Analysis device, analysis method, and program
The analysis device simulates robot behavior on three-dimensional point cloud models of workspaces to optimize robot and ground parameters, addressing the challenge of uneven terrain, thus improving robot design for varied ground patterns.
Patent Information
- Application Number
- JP2024001311
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2024-01-09
- Publication Date
- 2025-07-22
AI Technical Summary
Existing simulation technologies for robots fail to accurately account for the unevenness and hardness of ground surfaces, limiting their ability to design robots that can effectively operate in various ground patterns, including fields, gravel roads, asphalt roads, and unpaved grasslands.
An analysis device that acquires an aerial image of the workspace, generates a three-dimensional point cloud model of the workspace, simulates the behavior of a robot model on this model, and optimizes robot and ground surface parameters based on virtual behavior data to create a simulation environment that mimics real-world conditions.
Enables the design of robots that can better adapt to diverse ground conditions, enhancing their operational suitability and robustness across different terrains.
Smart Images
Figure 2025107830000001_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to an analysis device, an analysis method, and a program.
Background Art
[0002] In the face of a shortage of labor in agriculture, there is a demand to replace tasks that must be handled manually with robots. In particular, small robots have the advantage of being able to move even in fields in the late growth stage where it is difficult to work with tractors, and can also be used for tasks that require forces such as transportation and soil measurement, which are difficult for drones.
[0003] When making a robot work in a field or the like, the robot is likely to be affected by the ground (for example, the unevenness and hardness of the ground, etc.). In particular, the smaller the robot, the more significantly it is likely to be affected by the ground. Therefore, it is necessary to design the robot to cope with various ground patterns. To design the robot, in the real world, it may be considered to repeat the design, development, and verification experiments in the field many times, but since the development cost is too high, it is preferable to optimize the design of the robot in the virtual world using simulation.
[0004] As simulation technologies related to robots working in fields or the fields themselves, for example, the technologies described in Non-Patent Documents 1-2 are known.
Prior Art Documents
Non-Patent Documents
[0005]
Non-Patent Document 1
Non-Patent Document 2
Summary of the Invention
Problems to be Solved by the Invention
[0006] Non-Patent Document 1 discloses a simulation system for a field robot. In this simulation system, a lightweight mesh model is adopted as a model of the field. Non-Patent Document 2 discloses generating three-dimensional point clouds of a field and crops from a plurality of still images based on the principle of triangulation.
[0007] However, in these conventional technologies, various parameters such as the unevenness and hardness of the field ground cannot be appropriately determined, and the robot cannot be sufficiently designed to cope with various ground patterns of the field. Moreover, such problems are not limited to the field only, but are also common to roads paved with gravel or asphalt, and unpaved grasslands (ground overgrown with weeds) around the field.
[0008] The present invention has been made in consideration of such circumstances, and one of the objectives is to provide an analysis device, an analysis method, and a program that can more suitably design a robot by creating a simulation environment of the robot considering the influence of the ground.
Means for Solving the Problems
[0009] One aspect of the present invention is an analysis device including an acquisition unit that acquires an image of a robot's workplace imaged from above, a generation unit that generates a workplace model, which is a virtual model of the workplace and is represented by a three-dimensional point cloud, from the image, a simulation unit that calculates, by simulation, the behavior of a robot model, which is a virtual model of the robot, when the robot model is operated on the workplace model, and an optimization processing unit that optimizes a robot parameter, which is a parameter related to the design and control of the robot model, and a ground surface parameter, which is a parameter related to the ground surface of the workplace model, based on virtual behavior data, which is data indicating the behavior of the robot model.
[0010] Another aspect of the present invention is an analysis method using a computer. The analysis method includes acquiring an image of a robot's workplace imaged from above, generating a workplace model, which is a virtual model of the workplace and is represented by a three-dimensional point cloud, from the image, calculating, by simulation, the behavior of a robot model, which is a virtual model of the robot, when the robot model is operated on the workplace model, and optimizing a robot parameter, which is a parameter related to the design and control of the robot model, and a ground surface parameter, which is a parameter related to the ground surface of the workplace model, based on virtual behavior data, which is data indicating the behavior of the robot model.
[0011] Another aspect of the present invention is a program for causing a computer to execute. The program includes acquiring an image of a robot's workplace captured from above, generating a workplace model, which is a virtual model of the workplace and is represented by a three-dimensional point cloud, from the image, calculating, by simulation, the behavior of a robot model, which is a virtual model of the robot, when the robot model is operated on the workplace model, and optimizing robot parameters, which are parameters related to the design and control of the robot model, and surface parameters, which are parameters related to the ground surface of the workplace model, based on virtual behavior data, which is data indicating the behavior of the robot model.
Advantages of the Invention
[0012] According to the above aspect, by creating a simulation environment for a robot considering the influence of the ground, the robot can be designed more suitably.
Brief Description of the Drawings
[0013]
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Figure 7
Figure 8
Figure 9
Figure 10
Figure 11
Figure 12
Figure 13
Embodiments for Carrying Out the Invention
[0014] Hereinafter, with reference to the drawings, embodiments of the analysis device, analysis method, and program of the present invention will be described.
[0015] [Overview of the Analysis Device] The analysis device 100 according to the embodiment acquires an image of a robot's workplace taken from above (hereinafter referred to as an aerial image IMG).
[0016] For example, the workplace may be a farm field provided with ridges for growing crops or the like. In this case, the robot may be, for example, a work robot that sprays fertilizers and water on crops, transports crops, or harvests crops in the farm field. Further, the workplace is not limited to the farm field, and may be a road paved with gravel or asphalt, or an unpaved grassland (ground overgrown with weeds) around the farm field. In this case, the robot may be, for example, a work robot that performs various operations (transportation, cleaning, traffic control, etc.) on the road or grassland. Hereinafter, as an example, it is assumed that the workplace is a "farm field" and the robot is a robot that performs various operations in the "farm field" for explanation.
[0017] The analysis device 100 generates a virtual model of the farm field (hereinafter referred to as the farm field model MDL1) from the aerial image IMG. The analysis device 100 operates a virtual model of the robot (hereinafter referred to as the robot model MDL2) on the generated farm field model MDL1 by simulation, and calculates (simulates) the behavior of the robot model MDL2. The farm field model MDL1 is an example of a "workplace model".
[0018] Then, based on the data indicating the behavior of the robot model MDL2 obtained by simulation (hereinafter referred to as virtual behavior data), the analysis device 100 optimizes various parameter sets set during the simulation.
[0019] The parameter sets to be optimized include, for example, robot parameters and ground surface parameters.
[0020] Robot parameters are parameters related to the design and control of the robot model MDL2. For example, robot parameters may include the relative positions and postures between the joints of the robot, the rotational axis vectors of the joints, the masses of the links, the elastic forces of the links, the moments of inertia, the center of gravity positions, the maximum torques of the actuators of the robot, the lengths of the legs of the robot, and the like.
[0021] Ground surface parameters are parameters related to the ground surface of the field model. For example, ground surface parameters may include the coefficient of friction, the coefficient of kinetic friction, the hardness of the ground surface in the vertical and translational directions, the viscosity of the ground surface in the vertical and translational directions, and the like.
[0022] By optimizing the robot parameters, ground surface parameters, etc. in this way, a simulation environment closer to the real world can be constructed, and by using this simulation environment, the robot can be designed more suitably.
[0023] [Configuration of the Analysis Device] FIG. 1 is a diagram showing an example of the configuration of the analysis device 100 according to the embodiment. As shown in the figure, the analysis device 100 includes, for example, a communication interface 110, an input interface 120, an output interface 130, a storage unit 140, and a processing unit 150.
[0024] The communication interface 110 includes, for example, a NIC (Network Interface Card), a wireless communication module including a receiver and a transmitter, etc. The communication interface 110 communicates with external devices via networks such as a LAN (Local Area Network) or a WAN (Wide Area Network). The external devices are, for example, a flying object (such as a drone, an unmanned aircraft, etc.) that captures the aerial image IMG described later, a computer (such as a smartphone or a tablet terminal) available to the user, etc.
[0025] The input interface 120 receives various input operations from the user, converts the received input operations into electrical signals, and outputs them to the processing unit 150. For example, the input interface 120 includes a mouse, a keyboard, a trackball, a switch, a button, a joystick, a touch panel, etc. The input interface 120 may be a voice user interface that receives voice input including, for example, a microphone, etc.
[0026] The output interface 130 includes, for example, a display and a speaker, etc. The display displays an image generated by the processing unit 150, a GUI (Graphical User Interface) for receiving various input operations from the user, etc. For example, the display is an LCD (Liquid Crystal Display), an organic EL (Electro Luminescence) display, etc. The speaker outputs the information input from the processing unit 150 as sound.
[0027] The storage unit 140 is realized by, for example, an HDD (Hard Disc Drive), a flash memory, an EEPROM (Electrically Erasable Programmable Read Only Memory), a ROM (Read Only Memory), a RAM (Random Access Memory), etc. The storage unit 140 stores various programs such as firmware and application programs.
[0028] The processing unit 150 includes, for example, an acquisition unit 151, a generation unit 152, a simulation unit 153, an optimization processing unit 154, and an output control unit 155.
[0029] The components of the processing unit 150 are realized, for example, by a processor such as a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit) executing a program stored in the storage unit 140. Also, some or all of the components of the processing unit 150 may be realized by hardware such as LSI (Large Scale Integration), ASIC (Application Specific Integrated Circuit), FPGA (Field-Programmable Gate Array), or SOC (System On Chip), or may be realized by the cooperation of software and hardware.
[0030] [Processing Flow of the Analysis Device] Hereinafter, a series of processing flows of the analysis device 100 will be described with reference to the flowchart. FIG. 2 is a flowchart showing an example of a series of processing flows of the analysis device 100 according to the embodiment. The processing of this flowchart may be repeated, for example, at a predetermined cycle.
[0031] First, the acquisition unit 151 acquires the aerial image IMG of the farmland (step S100).
[0032] FIG. 3 is a diagram showing an example of the aerial image IMG of the farmland. As shown in the figure, when an aircraft (for example, a drone, etc.) takes an aerial photo of the farmland using a three-dimensional laser scanner or the like, the acquisition unit 151 may communicate with the aircraft via the communication interface 110 and acquire the aerial image IMG of the farmland from the aircraft.
[0033] In addition, when the aerial image IMG of the farmland is stored in an external server or the like, the acquisition unit 151 may access the external server via the communication interface 110 and acquire the aerial image IMG of the farmland from the external server. Further, when the user inputs the aerial image IMG of the farmland to the analysis device 100 using the input interface 120, the acquisition unit 151 may acquire the aerial image IMG of the farmland from the input interface 120.
[0034] Next, the generation unit 152 generates a farmland model MDL1, which is a virtual model of the farmland, from the aerial image IMG of the farmland (step S102).
[0035] FIG. 4 is a diagram showing an example of the farmland model MDL1. The farmland model MDL1 in the present embodiment is represented by a three-dimensional point cloud model that approximates the farmland as a set of innumerable points. The farmland model MDL1 may include not only three-dimensional information such as width, depth, and height, but also color information such as RGB. Further, the farmland model MDL1 includes surface parameters. The surface parameters are parameters related to the ground surface of the farmland as described above, and include, for example, friction, viscosity, hardness, and the like. The height component of the three-dimensional information may be treated as one of the surface parameters.
[0036] Next, the simulation unit 153 simulates the behavior of the robot model MDL2, which is a virtual model of the robot, on the farmland model MDL1 (step S104). The robot model MDL2 in the present embodiment is represented by a mesh model (polygon model) that approximates the robot with points and planes. This mesh model may be generated by, for example, three-dimensional CAD (Computer Aided Design) software.
[0037] FIG. 5 is a diagram showing an example of the simulation. When the robot is a traveling robot that travels between the ridges of the farmland, as shown in the figure, the simulation unit 153 simulates the robot model MDL2 to travel between the ridges of the farmland model MDL1. Details of the simulation will be described later with reference to separate figures.
[0038] Next, the optimization processing unit 154 optimizes various parameter sets set at the time of simulation based on the data indicating the behavior of the robot model MDL2 simulated by the field model MDL1, that is, the virtual behavior data (step S106).
[0039] Next, the output control unit 155 outputs the optimized parameter set (step S108). For example, the output control unit 155 may display the optimized parameter set on the display of the output interface 130. Further, the output control unit 155 may transmit the optimized parameter set to an external computer available to the user via the communication interface 110. Thereby, a series of processes of the flowchart is completed.
[0040] [Overview of Simulation] FIG. 6 is a diagram for explaining the overview of the simulation. As shown in the illustrated overview, the behavior of the robot model MDL2 is simulated by the field model MDL1 (F. simulator in the figure). The simulation of this behavior is, for example, the calculation of the state S and the wheel torque τ of the robot model MDL2.
[0041] First, the simulation unit 153 calculates the wheel torque τ as the behavior of the robot model MDL2 from the initial state S of the robot model MDL2 using a control system (C. control system in the figure) according to Equation (1). As the control system, for example, a state machine or a neural network may be adopted. The state S includes, for example, the position and orientation of the robot model MDL2.
[0042] [Equation]
[0043] Next, on the field model MDL1 temporarily fixed to arbitrary surface parameters (hereinafter referred to as surface parameters A for convenience), the simulation unit 153 calculates a state S (that is, the state next to the initial state) that is estimated to be possessed by the robot model MDL2 when the robot model MDL2 is caused to travel with the wheel torque τ calculated by the control system. The calculation for converting from this wheel torque τ to the state S may include dynamic calculation, integral calculation, kinematic calculation, collision calculation, and contact force calculation.
[0044] FIG. 7 is a diagram for explaining a method of converting from the wheel torque τ to the state S. As shown in the figure, when the wheel torque τ is output from a control system such as a state machine or a neural network, the simulation unit 153 calculates the angular acceleration q^(··) of the wheels of the robot model MDL2 from the wheel torque τ by dynamic calculation. The hat symbol (^) with the dot symbol (·) represents the first derivative, and the hat symbol (^) with two dot symbols (··) represents the second derivative.
[0045] For example, the simulation unit 153 may calculate the angular acceleration q^(··) of the wheels from the wheel torque τ according to Equation (2).
[0046]
Equation
[0047] Next, the simulation unit 153 calculates the angular velocity q^(·) and the angle q of the wheels from the angular acceleration q^(··) of the wheels by integral calculation.
[0048] For example, the simulation unit 153 may calculate the angular velocity q^(·) of the wheels by integrating the angular acceleration q^(··) of the wheels at a unit period ΔT on the order of milliseconds according to Equation (3), and further calculate the angle q by integrating the angular velocity q^(·) at the unit period ΔT.
[0049]
Equation
[0050] Next, the simulation unit 153 calculates the position x, velocity x^(·), and acceleration x^(··) of the wheels of the robot model MDL2 from the angle q, angular velocity q^(·), and angular acceleration q^(··) of the wheels by kinematic calculation.
[0051] For example, the simulation unit 153 may calculate the position x, velocity x^(·), and acceleration x^(··) of the wheels from the angle q, angular velocity q^(·), and angular acceleration q^(··) of the wheels according to Equation (4).
[0052]
Number
[0053] Next, the simulation unit 153 calculates the collision amounts y and y^(·) from the position x, velocity x^(·), and acceleration x^(··) of the wheels by collision calculation.
[0054] Figures 8 to 10 are diagrams for explaining the details of the collision calculation. As shown in Figure 8, the simulation unit 153 calculates the collision amounts y and y^(·) based on the point cloud of the robot model MDL2 and the vertices of the mesh of the field model MDL1.
[0055] Specifically, the simulation unit 153 calculates the collision amount y of each point to the mesh based on the distance and velocity when each point of the field model MDL1 represented by the three-dimensional point cloud penetrates the mesh of the robot model MDL2. Then, as shown in Figure 9, the simulation unit 153 calculates the contact force f described later by integrating the collision amount y over the point cloud.
[0056] At this time, as shown in FIG. 10, the simulation unit 153 stores in the storage unit 140 a grid that divides the ground surface and an array having a size equal to the number of grids, and further stores in advance the positions of all point groups within each grid in the array. As a result, the point group of the field model MDL1 that collides with the mesh of the robot model MDL2 can be calculated at high speed from the position of the robot model MDL2.
[0057] Return to the description of FIG. 7. Next, the simulation unit 153 calculates the vertical component of the contact force f^(v) and the horizontal component of the contact force f^(h) from the collision amounts y and y^(·) by contact force calculation.
[0058] For example, the simulation unit 153 may calculate the vertical component of the contact force f^(v) and the horizontal component of the contact force f^(h) according to Equation (5).
[0059]
Equation
[0060] z represents the difference (displacement) between the collision position at a certain time t and the collision position at the time (t - ΔT) that is one unit cycle ΔT before the time t, and z^(·) represents the first derivative thereof. k represents the hardness of the ground surface parameter, d represents the viscosity of the ground surface parameter, and μ represents the friction of the ground surface parameter.
[0061] In this way, through a series of calculation processes, the contact force f of the wheel with respect to the ground surface is calculated from the wheel torque τ.
[0062] Return to the description of the outline of the simulation in FIG. 6. When the simulation unit 153 calculates the contact force f of the wheel from the wheel torque τ, it calculates the state S of the robot model MDL2 based on the contact force f. Then, the simulation unit 153 recalculates the wheel torque τ as the behavior of the robot model MDL2 using the control system (C. control system in the figure).
[0063] In this way, the simulation unit 153 repeats the process of changing the robot parameters at intervals of the period ΔT, and running the robot model MDL2 while changing the state S on the field model MDL1 fixed to a certain arbitrary ground parameter A. As a result, virtual behavior data showing consecutive states S in the time series of the robot model MDL2 having a certain robot parameter is generated.
[0064] Furthermore, the simulation unit 153 repeats the process of changing the robot parameters and running the robot model MDL2 on the field model MDL1 fixed to the same ground parameter A as the previous time. As a result, a plurality of virtual behavior data with different robot parameters from each other are generated under the same condition of the ground parameter A.
[0065] Next, the optimization processing unit 154 searches for the optimal robot parameters for the ground parameter A from a plurality of virtual behavior data with different robot parameters from each other using the objective function (Q. objective function in the figure). The objective function may be, for example, a function for minimizing the wheel torque τ or the running error, or a function for maximizing the running speed (in other words, a function for minimizing the working time). As a result, the optimal robot parameters for the ground parameter A are selected.
[0066] Receiving this, the output control unit 155 outputs the optimal robot parameters for the ground parameter A.
[0067] When the optimal robot parameters are output from the analysis device 100, the robot designer implements the robot based on the robot parameters. Then, the designer runs the implemented robot in an actual field (the field that is the target of the aerial image IMG that was the basis for the field model MDL1). That is, a verification experiment is conducted in which an actual robot is used to run in an actual field. As a result, data showing the state S of the robot actually measured by the sensors mounted on the robot (hereinafter, actual behavior data) is obtained when the robot runs in a field with unknown ground parameters.
[0068] Next, the acquisition unit 151 acquires actual behavior data. When the actual behavior data is acquired, the simulation unit 153 causes the robot model MDL2 temporarily fixed to arbitrary robot parameters to travel on each of a plurality of field models MDL1 having different ground surface parameters from each other. Thereby, virtual behavior data indicating the behavior of the robot model MDL2 simulated by a plurality of field models MDL1 having different ground surface parameters from each other is generated.
[0069] The optimization processing unit 154 calculates the difference (for example, traveling error, etc.) between each of the plurality of pieces of virtual behavior data generated under the condition that the robot parameters are the same and the ground surface parameters are different from each other, and the actual behavior data.
[0070] The optimization processing unit 154 estimates the ground surface parameters of the field where the robot actually traveled in the demonstration experiment by using the calculated difference and the objective function. Thereby, for example, among the plurality of pieces of virtual behavior data, it is estimated that the ground surface parameters of the virtual behavior data having the smallest difference from the actual behavior data are the same as or similar to the ground surface parameters of the field.
[0071] When the optimization processing unit 154 estimates the ground surface parameters of the field, it changes the temporarily fixed ground surface parameter A to the estimated ground surface parameter (hereinafter, referred to as ground surface parameter B for convenience).
[0072] The simulation unit 153 repeatedly causes the robot model MDL2 to travel while changing the robot parameters on the field model MDL1 in which the ground surface parameter is changed from A to B. Thereby, a plurality of pieces of virtual behavior data having different robot parameters from each other are generated under the same condition of the ground surface parameter B.
[0073] In this way, the simulation unit 153 generates a plurality of virtual behavior data with different ground surface parameters for each robot parameter by repeating the simulation while changing the robot parameters and the ground surface parameters.
[0074] Next, the optimization processing unit 154 determines a robot parameter that is robust against the ground surface parameters A and B based on the plurality of virtual behavior data generated for each robot parameter (O. Optimization in the figure). For optimization, for example, evolutionary optimization or deep learning may be used. By determining the robot parameter through such optimization, it is possible to design a robot that can run robustly in any field situation without being affected by the ground surface such as the hardness, viscosity, and height of the field.
[0075] Also, as described above, when a plurality of virtual behavior data with different ground surface parameters are generated, the optimization processing unit 154 may estimate the ground surface parameters of the field based on the difference between each of the plurality of virtual behavior data and the actual behavior data.
[0076] FIG. 11 is a diagram showing an example of actual behavior data. FIGS. 12 and 13 are diagrams showing an example of virtual behavior data. In the example of FIG. 11, it represents the change in state S when the robot travels in a field with unknown ground hardness. In the example of FIG. 12, it represents the change in state S when the robot model MDL2 travels on the hard ground field model MDL1. In the example of FIG. 13, it represents the change in state S when the robot model MDL2 travels on the soft ground field model MDL1.
[0077] In the illustrated example, the actual behavior data is closer to the virtual behavior data when the ground surface is hard than the virtual behavior data when the ground surface is hard. In such a case, the optimization processing unit 154 may estimate that the ground surface of the field is soft.
[0078] According to the embodiments described above, the analysis device 100 acquires an aerial image IMG of a farm field and generates a farm field model MDL1 from the aerial image IMG. The analysis device 100 calculates the behavior of the robot model MDL2 by simulation on the generated farm field model MDL1. Then, the analysis device 100 optimizes a parameter set including robot parameters and ground surface parameters based on virtual behavior data indicating the behavior of the robot model MDL2. As a result, a simulation environment for the robot considering the influence of the ground can be created. Consequently, the robot can be designed more suitably.
[0079] (Other Embodiments) Hereinafter, modifications of the above-described embodiments will be described. In the above-described embodiments, it has been described that the parameter set to be optimized includes, for example, robot parameters and ground surface parameters.
[0080] In addition to or instead of the above-described examples, the robot parameters may include various other parameters such as parameters related to the mesh of the robot model MDL2 (such as vertex positions of the mesh) and parameters related to the arrangement of cameras mounted on the robot.
[0081] In addition to or instead of the above-described examples, the ground surface parameters may include various other parameters such as parameters related to the point cloud of the farm field model MDL1 (such as point cloud positions).
[0082] Furthermore, in addition to robot parameters and ground surface parameters, the parameter set to be optimized may further include parameters of a control system that outputs torque τ from state S (for example, weight coefficients and bias components of a neural network). Also, the parameters of the control system may be treated as robot parameters in a broad sense.
[0083] As described above, the embodiments for carrying out the present invention have been described using the embodiments. However, the present invention is not limited to such embodiments, and various modifications and substitutions can be made without departing from the gist of the present invention.
Explanation of Reference Numerals
[0084] 100…Analysis device, 110…Communication interface, 120…Input interface, 130…Output interface, 140…Storage unit, 150…Processing unit, 151…Acquisition unit, 152…Generation unit, 153…Simulation unit, 154…Optimization processing unit, 155…Output control unit
Claims
1. An acquisition unit that acquires an image of a robot's workplace captured from above; A generation unit that generates a workplace model, which is a virtual model of the workplace and is represented by a three-dimensional point cloud, from the image; A simulation unit that calculates, by simulation, the behavior of a robot model, which is a virtual model of the robot, when the robot model is operated on the workplace model; An optimization processing unit that optimizes a robot parameter, which is a parameter related to the design and control of the robot model, and a ground surface parameter, which is a parameter related to the ground surface of the workplace model, based on virtual behavior data, which is data indicating the behavior of the robot model; An analysis device comprising the above.
2. The acquisition unit acquires actual behavior data, which is data indicating the behavior of the robot when the robot is operated in the workplace; The simulation unit repeats the simulation of operating the robot model having the fixed robot parameter while changing the ground surface parameter of the workplace model, thereby generating a plurality of virtual behavior data in which the ground surface parameters of the workplace model are different from each other; The optimization processing unit estimates the ground surface parameter of the workplace based on the difference between each of the plurality of virtual behavior data and the actual behavior data; The analysis device according to Claim 1.
3. The simulation unit repeats the simulation while changing the robot parameter and the ground surface parameter, thereby generating, for each robot parameter, a plurality of virtual behavior data in which the ground surface parameters are different from each other; The optimization processing unit determines a robot parameter that is robust with respect to a plurality of the ground surface parameters based on the plurality of virtual behavior data generated for each robot parameter; The analysis device according to Claim 1 or 2.
4. In the simulation, the simulation unit calculates the contact force of the workplace model on the ground surface by the entire robot model as the behavior; The analysis device according to Claim 1 or 2.
5. The robot model is represented by a mesh model; The simulation unit Based on the distance and speed when each point of the workplace model represented by the three-dimensional point cloud penetrates into the mesh of the robot model, calculate the contact force on the mesh at each point, By integrating the contact forces at each of the points in the point cloud, calculate the contact force on the ground surface of the workplace model by the entire robot model, The analysis device according to claim 4.
6. An analysis method using a computer, comprising: Obtaining an image of a robot's workplace taken from above; Generating a workplace model, which is a virtual model of the workplace and is represented by a three-dimensional point cloud, from the image; Calculating, by simulation, the behavior of the robot model when the robot model, which is a virtual model of the robot, is operated on the workplace model; Optimizing a robot parameter, which is a parameter related to the design and control of the robot model, and a ground surface parameter, which is a parameter related to the ground surface of the workplace model, based on virtual behavior data, which is data indicating the behavior of the robot model; An analysis method including the above steps.
7. A program for causing a computer to execute, comprising: Obtaining an image of a robot's workplace taken from above; Generating a workplace model, which is a virtual model of the workplace and is represented by a three-dimensional point cloud, from the image; Calculating, by simulation, the behavior of the robot model when the robot model, which is a virtual model of the robot, is operated on the workplace model; Optimizing a robot parameter, which is a parameter related to the design and control of the robot model, and a ground surface parameter, which is a parameter related to the ground surface of the workplace model, based on virtual behavior data, which is data indicating the behavior of the robot model; A program including the above steps.