Method for flexibly controlling a robot

Abstract

The invention relates to a method for flexibly controlling a robot (10) having at least one robot joint (11a, 11b) for use in a passive environment comprising a passive, articulated object which can be moved in a specified movement path, wherein the robot (10) assumes an actual pose Pt by following a specified target pose PS,t, having the steps of: a0) specifying the target pose PS,t of the robot, the target pose PS,t producing a movement of the robot (10) involving an interaction with the passive, articulated object in the environment, a) detecting the interaction of the robot (10) with the passive, articulated object, the robot (10) transmitting energy to the passive, articulated object in a Cartesian degree of freedom during the interaction, b) determining a physical effect of the passive, articulated object on the robot (10) on the basis of the energy which is transmitted through the specified movement path from the passive, articulated object into another Cartesian degree of freedom or into a plurality of other Cartesian degrees of freedom and which is transmitted to the robot (10), c) changing the specified target pose on the basis of the determined physical effect, and d) controlling the robot (10) by specifying the changed target pose as the target pose PS,t.

Description

[0001] Method for the compliant control of a robot

[0002] The invention relates to a method for the compliant control of a robot for use in a passive environment with a passive, articulated object.

[0003] Robot simulations distinguish between active and passive environments. Robots that share their workspace with humans, for example, are in an active environment. Robot manipulators in unfamiliar environments, such as a disaster area, can also be exposed to active elements that move in the robot's immediate vicinity without the robot's control.

[0004] To mitigate the consequences of collisions between robots and active elements in active environments, such as people, safety measures must be implemented. These are important for both autonomous and teleoperative robots to minimize damage caused by collisions, for example, when a collision is perceived too late by a person and can no longer be avoided.

[0005] As an alternative to current safety standards for active environments, which require a robot to stop as soon as a human is detected in the robot's vicinity, the applicant has developed a compliant control method, which is disclosed in DE 10 2021 102 509 Al.

[0006] DE 10 2014 114 234 A1 discloses a method for controlling a robot in the medical field, wherein a collision with an object is detected and the robot is controlled to reduce the risk of injury or damage. However, robot tasks involving interaction with an articulated object in a passive environment are very complex for the robot due to the kinematics of the articulated object, as the robot must recognize the precise kinematics of the articulated object in order to create a motion trajectory. With previously known methods for robot control in a passive environment, such robot tasks are difficult to control or must be modeled.

[0007] The object of the present invention is to provide a method for controlling a robot, thereby improving the control of the robot during interaction with an articulated object in a passive environment. It is further an object of the present invention to provide a robot equipped with such a control device.

[0008] The method according to the invention is defined by the features of claim 1. The robot according to the invention is defined by the features of claim 11.

[0009] The inventive method for the compliant control of a robot with at least one robot joint for use in a passive environment with a passive, articulated object that is movable in a predetermined path of motion, wherein the robot assumes an actual pose Pt by following a predetermined target pose Ps,t, comprises the following steps:

[0010] aO) Specifying a target pose Ps,t, where the target pose Ps,t causes a movement of the robot with an interaction with the passive, articulated object of the environment,

[0011] a) Detecting the interaction of the robot with the passive, articulated object, wherein during the interaction the robot transfers energy to the passive, articulated object in one Cartesian degree of freedom,

[0012] b) Determining a physical effect of the passive, articulated object on the robot due to the energy transferred by the passive, articulated object through the specified path of motion into another Cartesian degree of freedom or into several other Cartesian degrees of freedom, which is transferred to the robot,

[0013] c) Changing the specified target position depending on the specific physical influence,

[0014] d) Controlling the robot by specifying the modified target pose as target pose Ps,t.

[0015] Passive environments are environments in which no external force source exerts force or energy on the robot. Any force or energy acting on the robot from the passive environment is generated by the robot itself and, after being transferred to the passive environment, is partially returned by it. An articulated object in the passive environment is an object that moves along a predefined trajectory when a force or energy is applied. The robot's task is to move the passively articulated object. To this end, the robot is given a target pose Ps,t, causing it to move in the direction of the target pose and interact with the passively articulated object. During this interaction, the robot transfers energy to the passively articulated object in one Cartesian degree of freedom.

[0016] The interaction of the robot with the passive, articulated object is detected in step a) of the method according to the invention. Detection can be achieved, for example, via the force of a controller, such as an impedance controller, and the robot's velocity. Both the force of an impedance controller and the robot's velocity are values ​​that are already known or at least determinable in conventional robot control systems. Alternatively or additionally, the interaction can also be detected by sensors located on or away from the robot. Sensors located on the robot can, for example, be attached to the robot's drives, measuring the forces acting on the drives as a result of the interaction. Sensors located away from the robot can, for example, be motion sensors in the robot's vicinity.It is also possible to measure the current in the drives to determine the interaction.

[0017] During the robot's interaction with the passive, articulated object, the robot transfers energy in one Cartesian degree of freedom to the passive, articulated object. The passive, articulated object can only move along a predetermined path. As a result, the energy transferred by the robot is at least partially transferred to at least one other Cartesian degree of freedom. The inventive method takes advantage of this fact and, in step b), determines a physical action of the passive, articulated object on the robot based on the energy transferred by the passive, articulated object along the predetermined path of movement to another Cartesian degree of freedom or to several other Cartesian degrees of freedom, which is then transferred back to the robot.In other words, in step b), the robot determines whether energy is transferred back to the robot from the passive, articulated object in another Cartesian degree of freedom.

[0018] Depending on the physical action determined in step b), the predetermined target pose is changed in step c), and the robot is controlled in step d) with the changed target pose. This allows the robot to be controlled along the motion trajectory necessary for the movement of the passive, articulated object. The motion trajectory is essentially developed by changing the target pose depending on the specific physical action of the passive, articulated object on the robot. This makes it possible to control the robot during interaction with the passive, articulated object without knowledge of the object's path of movement or any axes of rotation.

[0019] The robot possesses an actual position Pt at any given time, which defines its current position and orientation. The actual position Pt is the robot's pose at time t. The robot assumes the actual position Pt by following a predetermined target position Ps,t. At time t, the robot is given a target position Ps,t that corresponds to the desired position and orientation. The control process is continuous. Thus, the change in the actual position Pt over time can be considered the actual speed, and the change in the target position Ps,t can be considered the target speed. Due to the physical influence of the passive, articulated object on the robot, a correction speed is added to the target speed over time and used as the modified target speed for control.

[0020] The target position Ps,t can correspond to the actual position Pt when the robot is stationary. For example, an impedance controller moves the robot towards the target position. The passive, articulated object, due to its predetermined path of movement, can exert a physical force on the robot, pushing it away from the target position. When a physical force is detected, the robot's target position is changed to achieve compliance. This change is dependent on and a consequence of the physical force. In other words, when a physical force is detected, the robot is controlled with a modified target position, Ps,t, resulting in greater compliance. This improves the robot's interaction with the passive, articulated object, enabling the robot to perform its assigned task more effectively.If no physical influence is detected in step a), the robot continues to be controlled with the unchanged target position Ps,t.

[0021] Preferably, the following step b) is provided after step bl):

[0022] bl) Checking the energy and / or power acting on the robot through the physical action for one or more other Cartesian degrees of freedom, wherein

[0023] If the energy and / or power acting on the robot due to the physical action exceeds at least one predefined energy threshold Es and / or at least one predefined power threshold Ls in another Cartesian degree of freedom or in several other Cartesian degrees of freedom, steps c) and d) are performed; and if the energy and / or power acting on the robot due to the physical action does not exceed at least one predefined energy threshold Es and / or at least one predefined power threshold Ls, or no physical interaction is detected, the process returns to step aO) and the procedure steps are performed again.

[0024] Therefore, step c), and thus a change in the robot's target pose, only occurs if an interaction is detected in step a) and the energy or power exerted on the robot by the physical impact exceeds at least one predefined energy threshold Es and / or at least one predefined power threshold Ls in another Cartesian degree of freedom or in several other Cartesian degrees of freedom. Adjusting the target pose makes the robot's behavior more compliant. The physical impact can thus be sensed as energy or power, depending on the robot's implementation or the robot's sensors. Compliance occurs only if a predefined energy threshold Es or Ls is exceeded.A predefined power threshold Ls is exceeded to prevent the robot from reacting undesirably due to minor physical influences, such as wind or measurement noise. If the energy threshold Es or the power threshold Ls is exceeded due to the physical influence, the robot changes its target pose Ps,t. The new target pose Ps,t is determined based on the physical influence.

[0025] When it is determined that the energy threshold Es or the power threshold Ls is exceeded in another Cartesian degree of freedom or in several other Cartesian degrees of freedom due to physical action, the magnitude of the energy or power acting on the robot can be used. Normally, energy or power is a scalar. However, within the scope of the invention, energy or power is considered as a type of vector. Thus, based on the sign of the corresponding Cartesian degree of freedom, it can be determined whether energy or power is transferred back from the environment to the robot. Energy or power transferred from the robot to the passive, articulated object has a different sign than energy or power transferred back to the robot.

[0026] This consideration of the sign for the aforementioned values ​​can be carried out both in step a) when recognizing the interaction of the robot with the passive, articulated object and in step b) when determining the physical effect of the passive, articulated object on the robot.

[0027] The determination of the new target position in step c) can be achieved by integrating a robot velocity, measured during the external power input caused by the physical action of the passive, articulated object on the robot, into the target position. In other words, the robot's control system specifies a new position when the energy or power of the physical action exceeds the predefined energy or power threshold, thus making the system formed by the robot more compliant.

[0028] Preferably, the method is executed in a loop by repeating steps b) to d). This allows the target pose to be continuously changed depending on the physical action. If step bl) is also performed during the repetition, a cessation of the physical action of the passive, articulated object on the robot can be detected, so that the process returns to step aO) and further control continues without changing the target pose until a new physical action is detected.

[0029] When repeating process steps b) to d) with step bl), the robot can easily check in step bl) whether the physical action of the passive, articulated object on the robot has ended and, if so, adjust the control based on the then-current target position Ps,t without any further changes to the target position caused by the action. For this purpose, the robot is provided with a termination condition, which, when met, indicates that the physical action on the robot has ended. The termination condition can be met if no energy or power acting on the robot is sensed in any other Cartesian degree of freedom or in several other Cartesian degrees of freedom, or if the sensed energy or power no longer exceeds the at least one predefined energy threshold Es and / or the at least one predefined power threshold Ls.In other words, if a return transfer only occurs in the Cartesian degree of freedom in which, according to step a), the robot transfers energy to the passive, articulated object, the termination condition is met, so that the robot can be controlled without further changes to the desired pose.

[0030] For example, if a robot detects and determines the physical impact of a passive, articulated object on the robot (physical impact with energy or power above the predefined energy threshold Es and / or predefined power threshold Ls), and the physical impact subsequently ceases, the robot should transition to normal operation and recognize the cessation of the physical impact. For this purpose, any sensed energy or power below the energy threshold Es or power threshold Ls is disregarded. This allows the robot's efficiency to be advantageously increased.

[0031] The method according to the invention can, for example, also provide a termination condition for the entire motion task of the robot. If, for example, it is determined that predominantly or exclusively a physical influence of the passive, articulated object on the robot occurs in the Cartesian degree of freedom in which the robot transmits energy to the passive, articulated object, and this physical influence exceeds a certain magnitude, the motion task can be terminated even before the predetermined target pose is reached. For example, if the robot's motion task is the actuation of a lever, where the lever reaches a stop and can no longer be moved, the robot then registers that predominantly or exclusively a feedback transfer of energy is occurring in the Cartesian degree of freedom. In this case, further control of the robot to the predetermined target pose is prevented by the lever or stop.To avoid damage or deformation, the movement task can be terminated.

[0032] Furthermore, the compliant control can be designed as a force control. In particular, the force control can be designed as an indirect force control, preferably as an impedance control. Alternatively, the force control can be designed as a direct force control, preferably as a combination of force and position control.

[0033] Force control is the regulation of the force with which a robot interacts with its environment. Force control can prevent damage to both the robot and its surroundings. Furthermore, force control can reduce wear and tear on the robot.

[0034] Force control, preferably impedance control, has proven particularly advantageous. Impedance control regulates the robot's compliance, i.e., the relationship between force and position during an interaction or physical impact. Impedance is the robot's ability to counteract contact forces from its environment.

[0035] Furthermore, it has proven advantageous to design the force control as a direct force control, preferably as a combination of force and position control. In this way, both the force and the position are taken into account in the control system.

[0036] It can further be provided that in step a), the robot's interaction with the passive, articulated object is carried out using robot-determined forces and a robot velocity. The forces can be determined, for example, by an impedance controller and / or a force measuring device. The velocity can be determined, for example, using derived encoder values ​​at the robot's joints. It can further be provided that in step b), the determination of the physical action on the robot is carried out using robot-determined forces and a robot velocity, whereby an energy or power value is determined for one, several, or each Cartesian degree of freedom of the robot. The forces can also be determined, for example, by an impedance controller and / or a force measuring device.The speed can be determined, for example, via derived encoder values ​​at the robot's joints.

[0037] By determining physical impacts based on values ​​obtained by the robot's control system, the need for costly sensors on the robot can be largely eliminated. Sensing physical impacts can thus be implemented simply and cost-effectively.

[0038] Particularly preferred is the provision that for another Cartesian degree of freedom of the robot, or for some or all other Cartesian degrees of freedom, an energy threshold and / or a power threshold is specified, wherein in step c) the target pose is changed if in step bl) the energy or power acting on the robot due to the physical action, with respect to the other Cartesian degree of freedom or at least one of the other Cartesian degrees of freedom, exceeds the energy threshold and / or power threshold of the corresponding other Cartesian degree of freedom.

[0039] It may also be provided that an energy threshold and / or power threshold is assigned to one other Cartesian degree of freedom or to each other Cartesian degree of freedom of the robot, wherein the energy thresholds or power thresholds are selected and assigned from predefined energy thresholds or power thresholds.

[0040] By providing different energy and / or power thresholds for different Cartesian degrees of freedom of the robot, safety during robot operation can be further increased. This allows for differentiation between limbs strongly affected by a physical impact and those only slightly affected, or between strongly and slightly affected Cartesian degrees of freedom. In other words, a force from the passive, articulated object does not typically act on a robot simultaneously in every Cartesian degree of freedom. Therefore, it is sufficient for the robot joints that determine the position relative to the affected Cartesian degree of freedom to react compliantly.

[0041] Particularly when different energy or power thresholds are defined for different Cartesian degrees of freedom of the robot, the robot's compliance can be controlled very precisely and adapted to a specific physical impact. Furthermore, it has proven particularly advantageous to specify a larger number of predefined energy or power thresholds, from which, depending on the robot's motion task, its properties (e.g., the tool currently mounted on the end effector), or the passive, articulated object, energy or power thresholds can be selected and assigned to each other Cartesian degree of freedom of the robot.

[0042] Preferably, the energy threshold(s) or the power threshold(s) are discretely and / or continuously changeable.

[0043] Particularly in environments with environmental influences, such as wind, it has proven especially advantageous to provide energy and / or power thresholds that can be changed discretely and / or continuously. For example, a wind speed sensor located on or away from the robot can measure the actual wind speed in real time, allowing the energy and power thresholds to be adjusted accordingly. This prevents robot malfunctions and increases the reliability and efficiency of the process.

[0044] It may also be provided that the compliant control can be activated and deactivated in individual or multiple joints or Cartesian degrees of freedom, whereby when the compliant control is deactivated, the robot is rigidly controlled or controlled purely via an impedance controller.

[0045] The ability to activate or deactivate the compliant control allows the robot to be advantageously adapted to the environment or the passive, articulated object in which it is used. For example, if the robot is operated in an active environment, the execution of the method can be deactivated.

[0046] The procedure may also include a correction for the influence of gravity.

[0047] The weight of the robot's limbs, caused by gravity, should not lead to a false detection of physical impact. Therefore, correcting for the effects of gravity is advantageous when the robot is used in a gravitational environment.

[0048] The robot according to the invention has a control device, wherein the control device is configured to carry out a method according to the present invention.

[0049] Preferably, the robot is a teleoperated robot. Alternatively, the robot is an autonomous robot.

[0050] The inventive method is explained in more detail below with reference to figures.

[0051] Figures aa and bb show a schematic representation of a robot according to the invention in interaction with a passive, articulated object, and

[0052] Figure 2 shows a flowchart of the method according to the invention.

[0053] Figures 1a and 1b each depict a robot 10 with robot joints 11a, 11b and robot limbs 12a, 12b and an end effector 13. The robot 10 is located in a passive environment and uses the end effector 13 to grasp a passive, articulated object 14 in order to operate it. The passive, articulated object 14 is, for example, a lever that is mounted so that its upper end can move along a predefined path 14a. The robot 10 has the single task of moving the lever to its end position 14'.

[0054] Robot 10 has an actual position Pt at time t, which is determined by the position and orientation of robot 10. In other words, the actual position Pt includes the position of the end effector 13 as well as the position of the individual robot joints 11a, 11b and robot limbs 12a, 12b. Furthermore, robot 10 is given a target position Ps,t, 20. The target position 20 is the position to which robot 10 should move from its actual position Pt and includes, among other things, a change in the position of the end effector 13 in the x-direction by Ax, assuming a two-dimensional coordinate system, as schematically illustrated in Figures 1a and 1b. To reach the target position Ps,t, 20, the end effector is moved in the corresponding directions. The target position Ps,t, 20 is intended to allow robot 10 to actuate the lever.

[0055] Figure 1a depicts a state in which an interaction between the end effector 13 and the passive, articulated object 14 is just beginning. The robot 10 senses the interaction, a force, and the speed of the interaction using an impedance controller (not shown). The robot 10 exerts a force on the passive, articulated object 14 in one degree of freedom (x-direction) and transfers energy (large white arrow). Due to the elasticity of the robot 10, a kind of feedback occurs from the passive, articulated object 14 to the robot 10, in which a (small) amount of energy is transferred back to the robot 10 in that degree of freedom (black arrow). The transferred energy can be determined by its sign. This allows the robot 10 to detect the interaction with the passive, articulated object 14.

[0056] Since the passive, articulated object 14 can only move along the predetermined path of motion 14a, however, changing the position of the end effector 13 in the x-direction by Ax does not result in a trajectory of motion that can move the passive, articulated object 14 along the predetermined path of motion 14a.

[0057] Due to the predetermined trajectory, the passive, articulated object 14 can exert a physical influence on the robot, pushing it away from the desired pose Ps,t, 20, in the z-direction in the example shown. A portion of the energy transferred to the passive, articulated object 14 in one degree of freedom is thus converted and transferred back to the robot in another degree of freedom (z-direction) (black arrow in the z-direction in Fig. 1b), thereby causing a physical influence on the robot 10 in that other degree of freedom.

[0058] With respect to this physical action, the robot 10 is to react compliantly. Therefore, based on the physical action, the predetermined target position Ps,t 20 is changed to a new target position 20'. As a result, the actual position Pt of the robot follows the new target position 20' of the robot 10. The change in the target position means that the change in the position of the end effector 13 required to reach the target position 20' is now in the z-direction Az'. This changes the force required by the robot to change the position of the end effector 13 and thus the compliance of the robot. The change in the target position is dependent on the physical action.For the robustness of the control system, it is advantageous if the energy and / or power acting on the robot 10 through the physical action is compared with a predefined energy threshold Es and / or predefined power threshold Ls, and the target pose is only changed if the threshold(s) are exceeded.

[0059] This allows the robot to "develop" a movement trajectory that follows the predetermined movement path of the passive, articulated object 14.

[0060] The flowchart in Figure 2 shows the compliant control system according to the invention. The robot 10 is initially controlled rigidly, in which the actual position Pt follows a predetermined target position Ps,t at time t (step S101). If an interaction with the passive, articulated object 14 is detected (step S102), the energy or power acting on the robot in the other degrees of freedom (not the x-direction in Figures 1a, 1b) is checked, and it is verified whether the energy or power of the detected physical action is greater than a predefined energy threshold Es or greater than a predefined power threshold Ls, respectively (step S103). If no interaction is detected in step S102, or if no physical action is detected in other degrees of freedom in step S103, or if it is determined that the energy or power of the detected physical action is less than or equal to a predefined energy threshold Es or Ls, respectively, the robot is then controlled rigidly.If a predefined power threshold Ls is exceeded, the procedure according to the embodiment shown in Figure 2 continues with step S101, and rigid control based on the original target position Ps,t continues. This control continues as long as no or only a minor physical influence (less than the predefined energy threshold Es or a predefined power threshold Ls) is detected.

[0061] If a physical impact is detected and the energy or power of the detected physical impact exceeds the predefined energy threshold or the predefined power threshold Ls, the robot 10 should comply in the corresponding degree of freedom, changing the target position to a new target position Ps,t 20' (step S104). The change in the target position depends on the physical impact. Subsequently, the robot is controlled based on the changed target position Ps,t (step S105).

[0062] Detecting and determining the physical impact can generally be done using an impedance controller, which measures the robot's force and speed. Of course, other options are also conceivable, such as additional sensors or current measurement in the robot's drives.

[0063] robot

[0064] a robot joints

[0065] b Robot joints

[0066] a robot limb

[0067] b robot limbs

[0068] End effector

[0069] articulated object

[0070] a trajectory

[0071] ' Final position

[0072] Desired position

[0073] 'new target pose

Claims

PATENT CLAIMS 1. Method for the compliant control of a robot (10) with at least one robot joint (11a, 11b) for use in a passive environment with a passive, articulated object that can be moved in a predetermined path of motion, wherein the robot (10) assumes an actual position Pt by following a predetermined desired position Ps,t, comprising the steps: aO) Specifying the desired pose Ps,t of the robot, wherein the desired pose Ps,t causes a movement of the robot (10) with an interaction with the passive, articulated object of the environment, a) Detecting the interaction of the robot (10) with the passive, articulated object, wherein during the interaction the robot (10) transfers energy to the passive, articulated object in one Cartesian degree of freedom, b) Determining a physical effect of the passive, articulated object on the robot (10) due to the energy transferred by the passive, articulated object through the specified path of motion into another Cartesian degree of freedom or into several other Cartesian degrees of freedom, which is transferred to the robot (10), c) Changing the specified target position depending on the specific physical influence, d) Rules of the robot (10) by specifying the modified target pose as target pose Ps,t.

2. Method according to claim 1, characterized by step bl) after step b):bl) checking the energy and / or power acting on the robot (10) by the physical action for one or more other Cartesian degrees of freedom, wherein if the energy and / or power acting on the robot (10) through the physical action exceeds at least one predefined energy threshold Es and / or at least one predefined power threshold Ls in another Cartesian degree of freedom or in several other Cartesian degrees of freedom, steps c) and d) are carried out, and if the energy and / or power acting on the robot (10) through the physical action does not exceed at least a predefined energy threshold Es and / or at least a predefined power threshold Ls or no physical interaction is detected, the process returns to step aO).

3. Method according to claim 1 or 2, characterized by a repetition of steps b) to d).

4. Method according to one of claims 1 to 3, characterized in that the compliant control is designed as a force control.

5. Method according to claim 4, characterized in that the force control is designed as an indirect force control, preferably as an impedance control.

6. Method according to claim 4, characterized in that the force control is designed as a direct force control, preferably as a combination of force and position control.

7. A method according to any one of claims 1 to 6, characterized in that in step a) the interaction of the robot with the passive, articulated object is detected via forces determined by the robot and a velocity of the robot (10).

8. A method according to any one of claims 1 to 7, characterized in that in step b) the physical action on the robot (10) is determined via forces determined by the robot and a velocity of the robot (10), wherein an energy or power value is determined for one, several, or each Cartesian degree of freedom of the robot.

9. Method according to claim 7 or 8, characterized in that the forces determined by the robot are determined by an impedance controller and / or by a force measuring device.

10. Method according to one of claims 2 to 8, characterized in that the energy threshold(s) or the power threshold(s) are discretely and / or continuously changeable.

11. Robot (10) with a control device, wherein the control device is configured to perform a method according to any one of claims 1 to 10.

12. Robot according to claim 11, characterized in that the robot is a teleoperated robot.

13. Robot according to claim 11, characterized in that the robot is an autonomous robot.

14. Robot according to one of claims 11 to 13, wherein the compliant control is activatable and deactivatable.

Images