values ​​to be changed may be probabilistic elements specific to the learning model.

[0141] If feasible plan information is obtained as described above (step S306; YES), feasible plan information is generated from the input variables (step S307). The control device 5 controls the work machine 6 using the generated feasible plan information (step S310), and terminates its operation when the planned work is completed.

[0142] If, ultimately, a feasible solution is not obtained (step S306; NO), and the number of iterations is not less than or equal to a predetermined value (step S308; NO), the plan modification unit 342 generates proposed information and outputs it to the plan management device 1 (step S311).

[0143] (Effects of Embodiment 3) Embodiment 3 is characterized in that the feasibility determination unit 41 of Embodiment 1 or Embodiment 2 is replaced with a feasibility determination model 341 that is based on a learning model. The feasibility determination unit 41 used model information and task information stored in the memory device 3 to set an abstract model and target logic formula in advance and construct an optimization problem. Therefore, it is necessary to set an appropriate abstract model, target logic formula, and optimization problem in advance depending on the target task, work object, and work machine. This has the following disadvantages. The first is that it relates to the operator who performs the setting. In addition to requiring manpower for setting, specialized knowledge is also required for setting, which can lead to dependence on the operator. The second is that the abstract model and target logic formula are expressed in mathematical formulas such as functions and inequalities. Compared to cases where the work machine and work object have linear dynamics and the work machine's response to the work object is linear and reversible, when the dynamics are nonlinear or the response is irreversible, there are limitations to describing them with functions and inequalities, which can lead to decreased accuracy and settings with low generality, i.e., settings that cannot be applied when the environment changes. The third issue, related to the second, is how to handle changes in the target task, work object, or work machine. If any or more of these elements change simultaneously, the established abstract model, target logic formula, and optimization problem may become invalid, requiring them to be reconfigured each time a change occurs. In Embodiment 1 and Embodiment 2, it was mentioned that a learning model can be applied to the abstract model, but the application here is strictly to the abstract model alone.

[0144] To address the above challenges, this embodiment is characterized by replacing the input information, which is characterized by the target task, work object, and work machine, with a learning model that directly outputs (infers) actual work performance values. In this case, the learning model is an integrated model in which the dynamics of the work machine and work object, represented by an abstract model, and the constraints, which are represented by target logical formulas, are combined. Therefore, the first point mentioned above is addressed by eliminating the burden and dependency on the worker through the prepared training data and defined learning method. The second point is addressed by eliminating limitations imposed by artificially designing or setting the model and constraints, or by expressing them with functions or inequalities, because an inductive model based on data is learned. The third point can be addressed by preparing training data for different target tasks, work objects, and work machines in advance. Alternatively, by training with diverse data from various environments, a highly generalizable model can be obtained, that is, a model trained in one environment can perform well in other environments.

[0145] Next, we will explain the effect on the computation time during feasibility determination. Compared to the computation time required to solve the optimization problem by the feasibility determination unit 41 of Embodiment 1 or Embodiment 2, the inference time by the feasibility determination model 341 of this embodiment is significantly shorter. This is because a typical method of solving an optimization problem uses an iterative algorithm starting from the initial solution, and considers the solution when the evaluation function value converges to a predetermined value or less as a feasible solution, requiring computation time during this process. On the other hand, the feasibility determination model 341, when performing only inference using a learning model such as a neural network, does not include iterative processing, resulting in shorter computation time. Therefore, this embodiment can shorten the time from when initial planning information is obtained from the planning management device 1 until the work machine 6 starts work, that is, it has the effect of improving work efficiency.

[0146] <Embodiment 4> (Description of Configuration) Figure 11 is a block diagram showing an example of the configuration of another embodiment of the control system. The configuration of the control system 400 shown in Figure 11 is such that the planning adjustment device 4 of the control system 100 of Embodiment 1 is replaced by a planning adjustment device 404 equipped with a plurality of planning adjustment units. The planning adjustment device 404 includes a planning information conversion unit 440 and m (m is an integer of 2 or more) m-th planning adjustment units 40m. The first planning adjustment unit 401 shown in Figure 11 is an example where m=1 of the m-th planning adjustment unit 40m. The other configurations are the same as those in the control system 100, so their description is omitted.

[0147] Figure 12 is a block diagram showing an example configuration of the first plan adjustment unit 401, where m=1 for the m-th plan adjustment unit 40m. Each m-th plan adjustment unit 40m is the same as the configuration example in Figure 12. The first plan adjustment unit 401 includes a feasibility determination unit 441 and a plan modification unit 442. The feasibility determination unit 441 may be replaced with the feasibility determination model 341 of Embodiment 3. The functions realized by the first plan adjustment unit 401 are equivalent to those of the plan adjustment device 4 of the control system 100. That is, the first plan adjustment unit 401 outputs feasible plan information for executing the target task for the work object by controlling the work machine 6 with the control device 5 based on the initial plan information output by the plan management device 1, the information about the work object acquired by the observation device 2, the status information about the work machine 6, and the information stored in the storage device 3. The same applies to each m-th plan adjustment unit 40m. However, as shown in Figure 12, the first plan adjustment unit 401 and each of the m-th plan adjustment units 40m do not need to be equipped with a plan information conversion unit 440. This is because each of the m-th plan adjustment units 40m can utilize the output of the plan information conversion unit 440 provided by the plan adjustment device 404.

[0148] The reason why the plan adjustment device 404 has multiple plan adjustment units will be explained. Embodiments 1 to 3 consisted of a single plan adjustment device, so the initial plan information input from the plan management device 1 was limited to a single plan. On the other hand, this embodiment can handle multiple plan information. Here, single / multiple refers to at least the following two types of perspectives. The first is the spatial or temporal (period) hierarchy of the plan information. The former, spatial hierarchy, refers to plans for a wide area (global) and plans for narrower (local) areas into which that area is divided, and two or more levels refer to plans for further divided areas. The latter, temporal hierarchy, refers to long-term plans and short-term plans into which that period is divided, and similarly, two or more levels refer to plans for further divided periods.

[0149] The second case involves planning information for multiple work machines 6. The configuration shown in Figure 11 is an example of the case where there is only one work machine 6; therefore, if there is planning information for multiple machines, there will be multiple work machines 6 corresponding to the number of planning information entries.

[0150] If the initial plan information output by the plan management device 1 is plan information that includes the above hierarchy, the plan information conversion unit 440 of the plan adjustment device 404 divides the plan information based on the hierarchy of the initial plan information. For example, the plan information conversion unit 440 may divide the space of the work target defined in the initial plan information into a predetermined number of equal parts from a spatial perspective, or divide it into predetermined areas. Alternatively, the plan information conversion unit 440 may divide the initial plan information from a temporal perspective, treating the entire initial plan information as a long-term plan and the finer unit plans that constitute the long-term plan as short-term plans. For example, if the entire work consists of processes A, B, and C, the plan information conversion unit 440 may divide the initial plan information showing the processes of the entire work into short-term plans for each of processes A, B, and C. These divisions may be performed automatically by the plan information conversion unit 440, or the plan information conversion unit 440 may divide it in units instructed by the user. Other functions of the plan information conversion unit 440 are the same as in embodiments 1 to 3.

[0151] The first plan adjustment unit 401 and the m-th plan adjustment unit 40m acquire the relevant information from the output of the divided plan information conversion unit 440. For example, in the case of spatial hierarchy, the first plan adjustment unit 401 acquires global plan information, and the second plan adjustment unit 402 (when m=2) acquires local plan information. In this way, each m-th plan adjustment unit 40m acquires plan information divided by space, period, or work machine, and is responsible for processing that plan information. The individual functions of each plan adjustment unit are the same as in Embodiments 1 to 3.

[0152] Here, we will explain the features of having multiple planning adjustment units. The first planning adjustment unit 401 and the mth planning adjustment unit 40m can each function independently. For example, we will explain a case in which the first planning adjustment unit 401 determines the feasibility of the global planning information and the second planning adjustment unit 402 determines the feasibility of the local planning information for the initial planning information output by the planning management device 1, which includes spatial hierarchy.

[0153] The first planning adjustment unit 401 determines the feasibility of the target task for the entire target area as global planning information. At this time, the global planning information, which has been divided by the planning information conversion unit 440 based on spatial hierarchy, is used to determine the feasibility of local planning information for each divided area to which a target has been assigned. For example, the global planning information and other input information provided to the first planning adjustment unit 401 are designated as first planning information, and the local planning information and other input information provided to the second planning adjustment unit 402 are designated as second planning information.

[0154] First, in the first plan adjustment unit 401, based on the first plan information, the feasibility determination unit 441 or the plan modification unit 442 determines that an area is feasible, and then the feasibility of that area is determined in the second plan adjustment unit 402 based on the second plan information. If feasibility is determined, control is executed based on the feasible plan information. On the other hand, for areas that the first plan adjustment unit 401 did not determine to be feasible, the plan modification unit 442 outputs proposed information and requests the plan management device 1 to replan. After replanning, the process returns to the determination by the first plan adjustment unit 401, and this is repeated thereafter. In this way, the process of determining the feasibility of a global plan and requesting replanning as necessary, and the process of controlling local units that have been determined to be feasible based on local feasible plan information, can be executed in stages or in parallel. More specific examples will be described later.

[0155] The above is merely an example, and the number of planning adjustment units (value of m), the types and number of planning information handled, and their hierarchical structure are not limited to those described above.

[0156] (Description of Operation) The processes performed by the control system 400 will be described with reference to Figure 13. Figure 13 is a flowchart showing an example of the operation of the control system 400 (steps S401 to S411). Operations, supplementary information, and conditions that are the same as those of the control system 100 according to Embodiment 1 described in Figure 5 will be omitted. In the following description, the cases where m = 1 and 2, that is, the case where there are two planning adjustment units, will be used as examples.

[0157] The planning adjustment device 404 acquires initial planning information from the planning management device 1, information about the work object (target information) from the observation device 2, and status information from the work machine 6 (step S401).

[0158] Next, each m-th plan adjustment unit 40m of the plan adjustment device 404 sets an abstract model and a target logical formula based on the initial plan information and the model information and task information stored in the storage device 3 (step S402).

[0159] The planning information conversion unit 440 generates divided planning information based on the initial planning information, target information, and state information (step S403). In each m-th planning adjustment unit 40m, the initial solution is set in the case of Embodiment 1 or Embodiment 2, and the input variables are set in the case of Embodiment 3.

[0160] The first plan adjustment unit 401 (m=1) determines feasibility based on the first plan information, which is divided plan information (step S404). In the case of Embodiment 1 or Embodiment 2, the optimization problem is solved, and in the case of Embodiment 3, the actual work value is inferred. If the feasibility determination unit 441 does not determine that it is feasible, the conditions etc. are changed within the range where changes are permitted, as in Embodiments 1 to 3 (for example, step S108), and the determination is made (for example, step S107). If the plan modification unit 442 also does not determine that it is feasible, it is determined that it is not feasible (step S405; NO). In other words, the feasibility determination (step S404) includes the determinations of the feasibility determination unit 441 and the plan modification unit 442.

[0161] If it is determined to be feasible (step S405; YES), the second plan adjustment unit 402 (m=2) determines feasibility based on the second plan information, which is the divided plan information (step S406). The determination is the same as that made by the first plan adjustment unit 401 in step S404 described above.

[0162] If it is determined to be feasible (step S407; YES), feasibility plan information is generated (step S408), and the work machine is controlled (step S409).

[0163] If all the divided plans are completed (step S410; YES), the operation ends; otherwise, the process returns to the determination of the second plan adjustment unit 402 (step S406).

[0164] On the other hand, if the first plan adjustment unit 401 does not determine that it is feasible (step S405; NO), the plan modification unit 442 generates proposed information and requests the plan management device 1 to replan (step S411). Based on the replanning result of the plan management device 1, the process returns to generating plan information (step S403) and the subsequent operations may be repeated. Also, if the second plan adjustment unit 402 does not determine that it is feasible (step S407; NO), the process returns to generating plan information by the first plan adjustment unit 401 (step S403). This is because, due to the hierarchical structure, the second plan information is influenced by the first plan information. In other words, the first plan information is modified so that the second plan adjustment unit 402 determines that it is feasible. At this time, the plan modification unit of the second plan adjustment unit 402 may output proposed information to the first plan adjustment unit 401, and the first plan adjustment unit 401 may modify the first plan information using that proposed information. Here, when generating the first plan information, if control of a portion of the divided area has been completed, the planned first plan information is considered to have been realized, or the actual controlled area is reflected. In other words, when modifying the first plan information, the areas where control has been completed are treated as known values ​​and are not changed.

[0165] (Effects of Embodiment 4) Embodiment 4, by having multiple planning adjustment units, has the following effects. First, it can handle large-scale planning information in space and time by dividing it. When the initial planning information output by the planning management device 1 is large in space and time, specifically when it is a plan for a wide target area or a plan that spans a long period of time, the processing by a single planning adjustment unit in Embodiments 1 to 3 may increase the computational load. This is because as the spatial and temporal scale increases, the number of dimensions of the abstract state increases, so the optimization solution in Embodiments 1 to 2 and the learning and inference by the learning model in Embodiment 3 also increase the computational load. This embodiment can solve the problem of increased computational load by dividing the planning information by focusing on the spatial and temporal hierarchy of the initial planning information or the number of work machines, and processing each in an independent planning adjustment unit. In other words, it is effective in improving work efficiency because it does not cause waiting for calculations.

[0166] The second advantage is that calculation waiting time can be reduced by processing in stages or in parallel across multiple planning adjustment units. In the spatially hierarchical planning described above, the feasibility of the global planning information is determined first, but it is not necessary to continue calculations until all divided areas are determined to be feasible. The feasibility of the local plans can be determined and control can be started sequentially for the areas that have been determined to be feasible. Furthermore, even if the global plan is determined to be unfeasible, the output of proposed information and replanning can be requested from the planning management device 1 in parallel with the execution of the local plan determination and control. Therefore, the effect is that calculation waiting time can be reduced and work efficiency can be improved.

[0167] <Application Examples> The following describes application examples based on Embodiments 1 to 4.

[0168] (Application Example 1) Application Example 1 shows an example of a backhoe 60, where the work machine 6 in Embodiments 1 to 3 is a construction machine. Figure 14 is a diagram showing an example of the configuration of the control system 500 according to the first application example. As shown in Figure 14, the first application example consists of a plan management device 1, an observation device 20, a storage device 3, a plan adjustment device 4, a control device 50, a backhoe 60, and a target area 61 representing the soil to be worked on. The plan management device 1, the storage device 3, and the plan adjustment device 4 are shown as independent devices, but their installation location and configuration method are not limited in each embodiment. For example, they may consist of a PC (personal computer) or server installed in a monitoring room or control room on site, connected to the backhoe 60 by communication capable of exchanging electrical signals. Alternatively, they may be implemented on a server on a network or in the cloud via these devices. The control system 500, observation device 20, control device 50, and backhoe 60 are examples of the control system 100, observation device 2, control device 5, and work machine 6 illustrated in the block diagrams of Figures 1, 6, and 8, respectively.

[0169] The observation device 20 is illustrated as an example of a device mounted on a backhoe 60, but it may be fixed to the environment or mounted on other work machinery. There may be multiple observation devices 20. Furthermore, the observation device 20 may be a sensor capable of acquiring three-dimensional information about the work object, preferably a 3D-LiDAR, and is not limited to each embodiment.

[0170] The control device 50 is illustrated as an example of a device mounted on the backhoe 60, but its specific form and mounting method are not limited to each embodiment. The control device 50 has at least a communication function, preferably a wireless communication function, that can exchange electrical signals with the planning adjustment device 4, and has the function of receiving planning information output by the planning adjustment device 4 and outputting control commands to control the control mechanism of the backhoe 60. For example, it may be a PC (personal computer), a notebook PC, or a tablet terminal or smartphone type terminal.

[0171] The backhoe 60 has a mechanism that is controlled (automatically / autonomously operated) based on control commands output by the control device 50. That is, the control device 50 can control the controlled parts (actuators) of the backhoe 60 via electrical control commands. The control mechanism and method are not limited to each embodiment, but for example, a device that physically drives the operating lever by an external control command (remote control device), a device that controls the hydraulic control mechanism that drives the controlled parts (actuators) of the backhoe 60 by a control command, or a device that controls an electrically driven control mechanism by a control command can be used. An example of one backhoe 60 is shown, but there may be multiple units.

[0172] The target area 61 is illustrated as a rectangular area representing the soil to be worked on, but its shape and number are not limited. The objective task of the backhoe 60 for the target area 61 could be, for example, excavating a predetermined volume of soil or a predetermined area from the target area and releasing it to a predetermined location. In this case, the objective task can be achieved through subtasks of movement, excavation, and release. The objective task and subtasks are not limited to the above example and are not restricted in each embodiment.

[0173] The functions of each component in Application Example 1 correspond to the functions of each component in any of Embodiments 1 to 3, therefore, their explanation is omitted here.

[0174] Figure 15 illustrates the operation flow of Application Example 1 when Embodiment 1 or 3 is applied. On the left side of Figure 15, the planning management device 1, planning adjustment device 4, control device 50, and backhoe 60 are shown in order from upstream to downstream, based on the flow of information in the system. On the right side of Figure 15, the time-series flow of information is schematically represented with the horizontal axis being time. Vertical arrows in Figure 15 indicate the flow of information, and horizontal arrows also indicate the flow of information. Here, the operation flows for the following three cases (Figure 10 (a), (b), and (c)) are illustrated. The basic operation flow corresponds to the operation flow in either Embodiment 1 or 3, so the description is limited to the parts specific to this application example.

[0175] Firstly, Figure 15(a) shows the information and time flow when the initial plan information output by the plan management device 1 is determined to be feasible by the feasibility determination unit 41 of the plan adjustment device 4. In other words, this is the case when, as a result of inputting the initial plan information as the initial solution, feasible plan information is obtained as a neighboring solution to the initial solution. In this case, the feasible plan information output by the feasibility determination unit 41 is output to the control device 50, and the control device 50 generates an operation command from the feasible plan information and controls the backhoe 60 based on that operation command.

[0176] Secondly, Figure 15(b) shows the information and time flow when the initial plan information output by the plan management device 1 is determined to be unfeasible by the feasibility determination unit 41 of the plan adjustment device 4, and optimization is repeated while modifying the initial solution etc. by the search unit 45 of the plan modification unit 42, and as a result it is finally determined to be feasible. In other words, even if the initial plan information is input as the initial solution, a feasible solution is not obtained, and a feasible solution is obtained after searching for a solution by the search unit 45. In the search for a solution, values ​​that are allowed to be changed are changed within a range that does not change the target task. For example, the initial position and attitude of the work machine 6, and parameters that adjust the excavation trajectory. After that, similarly, the feasibility determination unit 41 outputs the feasible plan information to the control device 50, and the backhoe 60 is controlled.

[0177] Thirdly, Figure 15(c) shows the information and time flow when the plan information output by the plan management device 1 is determined to be unfeasible by the feasibility determination unit 41 of the plan adjustment device 4, and even after further changes to the initial solution etc. by the search unit 45 of the plan modification unit 42, a feasible plan cannot be obtained in the end. In other words, it means that simply changing the values ​​that can be changed by the search unit 45 is not enough to make it feasible. In this case, the proposal unit 46 of the plan modification unit 42 outputs proposal information (not shown) to the plan management device 1, and the plan management device 1 performs replanning based on the proposal information. This proposal information includes values ​​that cannot be changed, i.e., variables that affect the target task. For example, target soil volume and area. Based on the replanning information obtained as a result of replanning, the plan adjustment device 4 determines feasibility again. In the example of Figure 15(c), if this replanning information is determined to be feasible, that is, the following example of the operation flow is the same as in Figure 15(a). However, the subsequent operation flow may be the case of Figures 15(b) or (c). While the planning management device 1 is replanning, control of the backhoe 60 by the control device 50 is temporarily suspended. During this time, control of the backhoe 60 based on other information, execution of pre-set operations, or human intervention may be performed, and this is not limited to each embodiment.

[0178] The characteristics of the three cases illustrated above, Figures 15(a), (b), and (c), are explained below. In the case of Figure 15(a), the processing of the plan adjustment device 4 is completed by the feasibility determination unit 41 alone, so control is started in the earliest time. On the other hand, in the case of Figure 15(b), the processing of the plan modification unit 42 is added to the processing of the feasibility determination unit 41 in the plan adjustment device 4, so the processing time is longer than in the case of Figure 15(a). As a result, the start of control is delayed compared to Figure 15(a). Furthermore, in the case of Figure 15(c), a replanning request is made to the plan management device 1, so the start of control is delayed even further than in the case of Figure 15(b). The time required for replanning in the plan management device 1 depends on the processing performance (specifications) of the plan management device 1, the target task, and the plan information to be output, but in each embodiment, the processing of the plan management device 1 is not considered.

[0179] Here, we will explain the factors that cause differences in cases like those shown in Figures 15(a), (b), and (c) above. These factors can be broadly divided into fixed factors resulting from differences in the field environment and initial conditions, and dynamic factors during control. First, we will describe the former, the fixed factors.

[0180] Since the plan adjustment device 4 acquires the initial plan information output by the plan management device 1, the time at which the feasibility determination unit 41 of the plan adjustment device 4 processes the plan is always later than the time at which the plan management device 1 generates the initial plan information. In other words, there is a difference between the time at which the plan management device 1 generates the initial plan information and the time at which the feasibility determination unit 41 determines feasibility. Due to this time difference, the site conditions, such as the state of the target area 61 or the state of the backhoe 60, may have changed. At this time, the plan adjustment device 4 acquires the environmental information acquired by the observation device 20 and the state information of the backhoe 60 when it starts processing, so it can take into account the current situation at the site. In other words, the plan management device 1 does not necessarily generate plan information that takes into account the current situation at the site. This difference in site conditions is one of the factors that causes the differences in cases like those shown in Figures 15(a), (b), and (c) above.

[0181] A concrete example of differences in site conditions is when the soil shape of the target area 61 differs. In this case, possible causes include the effects of other work machinery or people on the soil in the target area 61, or the effects of vibrations and wind from other work machinery at the site. Typically, these effects can become larger the greater the difference between the time planned by the planning management device 1 and the time determined by the planning adjustment device 4, that is, the more time has passed. Such differences can be addressed by acquiring information about the target area 61 using the observation device 20 when the planning adjustment device 4 starts processing.

[0182] Other specific examples of differences in on-site conditions include differences in the state of the backhoe 60, specifically its position (coordinates) and orientation. When the planning management device 1 generates initial planning information, it sets the position and orientation of the backhoe 60 based on its position and orientation at the end of the previous operation, or information acquired at the time of plan generation. However, the position and orientation of the backhoe 60 may change after the planning management device 1 has generated the plan. For example, an error may occur between the target position and the actual position during the previous movement operation (subtask). Such differences can be addressed by acquiring the position and orientation information of the backhoe 60 when the planning adjustment device 4 starts processing.

[0183] The following describes the problems that arise when this application example is not applied in cases where differences in site conditions occur as described above. For example, if the soil shape of the target area 61 is different, and the excavation operation (subtask) is executed based on the initial planning information generated by the planning management device 1, the action will be taken on a soil shape different from that planned, resulting in a difference between the excavated soil shape and the planned shape. As a result, the objective task of excavating a predetermined volume of soil or a predetermined area from the target area may not be achieved. For example, if the position and orientation of the backhoe 60 are different, even if there is no difference in soil shape, a difference will occur between the planned excavation position and the actual excavation position. In other words, a control error will occur. Therefore, as a result, a difference will occur between the excavated soil shape and the planned shape, and similarly, the objective task may not be achieved.

[0184] As mentioned above, if the shape of the excavated soil differs from the plan, one might consider redoing the work. However, since the soil being worked on exhibits an irreversible response to the action, it is difficult to return it to its pre-work state, or it would require extra work time. On the other hand, one might consider revising the subsequent plan based on the actual results. In this case, one could acquire the current state information of the target area 61 and replan using the plan management device 1. However, this requires time to acquire the information of the target area 61 and computation time to generate a new feasible plan that satisfies the objective task based on that information. Therefore, in either case, the challenge with methods that address the issue after the work has been performed is that they reduce work efficiency.

[0185] Next, we will explain the effects of applying this example. The cases in Figures 15(a), 15(b), and 15(c) show that the countermeasures change automatically depending on the magnitude of the difference in on-site conditions. In the case of Figure 15(a), feasible plan information was output solely through the processing of the feasibility determination unit 41. In other words, because the difference in on-site conditions was relatively small, or the range of the difference was limited, it was possible to obtain feasible plan information that did not deviate much from the initial plan information. On the other hand, in the case of Figure 15(b), a feasible plan could not be obtained with the initial solution, and a feasible solution was obtained by modifying the initial solution etc. by the plan modification unit 42. In other words, because the difference in on-site conditions was relatively larger than in the case of Figure 15(a), a feasible solution close to the initial plan information could not be obtained, and a feasible solution was obtained after searching for a solution by the plan modification unit 42. Furthermore, in the case of Figure 15(c), a feasible solution could not be obtained even after searching for a solution by the plan modification unit 42 due to a larger difference in on-site conditions. In this case, the calculation time for replanning can be shortened by having the suggestion unit 46 output the values ​​that should be changed among the variables affecting the target task as suggestion information to the planning management device 1.

[0186] Based on the above, this application example automatically changes the countermeasures to minimize changes from the initial plan information in response to differences in on-site conditions. In other words, the countermeasure is selected according to the tolerance of the differences in on-site conditions obtained as continuous values. Therefore, as shown on the horizontal axis of Figures 15(a), (b), and (c), it is possible to generate plan information that minimizes working time according to differences in on-site conditions. That is, the effect is that efficient work can be achieved.

[0187] Next, we will explain how to deal with dynamic factors during control, which are among the factors that cause differences in on-site conditions. Figure 16 shows the operation flow of Application Example 1 to which Embodiment 2 is applied, illustrating how to deal with dynamic factors during control. Similar to Figure 15, the left side shows each device based on the information flow of the system, and the right side shows time-series information. In Figure 16, the case from initial planning information to the first control (operation) is similar to that in Figure 15(a). After the operation is completed, the performance evaluation unit 47 of the feasibility determination unit 41 acquires performance information. Here, the performance information is, for example, the soil shape of the target area 61 acquired by the observation device 20, or the excavation volume and area. The excavation volume and area may be calculated based on the information acquired by the observation device 20, or they may be calculated using other measurement means, and the method of acquiring performance information is not limited to each embodiment.

[0188] The performance evaluation unit 47 modifies the field parameters in the feasibility determination unit 41 based on the acquired performance information. The field parameters are, for example, parameters inherent in the model of the soil being worked on, and may be included as model information in the target model information Ia2 or response model information Ia3. Specifically, these include, but are not limited to, viscosity parameters representing the properties of the soil being worked on, or parameters representing the diffusion of soil due to the excavation action. Since the parameter changes made by the performance evaluation unit 47 incorporate the actual actions of the backhoe 60, in other words, information about the results of the actions, the accuracy of the target model or response model in the feasibility determination unit 41 can be improved. That is, in subsequent determinations by the feasibility determination unit 41, the model with improved accuracy due to the parameter changes can be used. Figure 16 illustrates a case in which subsequent plans involve plan modifications, similar to Figure 15(b), but is not limited to this.

[0189] This section explains the relationship and effects of changing field parameters based on actual performance information and addressing dynamic factors during control. Even if there are no or acceptable differences in fixed field conditions, differences between the target model or response model and the actual work object will result in discrepancies between the plan and the actual action, i.e., the results. Examples include differences in soil shape after each excavation or differences in excavation volume each time. These dynamic differences accumulate during control, potentially creating a difference between the plan and the results after the final control (action) is completed. Since this difference is fed back after the control is completed, no effect is obtained for completed work objects, but the accuracy of subsequent action plans can be improved. Even if the properties of the work object are not uniform within the target area 61, i.e., they change spatially, the model for the next work area can be modified from the results of the previous work area, thus improving the accuracy of the action plan. Therefore, the effect of applying this example is that even when there are dynamic differences in field conditions, highly accurate work, i.e., work with a high degree of achievement of the target task, can be achieved.

[0190] (Application Example 2) Application Example 2, like Application Example 1, shows an example of a backhoe 60, where the work machine 6 in Embodiment 4 is a construction machine. An example of the configuration is the same as in Figure 14 shown in Application Example 1, so it is omitted. Figure 17 shows an example of operation when Embodiment 4 is applied.

[0191] Figure 17 shows an example of an excavation plan for a target area 61 using a backhoe 60, assuming the plan information includes spatial hierarchy. The upper part of Figure 17 shows the first plan adjustment unit 401 and an example of first plan information as global plan information whose feasibility is determined by the first plan adjustment unit 401. The middle part shows the second plan adjustment unit 402 and an example of second plan information as local plan information whose feasibility is determined by the second plan adjustment unit 402. The lower part shows the backhoe 60 controlled based on the feasible plan information obtained from the second plan adjustment unit.

[0192] First, the first planning adjustment unit 401 determines whether the total excavation amount for the entire target area 61 is feasible. At this time, each excavation plan (the dotted-dotted rectangular area in Figure 17) becomes a spatially divided unit area of ​​local plans, and the excavation volume of each unit area is represented as V[j] (j=1 to M: M is an integer of 2 or more). Then, the total excavation amount is calculated by summing the volumes V[j] of each unit area for j=1 to M. Therefore, the determination by the first planning adjustment unit 401 based on the first planning information is based on whether the total excavation amount can satisfy the target excavation amount and whether the volume V[j] can be satisfied in each unit area. This determination corresponds to step S404 in the flowchart of Embodiment 4 shown in Figure 13. For example, the upper part of Figure 17 shows an example in which the volume V[1] of the area j=1 is determined to be feasible, and the volume V[2] of the area j=2 is determined to be unfeasible. In this case, for the region where j=1, the process proceeds to the feasibility determination process based on the second plan information by the second plan adjustment unit 402.

[0193] The second planning adjustment unit 402, for each divided unit area, determines the excavation volume V of each unit area j for a more detailed excavation plan (solid rectangular area in Figure 17). ｊDetermine whether [i] (i = 1 to m: m is an integer greater than or equal to 2) is feasible. Excavation volume V of each unit region j ｊ [i] is, for example, the value obtained by dividing V[i] defined in the first planning information by the number of finer unit areas j. Therefore, the determination by the second planning adjustment unit 402 based on the second planning information is whether the volume V[i] defined in the first planning information is satisfied and whether each excavation plan has an excavation volume V ｊ This is based on whether [i] can be satisfied. This determination corresponds to step S406 in Figure 13. For example, the middle section of Figure 17 shows the case when it is determined to be feasible, and the lower section of Figure 17 shows how the backhoe 60 is controlled based on the second planning information. This corresponds to steps S408 and S409 in Figure 13.

[0194] On the other hand, the region j=2 in the upper part of Figure 17 is determined to be unfeasible, and the plan management device 1 is requested to replan. As a result of the replanning, first plan information with a different unit area is obtained for the excavation volume V'[2]. From here on, the first plan adjustment unit 401 similarly makes a judgment on this updated first plan information, and if it is determined to be feasible, the process proceeds to the second plan adjustment unit 402, which also makes a judgment, and if it is determined to be feasible, the backhoe 60 is controlled. In this way, the feasibility of all unit areas j is determined, and once the operation of the backhoe 60 is completed, the process ends.

[0195] The above describes an example of the operation of Application Example 2 with reference to Figure 17. However, the application example using multiple planning adjustment units in Embodiment 4 can be applied not only to cases with spatial hierarchy, but also, as mentioned above, to cases with temporal hierarchy or multiple work machines. Figure 17 shows an example of two planning adjustment units, but it is not limited to this.

[0196] Examples of cases to which application examples 1 or 2 described above can be applied include river construction, coastal construction, dam construction, forestry civil engineering, road construction, and tunnel construction. In particular, the effects of each embodiment can be expected in cases that involve excavating irregularly shaped or deformed objects such as soil, gravel, or natural ground using backhoes or similar equipment.

[0197] In addition to the application examples described above, this technology can also be applied to cases using work machines such as robot arms, as exemplified in Patent Document 1, such as sheet metal processing, and is particularly effective for tasks where changes or responses to actions by the work machine are irreversible.

[0198] Figure 18 is a diagram showing the configuration of a control system according to Embodiment 5. The control system 800 comprises a determination means 801 and a modification means 802. The determination means 801 determines, before operation starts, whether the objective task of the work can be achieved, based on plan information for the work performed on the work object by the work machine, target information for the work object, and information about the work machine. If the determination means 801 determines that the objective task cannot be achieved, the modification means 802 outputs proposed information representing a modified objective task in which the objective task has been changed. Based on the proposed information, the control system 800 creates modified plan information representing a plan for controlling the operation of the work machine, and controls the work machine according to the created modified plan information.

[0199] Figure 19 is a flowchart showing an example of the operation of the control system according to Embodiment 5. The computer determines, before starting the operation (step S801), whether the objective task of the work can be achieved, based on the plan information for the work performed on the work object by the work machine, the target information for the work object, and the information of the work machine. If it determines that the objective task cannot be achieved, it outputs proposed information representing a modified objective task in which the objective task has been changed (step S802). Based on the proposed information, it creates modified plan information representing a plan to control the operation of the work machine, and controls the work machine according to the created modified plan information (step S803).

[0200] In the embodiments described above, a portion of the control systems 100 and 800 may be implemented using a computer. In this case, the program for implementing this function may be recorded on a computer-readable recording medium, and the program recorded on this recording medium may be loaded into a computer system and executed. The term "computer system" as used herein refers to a computer system built into the control systems 100 and 800, and includes hardware such as an OS (Operating System) and peripheral devices.

[0201] "Computer-readable recording media" refers to portable media such as flexible disks, magneto-optical disks, ROMs, and CD-ROMs, as well as storage devices such as hard disks built into computer systems. Furthermore, "computer-readable recording media" may also include those that dynamically hold programs for a short period of time, such as communication lines used when transmitting programs over networks such as the Internet or communication lines such as telephone lines, and those that hold programs for a certain period of time, such as volatile memory inside computer systems that act as servers or clients in such cases. The above-mentioned programs may be for the purpose of realizing some of the functions described above, and may also be able to realize the above-mentioned functions in combination with programs already recorded in the computer system.

[0202] Some or all of the control systems 100 and 800 in the embodiments described above may be implemented as integrated circuits such as LSIs (Large Scale Integration). Each functional part of the control systems 100 and 800 may be individually implemented as a processor, or some or all of them may be integrated into a single processor. The method of implementing the integrated circuit is not limited to LSIs; it may also be implemented using dedicated circuits or general-purpose processors. If an integrated circuit technology that can replace LSIs emerges due to advances in semiconductor technology, an integrated circuit using that technology may be used.

[0203] As described above, several embodiments relating to this disclosure have been explained, but all of these embodiments are presented as examples and are not intended to limit the scope of the invention. These embodiments can be carried out in various other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their variations are included in the scope and spirit of the invention, as well as in the claims and their equivalents.

[0204] <Note> The control system, control method, and computer-readable recording medium described in the embodiment can be understood, for example, as follows.

[0205] (1) The control system according to the first embodiment includes: a determination means for determining whether the objective task of the work can be achieved before the start of operation, based on plan information for the work to be performed on a work object by a work machine, target information for the work object, and information about the work machine; and a modification means for outputting proposed information representing a modified objective task in which the objective task has been changed, if it is determined that the objective task cannot be achieved; and based on the proposed information, a modified plan information representing a plan for controlling the operation of the work machine is created, and the work machine is controlled according to the created modified plan information.

[0206] (1') The control system according to the first embodiment includes: determination means for determining whether it is possible to realize an objective task including continuous values ​​for a work object whose response to the action of the work machine is irreversible, based on given initial plan information, target information for the work object, and information for the work machine, before starting operation; modification means for outputting plan information obtained by modifying the initial plan information within the scope that does not change the objective task, or proposed information regarding a change to the objective task, if it is determined that it is not possible to realize the objective task; and control means for controlling the work machine based on feasible plan information, which is the plan information determined to be feasible, or the feasible plan information replanned based on the proposed information.

[0207] (2) The control system according to the second embodiment is the control system described in (1) and (1'), wherein the conditions for determining whether the implementation is possible include at least that the planning information is executable and that the constraints defined in the objective task are satisfied, and the means for making the determination determines whether the implementation is possible based on the difference between the information included in the planning information and the current information regarding the work machine and the work object, or the tolerance of the difference.

[0208] (3) A control system according to the third embodiment is a control system according to (1) to (2) and (1'), further comprising a storage means that stores, as model information, at least work machine model information of the work machine, target model information of the work object, and response model information of the response to the action, and as task information, at least constraint information indicating constraint conditions defined in the objective task, subtask information obtained by decomposing the objective task, and action information indicating the operation of the work machine on the work object, wherein the determination means sets an abstract model by abstracting the work machine and the work object using the model information, and determines whether the realization is possible using the task information.

[0209] (4) A control system according to the fourth embodiment is the control system described in (3), wherein the determination means converts a given initial planning information into abstracted planning information for the work machine and the work object based on the abstract model.

[0210] (5) The information processing system according to the fifth embodiment is a control system according to (3) to (4), further comprising performance evaluation means for acquiring performance information after the work by the work machine is completed, and changing the parameters included in the abstract model of the work machine and the work target based on the performance information.

[0211] (6) The information processing system according to the sixth aspect is the control system described in (3) to (5), wherein the determination means determines conditions for satisfying the objective task based on the information stored in the storage means and the abstract model, and uses signal temporal logic to describe the conditions.

[0212] (7) The information processing system according to the seventh aspect is the control system described in (1) to (6) and (1'), wherein the determination means includes a feasibility determination model that receives input variables set based on the plan information, the target information and the work machine information, and outputs work performance values, and the modification means includes a plan modification means that, when it is determined that the feasibility is not possible, proposes changing the input variables that do not affect the objective task or changing the variables that do affect the objective task.

[0213] (8) An information processing system according to the eighth aspect is a control system according to (1) to (7) and (1'), further comprising: a conversion means for converting the planning information into planning information divided by space, time, or work machine; and a second determination means for determining the feasibility of the objective task for each of the divided pieces of planning information.

[0214] (9) The control method according to the ninth aspect involves a computer determining, before starting operation, whether the objective task of the work can be achieved based on plan information for the work performed on the work object by the work machine, target information for the work object, and information about the work machine. If it is determined that the objective task cannot be achieved, the computer outputs proposed information representing a modified objective task in which the objective task has been changed. Based on the proposed information, the computer creates modified plan information representing a plan for controlling the operation of the work machine, and controls the work machine according to the modified plan information created.

[0215] (9') The control method according to the 9' embodiment involves a computer determining, before starting operation, whether it is possible to achieve a target task including continuous values ​​for a work object whose response to the action of the work machine is irreversible, based on given initial plan information, target information for the work object, and information for the work machine. If it is determined that it is not possible to achieve the target task, the computer outputs plan information obtained by modifying the initial plan information within the scope that does not change the target task, or proposed information regarding the modification of the target task. The computer then controls the work machine based on feasible plan information, which is either the plan information determined to be feasible, or the feasible plan information replanned based on the proposed information.

[0216] (10) The computer-readable storage medium according to the tenth embodiment is a computer-readable recording medium that records a program which causes a computer to record a program which determines, before the start of operation, whether or not the objective task of the work can be achieved based on plan information for the work to be performed on a work object by a work machine, target information for the work object, and information on the work machine, outputs proposed information representing a modified objective task in which the objective task has been changed if it is determined that the objective task cannot be achieved, creates modified plan information representing a plan to control the operation of the work machine based on the proposed information, and executes a process to control the work machine in accordance with the created modified plan information.

[0217] (10') The computer-readable storage medium according to the 10' embodiment is a computer-readable recording medium that records a program which causes a computer to execute a process that causes the

[0218] According to the control system, control method, and computer-readable recording medium described above, the target task can be achieved even when the changes or responses to the actions of the work machine are irreversible.

[0219] 1...Planning Management Device 2, 20...Observation Device 3...Storage Device 3a...Model Information 3b...Task Information 4...Planning Adjustment Device 5, 50...Control Device 6...Work Machine 40, 340, 440...Planning Information Conversion Unit 41...Feasibility Determination Unit 42, 342...Planning Modification Unit 43...Target Logical Formula Generation Unit 44...Planning Determination Unit 45...Search Unit 46...Proposal Unit 47...Performance Evaluation Unit 60...Backhoe 61...Target Domain Ia1...Work Machine Model Information Ia2...Target Model Information Ia3...Response Model Information Ib1...Constraint Condition Information Ib2...Subtask Information Ib3...Action Information 100, 200, 300, 400, 500...Control System 204, 304, 404...Planning Adjustment Device 341...Feasibility Determination Model 343...Learning Unit 344...Inference Unit 401...First planning adjustment unit 40m...mth planning adjustment unit 800...Control system 801...Determination means 802...Correction means

Claims

1. A control system comprising: a determination means for determining whether the objective task of the work can be achieved before the start of operation, based on plan information for the work to be performed on a work object by a work machine, target information for the work object, and information about the work machine; and a modification means for outputting proposed information representing a modified objective task in which the objective task has been changed, if it is determined that the objective task cannot be achieved; and a modified plan information representing a plan for controlling the operation of the work machine based on the proposed information, and a control system for controlling the work machine according to the modified plan information created.

2. The control system according to claim 1, wherein the conditions for determining whether the implementation is possible include at least that the planning information is executable and that the constraints defined in the objective task are satisfied, and the means for determining whether the implementation is possible is determined based on the difference or tolerance of the difference between the information contained in the planning information and the current information regarding the work machine and the work object.

3. A control system according to claim 1, further comprising: a storage means for storing, as model information, at least a work machine model information of the work machine, an object model information of the work target, and a response model information of the response to the action; and as task information, at least constraint information indicating constraint conditions defined in the objective task, subtask information obtained by decomposing the objective task, and action information indicating an action performed by the work machine on the work target, wherein the determination means uses the model information to abstract the work machine and the work target to set an abstract model, and uses the task information to determine whether the realization is possible.

4. The control system according to claim 3, wherein the determination means converts a given initial planning information into abstracted planning information for the work machine and the work object based on the abstract model.

5. The control system according to claim 3, further comprising: performance evaluation means for acquiring performance information after the work by the work machine is completed, and modifying the parameters included in the abstract model of the work machine and the work target based on the performance information.

6. The control system according to claim 3, wherein the determination means determines conditions for satisfying the objective task based on the information stored in the storage means and the abstract model, and uses signal temporal logic to describe the conditions.

7. The control system according to claim 1, wherein the determination means includes a feasibility determination model that receives input variables set based on the plan information, the target information, and the work machine information, and outputs work performance values, and the modification means includes a plan modification means that, when it is determined that the feasibility is not possible, proposes changing input variables that do not affect the objective task, or changing variables that do affect the objective task.

8. The control system according to claim 1, further comprising: a conversion means for converting the planning information into planning information divided by space, time, or work machine; and a second determination means for determining the feasibility of the target task for each of the divided pieces of planning information.

9. A control method comprising: a computer determining, before starting operation, whether the objective task of the work can be achieved based on planning information for the work performed on the work object by the work machine, target information for the work object, and information about the work machine; outputting proposed information representing a modified objective task in which the objective task has been changed if it is determined that the objective task cannot be achieved; creating modified plan information representing a plan for controlling the operation of the work machine based on the proposed information; and controlling the work machine according to the created modified plan information.

10. A computer-readable recording medium that records a program that causes a computer to perform the following: determine whether or not the objective task of the work can be achieved before the start of operation, based on planning information for the work to be performed on a work object by a work machine, target information for the work object, and information about the work machine; output proposed information representing a modified objective task in which the objective task has been changed if it is determined that the objective task cannot be achieved; create modified plan information representing a plan to control the operation of the work machine based on the proposed information; and execute a process to control the work machine according to the created modified plan information.

Images