Robot control
Overview of the disclosure
The disclosure relates to a means for controlling a welding robot to improve the feedback control at the welding head. In particular, the disclosure relates to a human interface correction method of controlling the welding head of a robotic welding tool.
Background of the disclosure
Welding is used in a large number of engineering processes to join components together and for repair of components. Rather than using manual welding which requires a skilled human operator there is growing work in using robotic welding tools. These are desirable for their ability to weld large areas quickly and repeatably. Many welding robots use robots which have 6 degrees of freedom. Most applications of industrial robots focus on point-to-point movements for spot welding, pick and packing, assembly, etc. and lack of human input during the process. In order to improve the functionality of robotic welding much of the effort has been put into the design of robotic solutions with payload capacities improvements to assist a wide range of applications and industries. Despite all the developments in the structure of the robots themselves, their controllers typically work based on point-to-point vectoring of the desired paths. This controller method is acceptable for many applications, such as spot welding, assembly, part manipulations and even machining with such methods often being used in the automotive industry. Currently, most applications of industrial robots are based on movement among waypoints which results in relative "rugged" performance, especially when it comes to accurate and continuous path tracking. For example, current industrial welding robots are articulated arms with a pre-programmed set of movements based on separate waypoints, which lack the possibility to adjust the real-time path/velocity by skilled engineers. Another issue is the stability of moving the robotic end-effectors, these are very sensitive when it comes to "continuous" time-dependant processes such as welding/gluing/coating/water jet cutting. The reason for this sensitivity is that a small change in both the distance and/or the velocity between the end effector and the component may produce a considerable variation in the process outcomes; this can result in a component being acceptable and not acceptable for its desired purpose. Certain robotic energy-based operations (e.g. welding) require "smooth" and continuum end-effector manipulations and as such are formed around this. The use of these types of robots can overcome limitations in the shortage of the skilled workers, especially the experienced welder (e.g. left-handed welder). This is because the head can be manipulated to provide a smooth weld. It is further desirable that industrial robots can mimic or replicate the characteristics which human limbs possess by including feedback and accurate movability of the effector. This would increase the control and movability of the effector performance increases the smoothness of the process not only in terms of the path but also in velocity to increase the manufacturing quality and allow for fast dynamic response result in a set of challenging problems for robotic control. These systems need advanced control methods which can allow the robots to mimic the capabilities (e.g. path and velocity "smoothness") of human operators, which are able to accept real-time corrections to fulfil critical applications. Consequently, it is desirable to develop a human-robotic collaborative control system for accurate path tracking and subject to unknown external disturbances and multiple physical constraints.
Summary of the Disclosure
According to a first aspect of the disclosure there is presented a robotic welding tool comprising a robotic arm having a welding head attached to an end effector of the robotic arm, a robot controller, which provides signals to the robot arm to manipulate the motion of the robotic arm within three dimensional space, and a computer; the computer having a program which provides signals to the robotic controller to control the motion of the robotic arm, the computer having a memory in which the program can be stored for running the welding process, wherein the program for the weld comprises a series of spatial co-ordinates through which arm is required to move in order to perform the desired welding process, the computer further having an operator input for real-time control of the robotic welding head, and wherein the program weld instructions, the operator input and the position of the robot arm are fed into a model predictive controller which adjusts the co-ordinates for the welding head based upon the input signals.
The robotic arm may be coupled to a compensator, with the signal from the compensator being fed into the model predictive controller (MPC) and wherein the compensator and the robotic arm form a disturbance compensation loop with the input from the MPC.
The MPC may utilise a receding horizon strategy and provide the control input based on the online optimization.
The balance between the input from the program with the weld instructions and the human input may be achieved using such a MPC as the human input will update the current desire path of MPC.
A communication frequency of between 250 and 500 Hz may be used to control the position of the robotic welding head.
The welding head may be provided with at least one camera, the image from which is displayed to the operator to assist with movement of the robotic arm.
The image may be displayed to a virtual reality headset and the camera system provides a three dimensional image, which is processed by the computer before being displayed to the operator.
The operator may be provided with a user input control device, manipulation of which causes a signal to be fed into the MPC as the user input signals.
The welding end-effector may be provided with sensors for detecting pressure and/or movement of the head, and wherein the human input controller is provided with actuators that provide feedback to the controller and to the operator.
The input from the arm into the CPC and the signal from it back to the arm may be provided as tightened constraints.
The robot arm may be a 6 degrees of freedom robotic arm.
According to a second aspect of the disclosure there is provided a method of controlling a robotic arm welder according to any of the above first aspect, the method comprising: programming the computer with the co-ordinates for a desired weld path; the robot end effector is moved into the start position of the welding process; the welding process starts and a signal is fed to the operator through the human interface section; the robot follows the welding path as determined by the computer, with additional correction input from the operator via the human interface to correct the motion of the robot; the weld process is completed and the computer program instructs the robot arm to return to an initial position.
The welding path may be provided by a CPC which utilises a receding horizon strategy.
The component to be welded may be brought to a fixed position relative to the position of the robot.
Feedback may be provided through sensors being present on the welding effector.
The skilled person will appreciate that except where mutually exclusive, a feature described in relation to any one of the above aspects may be applied mutatis mutandis to any other aspect. Furthermore, except where mutually exclusive any feature described herein may be applied to any aspect and/or combined with any other feature described herein.
Brief description of the Drawings
Embodiments will now be described by way of reference only, with reference to the figures in which: Figure 1 presents a prior art example of a welding robot; Figure 2 presents an example of a welding robot according to the present disclosure; Figure 3 presents a system diagram of the operation of a welding robot according to the present disclosure; Figure 4 presents a flow chart of the operation of a welding robot according to the present disclosure.
Detailed Discussion Aspects and embodiments of the present disclosure will now be discussed with reference to the accompanying figures. Further aspects and embodiments will be apparent to those skilled in the art.
Figure 1 presents a prior art example of a welding robot. In this case the robot arm is a 6-D0F robot arm mounted to a base. The robot is connected to a robot controller and a computer. The computer is provided with enough memory and processing capability to run the program and any other software that is needed. The robot is controlled trough the computer providing an output to the robot controller which in turn controls the motion of the motors and actuators within the robot. The computer also controls the welding tool, which is mounted at the end of the robot arm in the position of the end effector. Manipulation of the shape of the arm by the three moveable sections allows the robot to be in the correct position and allows it to follow the path required by the welding tool. Each moveable section is coupled to a joint which can be manipulated by variations to angle and rotation allow for a change to the shape of the robot. Because three joints need moving and rotating to move the robot to the next point within the program means that all three actuators need to be addressed and means that the movement of the robot is complex and which slows the robot down. Consequently, in order to carry out a task the path that the arm needs to follow is programmed into a computer, the computer translates this path into a 3D series of movements of each of the sections the robot. The computer sends this out as a series of instructions to the actuator controllers which send the signals to the actuators to move the sections of the robot. Consequently, the welder only follows a set path. Therefore, to change or adapt the component that is to be welded a new path needs to be programmed into the computer; this will provide the co-ordinates to which the robotic welder as it is moved towards and along as the program is executed by the robot controller and the robot itself. Because the motion of the robot is pre-determined by the computer which sends the signals to the robot it means that there is required either accurate positioning of the component before the start of the process or accurate sensing of the position of the robotic system so that the position of the robot at each stage can be accurately determined. If this is not achieved, it will result in errors in the process which could result in faults in the final component.
Figure 2 presents an example of the welding robot according to the present disclosure. The robot configuration is similar to the prior art example that is presented in figure 1. The main difference between the figures is the presence of a user input device. There is also the option as shown of having a virtual reality headset for the operator. The presence of the input device, which may be a joystick, or any other haptic devices which can provide the operator with feedback and a sense of the motion of the tool. This could be through gyroscopes that resist motion or transducers to provide contact feedback. The end effector may be provided with motion and/or force sensors. The output of the sensors may be simulated by the input device; thus, providing the operator a sense of feedback. The end effector may also be provided with a camera. Alternatively, it may be provided with a plurality of cameras. The images from the camera(s) can be processed by the computer and displayed to the user. This display may be on a visual display unit. The visual display unit may be a monitor or a virtual reality headset. This can allow the user to visualise the area around the welding tool. The presence of the user input device feeds into the computer program and provides an input signal that allows the user to provide extra movement signals to the robot which can control the position of the tool. This signal may be fed in conjunction with the program with the user input being weighted to be more important than the signal from the computer program. Additionally or alternatively, the computer program may provide the signals to the robot controller to move the robot into position so that the operator can control the welding tool during the welding process and when the welding process is completed the program can then be used to safely move the robot away from the operating position.
Figure 3 presents a system diagram of the operation of the robot according to the present disclosure. A nominal model (311) which describes the relationship between the control input and the corresponding output position and velocity of the end-effector is established without considering the external disturbance and uncertainties. A model predictive controller (MPC, 307) is then developed based on the nominal model considering the tightened state constraints (306) and input constraints (309) instead of normal state constraints (305) and input constraints (308). The output of the M PC is the velocity which will drive the robot (302) move smoothly. To accommodate the disturbances that act on the robot, a compensator (303) is used to achieve a process tracking of the desired path. Since the tightened constraints are considered during the MPC design, the input of the compensator will not influence the online optimization of MPC. The MPC also receives the external signal that is received by the human interface device. The received signal is considered as a modification of the desired path. Once the human operator gives the input to improve the quality of the welding, the controller will track the new desired path. The real-time modification can be guaranteed by the fast dynamic response of the proposed controller. The degree to the sensitivity of the MPG to the human operator input can be set on the control computer. It can be set so that the system is mainly biased for small correction from the human operator or for larger correction and more control by the human operator. The overwrite by the human operator of the input signal provides the smooth and accurate control of the robot head that is required.
This set-up using a model predictive controller (MPC) with disturbance rejection strategy allows for accurate following of a track along a pre-planned trajectory; this allows the system to work whilst being subject to unknown external disturbances and multiple constraints. A pre-planned trajectory is considered as a desired signal fed into the controller. A number of pre-planned trajectories may be stored within the computer that is used to control the robot if it used on a number of different applications. The set-up allows for real-time feedback from the industrial robot, to be fed back into the MPC and this input can be compared against the pre-planned values. From this the controller determines the control action that is required based upon current system states and allows it to stabilize the system. The robotic system can be remotely driven by the MPC for accurate path tracking control. Additionally, as discussed above the MPC further has a signal; from a human operator, which further allows for real-time human modification. The human modification may be based upon on the outputted tracking performance. The control signal is provided as the speed for the x, y and z axis with these values containing two components. The two values are the nominal control input to stabilize the system and cope with the constraints, and the compensator control to cope with the external disturbances. The nominal control input is generated based on the nominal model without considering disturbances. The external disturbances act to drive the trajectory away from the desired or pre-planned path and as such the compensator is provided in the designed control scheme for accurate welding. Furthermore, the constraints considered to generate the nominal control input are also tightened from the original constraints to ensure that the control input can be bounded according to the constraints in all cases subject to external unknown disturbances. The MPC may work using a receding horizon strategy. The receding horizon strategy works by generating a control sequence by online optimization, and only a first element of the control sequence is applied to the system. This optimization provides smooth movement which is needed in continuous welding and gluing for complex components. By using the compensation control input, the unknown external disturbances can be eliminated, and the closed-loop dynamics of the actual model approximates the nominal model. Therefore, the use of such a control system means that the robotic system is robust against external disturbances during the path tracking process; this further improves the control performance of the robotic system. By utilising real-time input into the robotic control system, the controller is able to meet the unique requirements of human-robot collaboration for use in industrial manufacturing and inspection. The input from the operator is treated as a modification of the pre-planned path and the robot can be driven to follow the new path with a fast dynamic response which is guaranteed by the proposed controller. The human-robot collaboration requires high communication frequency (up to 500Hz) between the robot and the remote PC for real-time data exchanging. The communication frequency of the control system is suitable for other devices such as the haptic control device may have a minimum 250 Hz frequency. The control algorithm is programmed can be based on Python and can be executed on standard operating systems so that it can be applied straightforwardly to many kinds of industrial robots.
Figure 4 presents a flow chart of the method of operation of the welding robot. In step 401 the computer program is programmed to perform a weld path. This follows the path that the weld needs to follow that in order to join the components together. In step 402 the robot end effector is moved into the start position of the welding process. In step 403 the welding starts and a signal is fed to the operator through a camera within the operating area. Feedback may also be provided through sensors being present on the welding effector. In step 404 the robot follows the welding path as determined by the computer, with additional input from the operator to correct the motion of the robot, from that performed by the computer program alone. The bias to which the operator can correct the program can be predetermined before the operation is stared. The input of the operator is fed into the MPC which balances the signal from this with that of the computer program. In Step 405 the weld process is completed and the computer program instructs the robot arm to return to an initial position. The initial position is the safe position from which the component to be welded can be placed in position for the weld to be carried out. The robotic welder may be part of a production line, so that the component is delivered into a working position relative to the robot. Alternatively, the component can be brought into the working are of the robot individually, or the robot being moved into position relative to the component to be processed. The computer may additionally have a number of processes pre-programed into it which can be selected by the operator prior to the process starting and from which the input of the operator is able to correct the weld path of the particular welding job that needs to be performed. The operator may not need to be present along with the robot. The operator may be able to perform the task remotely, this can be done by utilising multiple computers one with one being positioned with the robot and is used to control the robot and having the MPC. A second computer may then be connected through a suitable networking connecting means, such as Bluetooth or a secure internet connection or through a network cable. This remote control of the welder allows the robot to weld in areas where a human cannot operate or in which it is dangerous for a human operator to work in. Machine learning can be utilised to capture and further mimic the motion of human operators in applications highly relying on human experience, such as welding. Other learning methods and Al can be employed for computer vision and uncertainties in the process to further improve the system performance in general.
A human computer robot has been shown using a model predictive control which combines a pre-determined input with a human operator controller to correct the motion of the welding torch. This can be achieved by using a MPC and sliding mode disturbance compensator. The MPC can provide a smooth speed control and fast dynamic response and the disturbance compensator can deal with the external disturbance to guarantee the accurate path tracking. To rule out the possibility of the constraint violation caused by disturbance compensation, tightened constraints are formulated to generate the control input signal. Thus, the proposed controller drives the robotic system remotely with enhanced smoothness and real-time human modification on the outputted performance so that the human experience can be fully transferred to robotic systems. The proposed human-robotic collaborative system is suitable for continuous welding of complex curves and other repair and inspection applications which relies on human experience.
It will be understood that the disclosure is not limited to the examples above-described and various modifications and improvements can be made without departing from the concepts described Wherein. Except where mutually exclusive, any of the features may be employed separately or in combination with any other features and the disclosure extends to and includes all combinations and sub-combinations of one or more features described herein.