Robot control system
By adopting a multi-processor structure in the robot control system, independently processing image analysis and motion control, the problem of large amount of image sensing and motion control functions in advanced autonomous action functions is solved, and more efficient real-time and robot action is achieved implement.
Patent Information
- Application Number
- CN202080009953.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Priority Date
- 2019-01-25
- Filing Date
- 2020-01-20
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2040-01-20
AI Technical Summary
In robots with advanced autonomous action functions, the computational volume of image sensing and motion control functions is large, resulting in the need to establish mechanisms to implement these functions immediately to ensure that the robot can perform complex actions such as dancing and other operations efficiently.
The control system adopts a multi-processor structure, including an application processor, an image sensor processor and a motion control processor. The image sensor processor independently performs image analysis and communicates directly with the motion control processor to reduce the load on the application processor and improve the Real-time.
By separating the processor's functions, the processing load of the application processor is reduced, the immediacy and efficiency of the system are improved, and the robot can respond quickly and perform complex actions efficiently.
Smart Images

Figure CN113316504B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to a system for controlling the motion of a robot. Background Art
[0002] Conventionally, various types of robots have been studied and developed. PTL 1 discloses a learning method for walking control of a humanoid type legged mobile robot. PTL 1 discloses a method for reinforcement learning of a stable walking trajectory along which a robot can walk when the robot cannot stably walk along a walking trajectory given in an initial stage.
[0003] [Citation List]
[0004] [Patent Document]
[0005] [PTL 1]
[0006] Japanese Patent Publication No. 2005-96068 Summary of the invention
[0007] [Technical issues]
[0008] Robotics is developing day by day. Although pet-type robots with four-legged walking have become the mainstream of commercial models, humanoid robots that perform complex operations such as dancing are becoming popular in recent years. It can be expected that by increasing arithmetic operation capabilities and improving learning models, robots themselves learn to improve existing functions and acquire new functions.
[0009] The purpose and mode of action of an autonomous robot depend on the application to be performed. An application for making a robot walk autonomously collects and analyzes environmental information around the robot, and calculates the rotation speed of multiple motors constituting the robot joints. A microcomputer for control is provided for each motor, and multiple microcomputers supply power to the motors in synchronization with each other according to the calculated rotation speed. For example, an application with the purpose of following a user determines the position of the user and obstacles in space based on a camera image, determines a path to follow the user, and calculates the rotation speed of the motors of the robot joints so that the robot moves on the path.
[0010] In the robot, the rotation speeds of a plurality of motors are calculated based on accurate data obtained by image sensing, and the motors are rotated at the calculated rotation speeds without any time delay to realize an action mode according to the purpose of the action. In the case of making the robot have a high-level autonomous action function, the amount of calculation applied becomes large. However, in order to allow the robot to act even in such a case as just described, it is necessary to establish a mechanism for immediately realizing the image sensing function and / or the motion control function.
[0011] [Problem Solution]
[0012] In order to solve the above problems, a robot control system according to one aspect of the present invention includes an application processor for processing an application, a sensor processor for acquiring image data from an image sensor and analyzing the image data, and a motion control processor for controlling a movable part of a robot. The motion control processor provides posture information for specifying the orientation of the image sensor to the sensor processor without passing through the application processor.
[0013] In addition, any combination of the above-mentioned constituent elements, and any combination of expressions of the present invention converted between methods, apparatuses, systems, computer programs, recording media recording readable computer programs, data structures, etc. are also effective as modes of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS
[0014] Figure 1 is a diagram depicting a general configuration of the entertainment system of the embodiment.
[0015] Figure 2 A view that depicts the hardware configuration of the control system. DETAILED DESCRIPTION
[0016] Figure 1 The general configuration of an entertainment system 1 of an embodiment is depicted. The entertainment system 1 includes a robot device 20 and a server device 10, which is a mobile object that can walk autonomously and is connected to the server device 10 via a network 2 such as the Internet through an access point (AP) 3 for communication.
[0017] The robot device 20 is configured as a humanoid robot and is owned by a user who is an owner. Preferably, the robot device 20 can identify the owner through facial authentication based on image analysis, voiceprint authentication based on voice analysis, etc. By identifying the owner, the robot device 20 can, for example, only accept instructions from the owner and act according to the owner's instructions.
[0018] Preferably, the robot device 20 has parts similar to those of a human being and has an external shape that gives a friendly feeling. The robot device 20 has a head, a neck, a trunk (chest, abdomen and back), upper limbs and lower limbs. The upper limbs can each have an upper arm, a forearm and a hand, and the lower limbs can each have a thigh, a calf and a foot. These parts are connected to each other by an actuator. The actuator includes at least a motor arranged at a joint portion as a movable part and a connecting rod mechanism that connects one motor to another motor. The robot device 20 moves according to the purpose of the action, while maintaining posture balance by driving the actuator.
[0019] The robot device 20 implements basic functions including walking and running functions and an obstacle avoidance function through a basic application program (hereinafter also referred to as a "basic application") that describes a control method of each actuator. Since the basic application is responsible for the basic functions of the robot device 20, it is preferably pre-installed in the robot device 20, and it may be incorporated into the middleware, for example.
[0020] Any application program other than the basic application program is an application program (hereinafter also referred to as an "application program of application"). The application program of application program realizes an additional function, such as a dance function. The application program of application program is provided from the server device 10 as needed and installed in the robot device 20. The robot device 20 acquires a new function by downloading and installing a new application program of application program.
[0021] Figure 2 2 is a diagram showing the hardware structure of the control system 4. The control system 4 of the present embodiment has a structure for directly realizing the image sensing function and the motion control function of the robot device 20. The control system 4 has a multi-processor structure, including an application processor 30 for processing an application program, a sensor processor 50 for processing image data, and a motion control processor 70 for controlling the motion of the movable part of the robot device 20. In the control system 4, the application processor 30 and the sensor processor 50 constitute an image analysis system, and the application processor 30 and the motion control processor 70 constitute a motion control system. The application processor 30, the sensor processor 50, and the motion control processor 70 are provided in the housing of the robot device 20, and two of the processors are connected to communicate with each other.
[0022] In this architecture, the application processor 30 operates as a main processor. The sensor processor 50 performs image analysis specified by the application processor 30 and provides the image analysis result to the application processor 30. The motion control processor 70 drives and controls the actuator at a rotation speed specified by the application processor 30. Since the sensor processor 50 and the motion control processor 70 perform corresponding processing respectively, the application processor 30 does not need to have an image sensing function and a motion control function, and the processing load on the application processor 30 can be reduced.
[0023] The application processor 30 runs a first operating system (hereinafter also referred to as "first OS") to process applications. The applications to be processed include basic applications and applications of applications. While processing a plurality of basic applications at the same time, the application processor 30 also processes applications of necessary applications. Therefore, the application processor 30 includes a plurality of central processing unit (CPU) cores therein. Preferably, the first OS is a general-purpose OS that allows various applications to be executed simultaneously.
[0024] The display device 32, the speaker 34, the communication device 36, and the microphone 38 are connected to the application processor 30. The display device 32 and the speaker 34 respectively output images and sounds generated by the application processor 30 to present information to the user. The communication device 36 establishes a connection with the AP 3 through a wireless local area network (LAN) and communicates with the server device 10 through the AP 3. The communication device 36 may have a portable phone communication function. The communication device 36 can download application programs and patch files from the server device 10. The microphone 38 collects ambient sound, converts it into a sound signal, and provides the sound signal to the application processor 30. The application processor 30 may have a sound recognition function so that it determines an action based on a user's voice instruction.
[0025] The sensor processor 50 is a processor that performs an image sensing function, and is connected to the image sensor 52, the distance sensor 54, and the event-driven sensor 56. The image sensor 52, the distance sensor 54, and the event-driven sensor 56 are incorporated into the robot device 20, and operate as a visual sensor that acquires peripheral information of the robot device 20, and the sensor processor 50 is responsible for the recognition function in terms of vision.
[0026] The image sensor 52 is a camera, and provides an RGB image captured at a predetermined period (e.g., 1 / 60 second) to the sensor processor 50. The distance sensor 54 may be a time-of-flight (TOF) distance image sensor, and provides a three-dimensional distance image to the sensor processor 50. The event-driven sensor 56 is a sensor that detects a change in the brightness value of a pixel of an imaging element and provides a combination of the detection time and pixel coordinates to the sensor processor 50. The distance sensor 54 and the event-driven sensor 56 are information sensors that detect information related to an imaging target included in image data.
[0027] The event-driven sensor 56 has a wider dynamic range than that of the image sensor 52, and can accurately detect the movement of the imaging target even in an environment where there is incident light that exceeds the dynamic range of the image sensor 52. In addition, since the event-driven sensor 56 can detect the movement of the imaging target with a higher temporal resolution than the image sensor 52, when the sensor processor 50 processes the image data acquired by the image sensor 52, it can complementarily utilize the movement information acquired by the event-driven sensor 56.
[0028] The sensor processor 50 runs a second operating system (hereinafter also referred to as a "second OS") to process image data acquired by the image sensor 52. The sensor processor 50 may have a neural network that recognizes a target object included in the image data by using a machine learning model. The sensor processor 50 acquires depth information of an imaging target from the distance sensor 54, acquires movement information of the imaging target from the event-driven sensor 56, and analyzes the image data acquired from the image sensor 52 by using the imaging target information. The application processor 30 specifies an image analysis function to be executed for the sensor processor 50 according to an application to be processed.
[0029] For example, when the application processor 30 processes an application for causing the robot device 20 to follow an owner as a follow target, it requests the sensor processor 50 to specify the position of the owner. Upon receiving the request, the sensor processor 50 sets a parameter set for image analysis for identifying the owner from a captured image to the neural network, and performs processing for identifying the owner by image analysis. It should be noted that although the parameter set for owner identification is stored in a storage device (not shown) connected to the sensor processor 50, it may be supplied to the sensor processor 50 from the application processor 30. The sensor processor 50 may perform processing for analyzing image data in consideration of imaging target information acquired by the distance sensor 54 and the event-driven sensor 56.
[0030] The sensor processor 50 may have a learning function for optimizing a parameter set of a neural network in order to implement an image analysis function dedicated to the application. In this example, preferably, the sensor processor 50 performs learning for optimizing a parameter set for owner recognition in advance, and optimizes the parameter set before the application processor 30 executes an application for following the owner.
[0031] The objective function (error function) in the learning function compares the output obtained by inputting the image for learning into the neural network with the correct value corresponding to the image (i.e., the fact that the target person is the owner) to calculate the error. The learning function calculates the gradient of the parameter based on the error by the gradient back propagation method, etc., and updates the optimized target parameter of the neural network based on the momentum method. By performing application-specific learning, an optimal parameter set can be created for each application, and an accurate image analysis function can be implemented. It should be noted that the sensor processor 50 can have different types of learning functions.
[0032] For example, when the application processor 30 processes an application for the robot device 20 to predict the destination of a specific moving object and reach the destination before the object reaches the destination, the sensor processor 50 predicts the movement trajectory of the specific object while performing image analysis to identify the object. Since the event-driven sensor 56 detects the movement information of the imaging target with high time resolution, the movement information can be appropriately utilized to perform accurate movement prediction. The sensor processor 50 predicts the movement trajectory of the specific object based on information from the image sensor 52, the distance sensor 54, and the event-driven sensor 56.
[0033] Furthermore, when the application processor 30 processes an application in which the robot device 20 performs movement synchronized with movement such as the owner's gesture, the sensor processor 50 can recognize the owner's movement while performing image analysis for recognizing the owner, and can also predict future movement. The sensor processor 50 can acquire the owner's movement information detected by the event-driven sensor 56 to recognize the owner's current movement, and predict the movement to be made based on the movement information obtained so far. If the sensor processor 50 predicts the owner's movement, the robot device 20 can move in anticipation of the owner's movement.
[0034] The second OS executed by the sensor processor 50 in the above manner is preferably an OS that controls a neural network for image analysis. The second OS is a dedicated OS suitable for a neural network, and therefore, the second OS is not a general-purpose OS and is different from the first OS.
[0035] The motion control processor 70 is a processor for controlling the motion of the movable part of the robot device 20, and is connected to a plurality of motion sensors 72, a plurality of touch sensors 74, and a plurality of microcomputers 76a to 76n (in the following description, when they are not specifically distinguished from each other, each of them is referred to as a "microcomputer 76"). One microcomputer 76 is responsible for driving one motor. The motion control processor 70 is responsible for the motion control function of the robot device 20, and controls the microcomputer 76 to supply a driving current to the motor.
[0036] Each motion sensor 72 includes a three-axis acceleration sensor and a three-axis gyroscope sensor. The motion sensor 72 can be set at each joint part, and provide the motion control processor 70 with sensor data indicating the position and posture of the joint part in three-dimensional space and / or sensor data indicating changes in position and posture. It should be noted that in this embodiment, the motion sensor 72 can be set at the position of the image sensor 52, so that the motion control processor 70 is provided with sensor data indicating the position and posture of the image sensor 52 in three-dimensional space and / or sensor data indicating changes in position and posture. Each touch sensor 74 is set on the outer surface of the robot device 20, and detects contact with the robot device 20.
[0037] The motion control processor 70 runs a third operating system (hereinafter also referred to as a "third OS") to control the motion of the movable part of the robot device 20, that is, the rotation of the motor. The role of the motion control processor 70 is to manage the microcomputer 76 in real time so that each motor rotates at the rotation speed calculated by the application processor 30, and its more important role is to prevent the robot device 20 from falling. To this end, as a prerequisite for meeting the motion requirements from the application processor 30, the motion control processor 70 usually monitors the sensor data of the motion sensor 72 and performs posture control applications for fall prevention. For example, when the robot device 20 hits an obstacle and is about to fall, the motion control processor 70 immediately performs motor control for fall prevention.
[0038] Thus, the motor control of the motion control processor 70 requires immediacy. Therefore, the third OS run by the motion control processor 70 is preferably a real-time OS. A real-time OS is an OS designed so that a specific application such as a gesture control application is executed at timing with high precision and can perform real-time processing of the application. Naturally, the third OS is different from the first OS and the second OS.
[0039] The application processor 30, the sensor processor 50, and the motion control processor 70 are configured in the manner described above. For example, when an application for causing the robot device 20 to follow an owner as a follow target is to be executed, the application processor 30 needs to find the position of the owner in space and the position of an obstacle in space that exists between the robot device 20 and the owner. Therefore, the application processor 30 instructs the sensor processor 50 to specify the positions of the owner and the obstacle included in the image data of the image sensor 52. In the following description, each of the owner and the obstacle that will be the recognition target of the application processor 30 is simply referred to as a "target".
[0040] The sensor processor 50 acquires image data from the image sensor 52, and performs image analysis specified by the application processor 30 by using the depth information of the distance sensor 54 and the motion information of the event-driven sensor 56. The sensor processor 50 provides the application processor 30 with an image analysis result, which includes the position coordinates of each target in a camera coordinate system with the image sensor 52 as the origin. The image analysis result also includes the depth information (distance information) of the target in the camera coordinate system. Therefore, the application processor 30 can specify the three-dimensional coordinates of the target in the camera coordinate system, which has the origin at the image sensor 52.
[0041] The motion control processor 70 provides the sensor processor 50 with posture information for specifying the orientation of the image sensor 52. The posture information may include information for specifying the position of the image sensor 52. In short, the posture information includes information for specifying the orientation and position of the optical axis in the space when the image sensor 52 acquires image data. The position of each of the joints and main parts in the robot device 20 is defined by a robot coordinate system in which the origin is a reference position in the robot body. The motion control processor 70 continuously calculates the three-dimensional coordinates of the reference position in the three-dimensional real space coordinate system (spatial coordinate system), and creates posture information for specifying the orientation and position of the image sensor 52 in the spatial coordinate system by specifying the orientation and position of the image sensor 52 in the robot coordinate system. The sensor processor 50 sends the image analysis result together with the posture information of the image sensor 52 to the application processor 30.
[0042] The application processor 30 specifies the orientation and position of the image sensor 52 in the space coordinate system by acquiring the posture information, and specifies the three-dimensional coordinates of the image sensor 52 in the camera coordinate system by acquiring the image analysis result. The application processor 30 thereby acquires the three-dimensional coordinates of the target in the space coordinate system.
[0043] The motion control processor 70 provides the sensor processor 50 with posture information obtained when the image sensor 52 acquires image data to be analyzed by the sensor processor 50. The motion control processor 70 can derive the posture information of the image sensor 52 from the sensor data of the motion sensor 72 provided at each joint portion. Note that when the motion sensor 72 is provided at the installation position of the image sensor 52, the motion control processor 70 can derive the posture information of the image sensor 52 by using the sensor data of the motion sensor 72.
[0044] The sensor processor 50 notifies the motion control processor 70 of the image capture time of the image data for analysis, and the motion control processor 70 provides the posture information of the image sensor 52 at the same time as the image capture time to the sensor processor 50. The motion control processor 70 continuously executes the posture control application, and calculates the position coordinates of the joint part and the main part in a sampling cycle that is much shorter than the image analysis cycle of the sensor processor 50 (for example, 1 / 60 second). The motion control processor 70 in the embodiment can derive the posture information of the image sensor 52 in the sampling cycle, and store the posture information together with the time information in a storage device (not shown), so that when it receives the notification of the image capture time of the image data, it reads out the posture information of the image sensor 52 at the same time as the image capture time from the storage device, and provides the posture information to the sensor processor 50.
[0045] During the analysis of the image data, the pose information from the motion control processor 70 is provided to the sensor processor 50. When the sensor processor 50 finishes analyzing the image data, it adds the pose information of the image sensor 52 as metadata to the image analysis result, and provides the obtained image analysis result to the application processor 30. The application processor 30 specifies the position of the target in space based on the image analysis result and the pose information.
[0046] The motion control processor 70 in this embodiment provides the posture information of the image sensor 52 to the sensor processor 50 not via the application processor 30. In short, the motion control processor 70 directly provides the posture information to the sensor processor 50 through a bus connecting the motion control processor 70 and the sensor processor 50 to each other.
[0047] In a common architecture design, since the application processor 30 as the main processor relays data transmission between other processors, the data transmission time is the sum of the relay time of the application processor 30. On the other hand, since in the control system 4, the sensor processor 50 and the motion control processor 70 are connected to each other through a bus, when the motion control processor 70 provides the sensor processor 50 with the posture information of the image sensor 52, directly providing the posture information to the sensor processor 50 achieves a shorter transmission time than sending the posture information via the application server 30. Therefore, in order to improve real-time performance, the motion control processor 70 directly sends the posture information of the image sensor 52 to the sensor processor 50.
[0048] During image analysis, the sensor processor 50 may issue a request to the motion control processor 70 to provide posture information of the image sensor 52. At this time, the sensor processor 50 may notify the motion control processor 70 of the image capture time of the image data to be used in the analysis. When the motion control processor 70 receives the request to provide posture information, it directly provides the posture information of the image sensor 52 to the sensor processor 50. When the motion control processor 70 is notified of the image capture time of the image data, it may thereafter cyclically provide the posture information according to the image analysis cycle of the sensor processor 50. For example, when the image analysis cycle of the sensor processor 50 is 1 / 60 seconds, the motion control processor 70 may provide the posture information to the sensor processor 50 in the same cycle from the starting point set at the notified image capture time. It should be noted that the motion control processor 70 may periodically provide the posture information to the sensor processor 50 in other ways, without relying on a request from the sensor processor 50.
[0049] The sensor processor 50 adds the posture information as metadata to the image analysis result, and provides the obtained image analysis result to the application processor 30. Therefore, the application processor 30 can specify the position of the target in the space. It should be noted that although the sensor processor 50 provides the image analysis result, it does not provide the image data used in the image analysis to the application processor 30. In the control system 4, the fact that the sensor processor 50 does not provide the image data to the application processor 30 reduces the risk of transmission delay from the sensor processor 50 to the application processor 30, so that the image analysis result is provided to the application processor 30 without time delay. In addition, the fact that the sensor processor 50 does not provide the image data to the application processor 30 prevents the image data including personal information from being leaked to the outside of the robot device 20.
[0050] It is worth noting that the sensor processor 50 can also provide image analysis results to the motion control processor 70. For example, when the motion control processor 70 has a tracking function of a predetermined target, it can control an autonomous tracking action by using the image analysis results of the sensor processor 50.
[0051] As described above, in the control system 4, the sensor processor 50 and the motion control processor 70 are provided to avoid a situation where the load is concentrated on the application processor 30. In addition, since each of the application processor 30, the sensor processor 50, and the motion control processor 70 runs an OS suitable for the corresponding processing, the real-time performance of the control system 4 is improved as a whole.
[0052] The present invention has been described based on the embodiment. The embodiment is exemplary, and it should be understood by those skilled in the art that various modifications are possible regarding the combination of components and processes, and such modifications also fall within the scope of the present invention. It is sufficient if the robot device 20 is a mobile robot, and the robot device 20 is not limited to a two-legged walking robot. In addition, the robot device 20 may be an industrial robot or the like.
[0053] In the present embodiment, although the motion control processor 70 provides the posture information of the image sensor 52 to the sensor processor 50, the posture information may be provided to the application processor 30. In this case, the application processor 30 does not transmit the posture information to the sensor processor 50 but holds the posture information, and associates the posture information with the image analysis result provided from the sensor processor 50. The application processor 30 refers to the image capture time of the image data used in the image analysis included in the image analysis result and the acquisition time of the sensor data used in the derivation of the posture information included in the posture information, associates the image analysis result and the posture information at the same time, and uses them for determining the position of the target in space.
[0054] In the present embodiment, an example is described in which the application processor 30, the sensor processor 50, and the motion control processor 70 are provided within the frame of the robot device 20. However, in a modified example, the sensor processor 50 and the motion control processor 70 may be provided within the frame of the robot device 20, and the application processor 30 may be provided outside the frame of the robot device 20. By providing the function of the application processor 30 outside the robot device 20, for example, within the server device 10, it is possible to provide the application processor 30 with high processing power, so that the application processor 30 can perform complex calculation processing.
[0055] Although it is described in this embodiment that the sensor processor 50 does not provide image data to the application processor 30, if a predetermined operation mode is set by the user, the sensor processor 50 may be enabled to provide image data.
[0056] When the main power of the robot device 20 is turned off, the event-driven sensor 56, the motion sensor 72, and the touch sensor 74 may be driven by using the backup power. The robot device 20 may be configured so that when at least one of the event-driven sensor 56, the motion sensor 72, and the touch sensor 74 detects predetermined information, the main power is automatically turned on.
[0057] [Industrial Applicability]
[0058] The present invention can be used in a system for controlling the motion of a robot.
[0059] [Reference Symbol List]
[0060] 1: Entertainment System
[0061] 2: Network
[0062] 3: AP
[0063] 4: Control system
[0064] 10: Server device
[0065] 20: Robotic Device
[0066] 30: Application Processor
[0067] 32: Display device
[0068] 34: Speaker
[0069] 36: Communication equipment
[0070] 38: Microphone
[0071] 50: Sensor processor
[0072] 52: Image sensor
[0073] 54: Distance sensor
[0074] 56: Event-driven sensors
[0075] 70: Motion Control Processor
[0076] 72: Motion Sensor
[0077] 74: Touch sensor
[0078] 76: Microcomputer
Claims
1. A robot control system, comprising: an application processor that operates as a main processor and processes applications; a sensor processor that acquires image data from an image sensor and analyzes the image data according to a designation from the application processor; as well as a motion control processor that controls the motion of the movable part of the robot according to the rotation speed specified by the application processor, wherein The motion control processor provides the sensor processor with posture information for specifying the orientation of the image sensor without via the application processor, and the sensor processor adds the posture information as metadata to an image analysis result and provides the obtained image analysis result to the application processor.
2. The robot control system according to claim 1, wherein: The posture information includes information for specifying the position of the image sensor.
3. The robot control system according to claim 1, wherein: The motion control processor provides the sensor processor with posture information obtained when the image sensor acquires image data to be analyzed by the sensor processor.
4. The robot control system according to any one of claims 1 to 3, wherein: The motion control processor derives the posture information from sensor data of a motion sensor provided at a joint portion of the robot.
5. The robot control system according to any one of claims 1 to 3, wherein: The motion control processor derives the posture information from sensor data of a motion sensor provided at a mounting position of the image sensor.
6. The robot control system according to any one of claims 1 to 3, wherein: The sensor processor requests the motion control processor to provide posture information during image analysis.
7. The robot control system according to any one of claims 1 to 3, wherein: The motion control processor periodically provides posture information to the sensor processor.
Citation Information
Patent Citations
Robot device and attitude control method for robot
JP2005096068A
Method for rapid calibrating hand-eye relationship of single eye vision sensor of welding robot
CN101053953A
Production apparatus
CN104275698A
Information processing device and information processing method
CN105917385A
Material sorting system and method
CN108607819A
