Robot system and method for controlling a robot system

Abstract

The invention relates to a robot system comprising at least one robot arm and a control unit, wherein the control unit is configured to - to operate the robot system in a first operating mode, a second operating mode and a third operating mode and to switch between the individual operating modes; - to determine an autonomy parameter that defines a gradually adjustable degree of autonomy for controlling the robot system within the second operating mode, - to control the robot system in the first operating mode depending on user input; - to control the robot system in the second operating mode depending on user input, a predefined movement sequence, and autonomy parameters; and - to control the robot system in the third operating mode depending on a predefined movement sequence; whereby - the control unit is set up to adjust the degree of autonomy of the control defined by the autonomy parameter in the second operating mode using a virtual bracket and by varying the parameters of the virtual bracket.

Description

[0001] The present invention relates to a robot system and a corresponding method for controlling a robot system. In particular, the present invention relates to a robot system that can be used in the field of telemedicine, enabling, for example, a surgical procedure by a surgeon located at a remote location. However, the present invention is not limited to such robot systems for medical use, but also encompasses, for example, robot systems used in industry for the production of goods.

[0002] In the robot systems known from the state of the art, it can be provided that the control of the robot system takes place in different levels of autonomy.

[0003] In a first level of autonomy (also referred to as the Level of Autonomy or LoA for short), the robot system or a defined point of the robot system (also referred to as the Interest Point or IP for short) is manually controlled by the user. An input device can be used for this purpose, which in telemedicine applications is often designed as a haptic input device. The input device can have various configurations. For example, joysticks can also be used to receive a control command from a user and generate a corresponding control signal to control the robot system.The input by a user can also be realized by the user moving the robot system and varying its orientation, whereby corresponding sensors (in particular acceleration sensors provided in the robot system) can detect a change in the position or orientation of the robot and generate a control signal depending on the detected change.

[0004] In a second level of autonomy, the robot system is controlled based on user input and a predefined motion sequence. This level of autonomy is often referred to as shared control. A motion sequence can be predefined according to which the robot system is ideally controlled. For example, the motion sequence can define a trajectory that describes a surgical incision or a suture to be performed during a predefined surgical procedure. Ideally, a scalpel held by the robot system would be guided along this trajectory to perform the predefined incision. However, in this level of autonomy, the user is given full control over the movement of the robot system. However, they are supported by the robot system in at least one degree of freedom.To implement control in the second level of autonomy, virtual fixtures (VFs) are often used to realize haptic augmentation in one or more degrees of freedom. These virtual fixtures typically have dead zones so that the user can compensate for any inaccuracies in the robot system. In the prior art methods, the virtual fixtures and their dead zones are designed to be constant and adapted to a specific application.

[0005] In a third level of autonomy, the robot system can be controlled autonomously according to a predefined motion sequence. In this third level of autonomy, the user has no influence on the predefined trajectory and / or the rotational movements defined by the predefined motion sequence.

[0006] While the robot systems known from the state of the art are already very promising in many applications, in practice it is desirable to provide an improved robot system that, on the one hand, enables an optimized transition between the autonomy levels and, on the other hand, allows improved control.

[0007] To achieve the above-mentioned object, the present invention proposes a robot system comprising at least one robot arm and a control unit, wherein the control unit is configured to: - to operate the robot system in a first operating mode, a second operating mode and a third operating mode and to switch between the individual operating modes; - to determine an autonomy parameter that defines a gradually adjustable degree of autonomy for the control of the robot system within the second operating mode; - to control the robot system in the first operating mode in response to a user input; - to control the robot system in the second operating mode depending on a user input, a specified movement sequence and the autonomy parameter; and - to control the robot system in the third operating mode depending on a predetermined movement sequence; where - the control unit is configured to set the degree of autonomy of the control defined by the autonomy parameter in the second operating mode using a virtual mount and by varying the parameters of the virtual mount.

[0008] The first operating mode describes a mode in which the robot system is controlled manually. The first operating mode can therefore also be referred to as a manual operating mode. The second operating mode describes a mode in which the robot system is controlled by shared control. This means that both user input and a predefined motion sequence are taken into account.

[0009] The user cannot therefore control the robot system completely manually and free from any restrictions. However, the control commands entered by the user are influenced by a guide. For example, it can be provided that the user's control commands are entered via a haptic input device, whereby the virtual fixture is used to guide the user's input or to prevent movement in certain areas. The use of the virtual fixture therefore restricts the user's free or manual control (at least in one degree of freedom). The degree of restriction of free control in the second operating mode or the degree of guidance in this operating mode by the system is defined in the method according to the invention by the autonomy parameter and adjusted by varying the parameters of the virtual fixture.The robot system according to the invention thus allows a gradual adjustment of the degree of autonomy within the second operating mode, thus enabling an improved adjustment of the degree of autonomy. In the third operating mode, the robot system is largely controlled autonomously. The robot system is essentially controlled depending on a predetermined motion sequence, wherein the motion sequence defines a translational movement and / or a rotational movement of a component of the robot system for a predefined process (e.g., for a surgical incision or for performing a desired suture).The third operating mode can therefore also be referred to as an autonomous operating mode, whereby the term "autonomy" refers to the fact that a movement occurs according to a predetermined movement sequence without the need for input from the user and without the user being able to significantly change a predetermined movement trajectory or a rotation sequence. As will be explained below, the user can preferably also make inputs in the third operating mode in order to, for example, manipulate the speed of the movement, while the movement trajectory remains unchanged. In some embodiments of the invention, it can be provided that the control unit is configured to control the robot system in the third operating mode exclusively as a function of a predetermined movement sequence.Alternatively, it may be provided that, in addition to the specified movement sequence, a user input is taken into account so that the user can make at least minor corrections to a process to be carried out and / or can adjust the speed of the movement.

[0010] The robot system according to the invention offers the possibility of switching between different operating modes and thus reacting to varying environmental conditions. The robot system according to the invention also offers the user the possibility of intervening in the control system at any time and adjusting the degree of autonomy as needed. For example, the degree of autonomy can be reduced if the control of the robot system proves to be unreliable, while the degree of autonomy can be increased if the control of the robot system is considered reliable. Furthermore, in the second operating mode, the autonomy parameter can be used to gradually adjust the degree of autonomy for the control of the robot system. This allows for optimized control of the robot system, enabling particularly fine adjustment of the degree of autonomy.

[0011] For example, the autonomy parameter can assume a value between 0 and 1, with a low value of the autonomy parameter close to 0 indicating largely manual control of the robot system, and a high value of the autonomy parameter close to 1 indicating largely autonomous control of the robot system. For example, the autonomy parameter can assume five, ten, 20, 30, 50, or 100 different values ​​between 0 and 1. This allows for a gradual adjustment of the degree of autonomy that can be adapted to a specific application scenario.

[0012] In the robot system according to the invention, it can be provided, in particular, that the robot system has an input device, which is preferably designed as a haptic input device. Alternatively, it can be provided that the input device is designed as a joystick.

[0013] The predefined motion sequence can be provided, for example, in the form of a function or in the form of a vector that describes the position of a point or the rotation of a component. If the robot system is in the third operating mode, the reference point to be controlled or the component of the robot system to be controlled can, for example, be moved autonomously along a predefined trajectory. The predefined motion sequence can also include rotation information. This allows the robot system to be controlled autonomously in all six degrees of freedom. In the third operating mode, in some embodiments, the user cannot make any changes to the predefined trajectory or the predefined rotation information. Alternatively, it can be provided that smaller corrections are also permitted in the third operating mode, in which case the dead zones are very small.If the robot system is in the second operating mode, the robot system can be controlled depending on the specified movement sequence, the user input, and the autonomy parameter, whereby the user can correct the movement specified by the robot system through their input. The degree of influence the user has on the control is defined by the autonomy parameter. As already explained above, the user can intervene relatively strongly in the movement sequence of the robot system if the autonomy parameter has a low value, while they can only intervene relatively slightly in the movement sequence of the robot system if the autonomy parameter has a high value. This makes it possible to gradually adjust the degree of autonomy within the second operating mode.According to the present invention, a virtual mount is used to adjust the degree of autonomy, the parameters of which can be varied in real time.

[0014] According to some embodiments of the robot system according to the invention, it can be provided that the autonomy parameter can be adjusted by user input. This allows the user to manually adjust the degree of autonomy. If, for example, the user determines that the control of the robot system is unreliable, they can reduce the degree of autonomy in order to be able to intervene more strongly in the movement sequence of the robot system. Alternatively, it can be provided that the degree of autonomy is adjusted automatically, which is explained in more detail below. For this purpose, context information can be evaluated that allows a statement to be made about how reliable the control of the robot system is at a specific point in time.

[0015] In the robot system according to the invention, it can be provided that the virtual mount is defined by a geometric body and by geometric parameters that describe the geometric body, wherein a dead zone is defined in the interior of the geometric body, within which user inputs for controlling the robot system are implemented without restriction, while user inputs that result in control outside the dead zone are restricted by the virtual mount. The virtual mount can preferably be defined by a cylinder, an ellipsoid, or a cuboid. The use of a virtual mount defined by a geometric body offers the advantage that the virtual mount and the degree of autonomy set with the virtual mount can be easily adapted to the current application scenario.In particular, the virtual mount can be defined by a cylinder, whereby the parameters of the virtual mount can be adjusted accordingly. If the configured cylinder defining the virtual mount has a relatively small cylinder radius, then the user's movement input is correspondingly severely restricted. It can be provided that the user uses a tactile input device, and that the user inputs are restricted by a feedback force if the user input causes a departure from the specified cylinder. This enables guided user input.Depending on the current application scenario, the cylinder radius can be adjusted so that it is set to a relatively small value in order to enable largely autonomous control of the robot system, while the cylinder radius is set to a comparatively large value in order to enable largely manual control of the robot system. In this way, the robot system according to the invention allows a gradual adjustment of the degree of autonomy within the second operating mode, whereby the degree of autonomy can be precisely adapted to different application scenarios. Analogous to the embodiment described above, it can also be provided that the virtual holder is defined by a cuboid or an ellipsoid. The corresponding parameters of the cuboid or the ellipsoid can also be varied in this case in order to adjust the degree of autonomy of the controller in the second operating mode.

[0016] In the robot system according to the invention, it can also be provided that the control unit is configured to switch the operating mode and / or the degree of autonomy manually and depending on a user input. For example, an input signal generated by an input device can be evaluated, wherein the control unit can be configured to switch between the first operating mode, the second operating mode, and the third operating mode depending on a received input signal. Likewise, the control device can be configured to increase or decrease the degree of autonomy in the second operating mode depending on an input signal. For example, it can be provided that ten different degrees of autonomy are provided within the second operating mode, wherein the user can switch up and down between the individual degrees of autonomy via a user input.This allows the user to adjust their influence on the control of the robot system or the level of autonomy of the controller according to the current situation. For example, it can be provided that the user monitors a task performed by the robot system and, in cases where faulty or at least inaccurate operation is observed, can reduce the level of autonomy or completely switch to the first operating mode. Likewise, in cases where reliable operation is observed, the level of autonomy can be increased or the controller can switch to the third operating mode (autonomous operating mode). In this way, the user is always in control of how autonomously the robot system should operate and how much the user can intervene in the control process.

[0017] According to some preferred embodiments of the robot system according to the invention, it can be provided that a selection of operating modes is suggested to the user by the robot system, which are assessed as useful in an application. It can also be provided that individual operating modes and / or levels of autonomy are not permitted at all if they appear unsuitable taking into account the current situation. The selection of the suggested or permitted operating modes and / or levels of autonomy can be made in particular depending on context information, which is described in the following section in connection with the automatic determination of the operating mode and / or the level of autonomy. In particular, the selection of the suggested or provided operating modes and / or levels of autonomy can be made depending on an uncertainty parameter U, which is also explained in detail below.This enables the robot system to provide data-based support to the user during the manual selection of the operating mode and / or the level of autonomy. In particular, it can be provided that certain operating modes or certain levels of autonomy are not permitted in cases of high uncertainty.

[0018] Furthermore, in the robot system according to the invention, it can be provided that the control unit is configured to determine the operating mode and / or the degree of autonomy automatically and depending on previously determined context information. The context information comprises data that characterizes the current operation of the robot system. This data can, for example, relate to individual components of the robot system or to its environment, thus allowing a characterization of the robot system or environment. For example, the context information can include measurement data that represent the position of a component of the robot system or a defined point within the robot system or the rotation angle of a component of the robot system. For example, the robot system can comprise a sensor unit for this purpose that generates corresponding context information.The sensor unit can be designed, for example, as an optical, electrical, or magnetic sensor unit. For example, the sensor unit can have a camera or a lidar sensor, which is used, for example, to determine the position of a reference point of the robot system or a rotation angle of a robot arm. For example, the tracking approaches known from the prior art can be used, wherein known position determination methods can be used to determine the position of the reference point. In a robot system used for surgical applications, it can be provided, for example, that the position and / or orientation of a scalpel is monitored and that the context information contains corresponding measurement data on the position and / or orientation of the scalpel. The context information can also include data describing the environment of the robot system.The context information can, for example, contain data on the current temperature or the current ambient light in the environment of the robot system. The context information can also contain data on whether an object to be processed by the robot system is in the correct position and has the correct orientation. In situations in which the autonomous control of the robot system is expected to fail (for example, due to a significant deviation of the current temperature from the normally intended temperature or if an object to be processed is not correctly secured in a holding device), the control unit can automatically change the operating mode and, for example, switch from the third operating mode (autonomous control) to the second operating mode (split control), or from the second operating mode to the first operating mode (manual control).

[0019] In the embodiment of the robot system according to the invention described above, it can therefore be provided that the control unit determines the operating mode and / or the degree of autonomy fully automatically depending on the determined context information. This can ensure particularly reliable and safe control of the robot system. According to one embodiment, it can also be provided that the current lighting conditions are measured (for example by measuring the intensity of the ambient light using a photodiode) and that the operating mode and / or the degree of autonomy is then selected depending on the measured lighting conditions. For example, it can be provided that in very poor lighting conditions orIn complete darkness, switching from the first operating mode to the second operating mode or from the second operating mode to the third operating mode increases the degree of autonomy of the controller. This allows for assistance from the robot system in cases where the user's visibility is limited and user input can be considered relatively unreliable. This allows for fully automatic adjustment of the operating mode and / or the degree of autonomy of the controller depending on the current context information.

[0020] Furthermore, in the robot system according to the invention, it can be provided that the control unit is configured to determine an uncertainty parameter U for automatically determining or selecting the operating mode and / or the degree of autonomy, wherein the uncertainty parameter U reflects the unreliability of measurement data recorded within the robot system. The uncertainty parameter U can therefore be calculated from the context information. As already explained above, the robot system can contain a sensor unit for recording measurement data that characterizes the current status of the robot system and its environment. The uncertainty parameter U can be determined from the recorded measurement data. The uncertainty parameter U can, in principle, change in each time step in which the associated sensor unit records measurement data.In particular, the uncertainty parameter U can characterize the uncertainty of an estimated position, an orientation, or a pose (position + orientation) of a component of the robot system. For example, the position of a reference point of the robot system can be determined by a camera system, a lidar system, or a magnetic position sensor, and the uncertainty parameter U can then be determined from the measurement data recorded within a measurement period. Various approaches can be chosen for this. In particular, the reliability of the measurement data can be verified by a plausibility check. If the measurement data show relatively strong outliers, this can be interpreted as an indication that the measurement data is rather unreliable.The measured data can also be compared with data typically expected in a specific application scenario (a reference point, for example, can be designed to be moved within a certain value range), whereby the discrepancy between the recorded measured data and the expected data can be used as an indication of the unreliability or uncertainty of the robot system. Consequently, the uncertainty parameter U can be set to a relatively high value. The strength of any outliers can be assessed, for example, by calculating the variance of the measured data. If many outliers are detected, or fewer outliers that deviate significantly from the expected values, the uncertainty parameter can assume a comparatively high value.If a high level of uncertainty (defined by a high value of the uncertainty parameter U) is detected in the robot system, the degree of autonomy can be reduced, giving the user greater influence over the control process. This can be achieved by varying the autonomy parameter. Alternatively, it is possible to switch from the third operating mode (autonomous mode) to the second operating mode (split control) or from the second operating mode to the first operating mode (manual mode). If the value of the uncertainty parameter U drops again after a certain period of time, the degree of autonomy can be increased accordingly by adjusting the autonomy parameter, ensuring largely autonomous control of the robot system without increasing the risk of faulty control.Alternatively, switching can be performed from the first operating mode to the second operating mode or from the second operating mode to the third operating mode. This allows for a fully automatic adjustment of the degree of autonomy and / or the operating mode, depending on the uncertainty parameter U, that is optimized for the current application scenario.

[0021] According to a preferred embodiment of the robot system according to the invention, it can further be provided that the control device is configured to determine the operating mode as a function of the determined uncertainty parameter U, wherein the control device is designed to: - to select the first operating mode if the uncertainty parameter U exceeds a first threshold U1; - to select the second operating mode, provided that the uncertainty parameter U lies between the first threshold U1 and a second threshold U2, where U1 is greater than U2; and - to select the third operating mode if the determined uncertainty parameter U falls below the second threshold value U2.

[0022] This enables automatic switching between operating modes based on the reliability of the measured data. This ensures that an optimal operating mode and thus an optimal level of autonomy (LoA) is always provided in every situation.

[0023] According to a preferred embodiment of the robot system according to the invention, it can be provided that the virtual holder is defined by a cylinder having a cylinder radius r d where the cylinder radius r din the second operating mode as a function dependent on a previously determined uncertainty parameter U rd=f(U) The function f(U) can be implemented as a linear function, an exponential function, or a polynomial function of nth order. The function f(U) can describe the behavior of the cylinder radius r d can be adjusted depending on the uncertainty parameter. According to one embodiment, the function f(U) can be implemented as a linear function, so that the cylinder radius r d according to rd=kr U can be defined. Where k denotes ra weighting factor. This allows for simple and efficient adjustment of the virtual mount in direct dependence on the determined uncertainty parameter U, ensuring automatic adjustment of the degree of autonomy within the second operating mode depending on the recorded measurement data.

[0024] Furthermore, the robot system according to the invention can be configured so that the control unit is configured to switch from the third operating mode to the second operating mode if the control unit detects a force exerted by the user via an input device that exceeds a predetermined force threshold. This enables the user to intervene in the control of the robot system even in autonomous operating mode (= third operating mode), particularly if unreliable operation is detected and the current control of the robot system does not meet the user's requirements.This mechanism, which allows user intervention even in autonomous operating mode, is particularly advantageous in surgical applications, as the user of the robotic system (usually a surgeon) has the possibility to intervene in the surgical process at any time and thus take over control (at least partially) in case of complications.

[0025] Furthermore, in the robot system according to the invention, it can be provided that the control unit is configured to control the robot system in a third operating mode such that a component of the robot system is moved according to a predetermined movement sequence, wherein the speed of the movement can be adjusted depending on a user input. While in the methods known from the prior art, user inputs are completely ignored in an autonomous operating mode, in this embodiment of the invention the user is given the expanded option of influencing the speed of the process performed by the robot system, while the movement trajectory and / or a rotation angle remain unchanged compared to the predetermined movement sequence.This ensures, on the one hand, that the specified movement sequence is maintained and, on the other hand, provides the user with the option of slowing down the process, for example, when a certain uncertainty in the control of the robot system is expected and of accelerating the process when the control of the robot system is considered reliable.

[0026] According to a preferred embodiment of the robot system according to the invention, it can also be provided that the robot system has an input device designed to detect a user input and, in response to the user input, to generate an input signal corresponding to the input and to transmit the input signal to the control unit. The input device can, in particular, be designed as a haptic input device. Such input devices enable interaction between a user and the robot system through touch sensors and tactile feedback. Haptic input devices make it possible to provide the user with tactile feedback, whereby the user can, for example, be guided according to a predetermined movement sequence and / or prevented from generating certain control signals.In this case, the Virtual Holder can be used to prevent the user from performing undesired movements and / or to promote a desired movement sequence by guiding the user's input according to a predetermined movement sequence and a permissible tolerance, which can be set by the parameters of the Virtual Holder.

[0027] According to some embodiments of the robot system according to the invention, control can be provided without a separate input device. For example, the control signals generated by a user input can be generated by force and / or torque sensors mounted on or in the robot system. This allows the user to generate the desired control signals by changing the position or inclination of the robot system. This can be particularly preferred for compact and portable robot systems.

[0028] In addition, to achieve the object described above, the present invention proposes a method for controlling a robot system comprising at least one robot arm and a control unit, the method comprising the following steps: - controlling the robot system in a first operating mode, a second operating mode and a third operating mode, wherein switching between the operating modes is possible; - Determining an autonomy parameter that is gradually adjustable and defines the degree of autonomy of the control of the robot system during an operating mode; - controlling the robot system in the first operating mode in response to a user input, provided the first operating mode is selected; - controlling the robot system in the second operating mode depending on a user input, a predetermined movement sequence and the autonomy parameter, if the second operating mode is selected; - controlling the robot system in the third operating mode depending on a predetermined movement sequence, provided that the third operating mode is selected; - wherein the degree of autonomy of the control of the robot system in the second operating mode is adjusted using a virtual holder and by varying the parameters of the virtual holder.

[0029] The method according to the invention can implement the same features that were described above in connection with the robot system according to the invention.

[0030] In particular, the method according to the invention can provide for the control unit to generate a control signal that causes switching between the operating modes. The control signal can be generated in response to a user input or automatically. If the control signal is generated automatically, it can provide for context information to be acquired and for the control signal to be generated depending on the context information.

[0031] In addition, the method according to the invention can preferably comprise the following method steps: - Generating context information containing data concerning the robot system or its environment; - Determining an uncertainty parameter U depending on the generated context information, wherein the uncertainty parameter U represents the unreliability of measurement data recorded by the robot system; and - Variation of the parameters of the virtual support depending on the determined uncertainty parameter U.

[0032] The present invention will be explained in more detail below with reference to further preferred embodiments, wherein the invention will be explained using application examples relating to a surgical robot system for illustrative purposes. However, it will be apparent to those skilled in the art that the technical features described below can also be used in other robot systems. Foreword to the preferred embodiments of the invention

[0033] Today, telemanipulation robot systems can be found in many operating rooms. The use of autonomy for minimally invasive robotic surgery (MIRS) remains an open field of research, as it is not trivial to provide meaningful control support while ensuring safe transitions between control by the robot system and the surgeon. The present invention presents a novel virtual fixture (VF) that preferably provides haptic support for shared control and task-level autonomy. Furthermore, in the present invention, uncertainty measurements from estimators for the robot system and its environment can influence the current autonomy level.The proposed virtual fixture ensures safe transitions between the robotic system and the surgeon. The proposed method is evaluated through experiments demonstrating the successful completion of a surgeon training task while exploiting shared control and task-level autonomy. I. Introduction

[0034] Minimally invasive robot-assisted surgery (MIRS) is widespread and has already become established in many operating rooms. It offers patients similar benefits to minimally invasive surgery: shorter recovery times and minimal scarring due to small incisions. At the same time, MIRS offers surgeons more ergonomics and improved hand-eye coordination. Commercially available surgical robots for MIRS are often designed as telemanipulation systems. They act as the surgeon's hands and eyes within the patient's body. This means that the surgeon directly controls the movement of the telemanipulation system in all available, usually six, degrees of freedom (DoF).

[0035] Recent scientific approaches aim to increase the autonomy of telemanipulation systems. Levels of autonomy (LoA) classify both the creation and selection of an operation plan, as well as the execution and monitoring of this plan. They range from low autonomy (L0) through robot assistance (L1), task-related autonomy (L2), supervised autonomy (L3), and high-level autonomy (L4) to full autonomy (L5) (see [1], [2], [3]). This present invention focuses in particular on the following LoAs: L0: The surgeon can freely telemanipulate the instrument tip within its inherent working area. L1: The robot system supports the surgeon during shared control through haptic augmentation. L2: The robot system controls all degrees of freedom to perform a previously defined task.

[0036] For this purpose, a novel Feed Forward Virtual Fixture (ffVF) is presented here. It implements a wrench that acts on the pose and deadband of an attractive virtual fixture (attractive VF or aVF). A deadband allows the user to compensate for inaccuracies in the robot system. The ffVF acts on the deadband parameters to scale them incrementally.

[0037] When the ffVF acts on the pose (position + orientation), interactive task-level autonomy can be achieved, allowing the user to fine-tune the resulting movement through interaction with an input device. Similarly, the ffVF is compatible with other virtual fixtures (VFs), particularly repulsive VFs (rVFs).

[0038] Surgery creates dynamic environments with unpredictable circumstances. Therefore, continuous and rapid transitions of control are required, in particular from control by the robot system (L2 or L1) to control by the surgeon (L0) at any given time (see [4], [5]). The present invention allows continuous transitions between all of the aforementioned LoAs, which can optionally be implemented by introducing an Autonomy and Shared Control Parameterization Engine (ASCOPE). ASCOPE can determine the appropriate LoAs and adjust the parameters of the ffVF accordingly. This can preferably be based on uncertainty measurements, particularly preferably based on context information of the robot system and its environment, as well as the current state of a task.

[0039] Some particularly preferred aspects of the present invention are listed below: • ASCOPE for the continuous adaptation of the LoAs depending on context information about the robot system and its environment, coupled with an uncertainty parameter • determining the parameterization of the novel ffVF based on the LoA and the current task state • ffVF based on a feed-forward wrench that enables interactive autonomy at the task level and smooth transitions between all relevant LoAs.

[0040] After a discussion of the state of the art compared to the contribution of the present invention in Section II, the following Section III provides detailed insights into the proposed methods, which are then evaluated in Section IV. Finally, Section V discusses the proposed methods. II. State of the art

[0041] This section provides an overview of related work. The first part focuses on L2 and L1, while the second part discusses the transitions between different LoAs.

[0042] Various approaches address L2. Hwang et al., Rosen et al., and Rahman et al. introduce task-level autonomy for a surgical pin transfer and compare the results of autonomous execution with human performance in a telemanipulation environment (see [7], [8], and [9]). Approaches for autonomous suture placement are presented in

[10] ,

[11] , and

[12] . Kam et al. present a method for suturing experiments on synthetic vaginal cuff fabrics (see

[13] ). The sutures are placed autonomously. Saedi et al. perform an autonomous intestinal anastomosis using the STAR system (see

[14] ). These works introduce systems for task-level autonomy (L2) without focusing on transitions to L1 or L0. User interactions with a haptic input device are not possible during autonomous operation at the L2 autonomy level.

[0043] In L1, the robot system provides the user with support in performing the task. In telemanipulation systems, one type of support is the automatic scaling of user input compared to the movement of the instrument tip (see

[15] ). Another type of support is VFs that enable haptic augmentation (see

[16] ,

[17] , and

[18] ). Attractive VFs haptically guide the user to a target structure, as described in

[19] . Repulsive VFs protect the user from entering a forbidden zone. Geometric primitives, such as meshes (see

[20] ), and visual servo information (see

[21] ) can determine the shape of a VF. In

[22] , uncertainties in visual pose recognition are used to parameterize a visual servo VF.In the method described in

[19] , the user is able to generate attractive VFs and adjust their geometry and position through a human-in-the-loop approach. These works provide support in L1 without the possibility of switching to L2.

[0044] Fontanelli et al. designed a state machine for needle insertion and removal (see

[23] ). The transitions are based on predefined gestures that the user performs to iterate through the different states. The control mode, or operating mode L0, L1, or L2, is preselected and does not change during the execution of a task. Some works discuss transitions between different LoAs. In

[24] , a trust-based supervised autonomous control strategy shifts control between the user during telemanipulation and an autonomy module that uses a shared spatial interpolator. The authors note that a sudden transition from autonomy to telemanipulation can lead to an increased error rate for the operator due to the user's lack of situation awareness. Ferraguti et al.ensure a fast and stable transition from L2 to L0 by compensating the pose offset between leader and follower (see

[25] ). Shamaei et al. introduce a dominance factor that enables the continuous change of authority between any autonomous agent and telemanipulation by a user. Additionally, an aggressiveness factor determines the allowed pace of the autonomous agent (see

[26] ). In

[27] , the robot system autonomously distributes authority between the human operator and itself for shared control. In

[28] , the user authorizes the transition from L1 to L2 by keypress, but not vice versa. These works ensure the transition either between L2 and L0 or between L1 and L2.

[0045] Although the state-of-the-art methods offer some promising approaches, they still do not offer a safe and convenient way to switch between the individual operating modes. Furthermore, the state-of-the-art methods do not allow for gradual adjustment of the degree of autonomy during shared control. III. Methods

[0046] The methods explained in this section, which describe preferred embodiments of the invention, are illustrated using the example of a so-called "Picka-Ring task" with a surgical instrument. The determination of the appropriate LoA is based on estimators for the robot system and its environment, as described in Section III-A and in Fig. 2a. Section III-B describes the details of the Autonomy and Shared Control Parameterization Engine (ASCOPE) used in Fig. 2b. The novel Feed Forward VF (ffVF) is presented in Section III-C and Fig. 2c graphically shown. As in the Fig. 2, estimators (as in [6]) provide context information C. When the context information C is paired with a corresponding uncertainty estimate U, ASCOPE can determine the current state of the task and the corresponding LoA and calculate the corresponding parameters P and the LoA-specific parameters P loa to parameterize VF. The pick task serves as an example of any task that can be performed by the robot system. A. Selecting the appropriate LoA

[0047] To select the appropriate LoA for a given task, context information about the environment can be used. It can be provided that each estimator i context information C ithat describe the current state of the robot and its environment. C i can be estimated with an uncertainty estimate U i For the autonomous execution of the pick task (predefined task for grasping an object), a high accuracy of the device pose (also called device posture) x with respect to the ring is an important prerequisite and is therefore taken as an example for an environment estimator. Accordingly, U i the uncertainty of the in C i contained estimated pose. From this, the combined uncertainty U U=maxi(kiUi) with U,Ui,ki∈[0,1]∩ℝ, which represents the maximum uncertainty of all estimators with k i , a predefined weighting factor. The appropriate LoA is determined by LoA={L2,if U≤Um2L1,if Um2< ​U≤Um1L0,if U>Um1 where U m2 the maximum uncertainty for L2 and U m1the maximum uncertainty for L1 is defined.

[0048] For example, if U < U m2 for the task pick, the system has a very good estimate. The uncertainty of the robot system can be considered relatively low. The task can therefore be completed autonomously, which is why L2 is chosen. If U > U m2 , then the uncertainty of the estimate can be considered very high. In this case, L0 is chosen because adequate guidance is unlikely to be achieved. L1 provides the user with a continuous measure for adjusting the split control, which can be selected as a function of U. This is discussed in Section III-C. B. Autonomy and parameters for shared control

[0049] Depending on the corresponding LoA, ASCOPE can parameterize shared control according to the progress of a task. A task defines the interaction of a reference object Oref with a target object O tar In the ring grip example, the ring serves as O tar and the surgical instrument as O ref A task consists of various subtasks. The Shared Control Templates (SCTs) introduced in

[29] model the various subtasks as states of a state machine. Fig. Figure 2b shows the pick-SCT with the states 'approach' and 'close'. The state picked (= taken) is the target state g .

[0050] Conditions on wrenches w, poses x and distances d ∈ ℝ between O ref and O tar allow the transition from one state to another. A wrench w = [f ; τ] ∈ ℝ 6 consists of forces f ∈ ℝ 3 and torques τ ∈ ℝ 3 The pose x refers to the rotation R ∈ ℝ 3×3 and the position p ∈ ℝ 3×1a homogeneous transformation matrix T ∈ SE(3) .

[0051] In

[30] , a SCOPE was introduced as a method for extending SCTs. In SCOPE, a set of parameters P is encoded in each state. These parameters are used to parameterize VFs and enable joint control depending on the current state of the task. ASCOPE adds the definition of LoA-specific parameters P loa to each state. This allows VFs to be parameterized not only depending on the current state of the task, but also depending on the corresponding LoA.

[0052] Algorithm 1: Pseudocode of the ASCOPE main loop for parsing SCTs, selecting the appropriate LoA, and parameterizing VFs. Extensions of SCOPE compared to ASCOPE are marked with *. Input: The SCT encoding the current task.

[0053] Output: The algorithm nominally ends when all states of the SGT have been processed.

[0054] Algorithm 1 describes ASCOPE. Execution starts with the first state defined in the SCT and stops when the target state g is reached, which was determined from the SCT using getGoal. getReferenceandTarget determines O ref and O tar based on the current task. The estimators provide the function getContextInformation, which calculates C based on O ref and O tar This allows the determination of state id by getCurrentState to ensure possible transitions to subsequent states. According to the state id the SCT offers the corresponding P with getParameter. As an extension to the previously introduced SCOPE, ASCOPE introduces the functions getUncertaintyEstimation, which determines U according to equation (1), and get-UMax, which determines U m1 and U m2defined by each state in the SCT. With getCurrentLoA, ASCOPE determines the appropriate LoA according to Eq. (2) based on U, U m1 and U m2 . getParameter calls P loa from the SCT based on state id and loa id off. P loa and P update the corresponding VF by parameterizeVF. C. Interactive feed forward VF for L2, L1 and L0

[0055] VFs implement a virtual spring-damper system (SDS) between the instrument TCP x and a virtual proxy. The deflection of the SDS generates virtual wrenches w. In a telemanipulation setup, these can be projected onto haptic input devices. An attractive VF (aVF) provides haptic guidance to the user, for example, to match the pose (position + rotation angle) of the instrument TCP with a desired pose.

[0056] In

[30] , a cylindrically shaped path-following aVF (also called path following aVF or paVF) is described, as in the Fig. 3a, along a generated spline s. The radius r and the length / define their shape. Their pose x a always lies on s and the longitudinal axis is aligned along s. r and I serve as translational dead zones. α defines a rotational dead zone. x moves freely within the dead zones. If a wrench w e (e.g. a user input on the haptic input device) causes x to leave the cylinder, the virtual proxy remains with the pose x p on the cylinder wall. The SDS between x p and x is deflected, causing w, which acts on x and is projected onto the haptic input device. w a = [f a , τ a ] acts on x pand has the same magnitude as w, but the opposite direction.

[0057] In the present invention, the paVF is preferably extended with a feed-forward behavior to encode various LoAs while simultaneously enabling user interactions, and is then called feed-forward VF (ffVF). Splines are preferably defined to represent the respective task. The presented ffVFs for L2 are described in Section III-C.1, L1 in Section III-C.2, and L0 in Section III-C.3. The transitions are explained in Section III-D. Table I provides an overview of the LoA-specific parameters P loa , which are presented in this section. 1) L2 - Task-level autonomy:

[0058] L2 requires precise autonomous guidance along an arbitrary spline. The dead zones r = I = α = 0 enclose the virtual representative x a with the position p aand force precise guidance through x p = x a For autonomous guidance along any spline s, x a move along s. A force ff‖ could cause this movement. Therefore, a virtual mass point with mass M and damping factor D is placed at x a introduced. The equation of motion Mp¨a+Dp˙a=ff‖. ff‖ can be considered as a constant force fc‖ along s, which causes a continuous movement. x p is connected to x by an SDS and therefore leads to autonomous leadership.

[0059] User interactions are a core component of ffVF. f e contains the sum of the user forces on the input device and the resistance forces, consisting of other VFs, friction, and damping. f e acts on x. It causes the deflection of the SDS between x and xp , which leads to f a leads. fa‖ is the component of f a along s. ff‖ as ff‖=fc‖+fa‖. Selecting this option ensures that the user can interact with the ffVF by accelerating or decelerating it and ensures consistency with other VFs. 2) L1 - Robot assistance:

[0060] At L1, the robot supports the user through shared control (also called joint control). The user controls the movement of x along the spline and receives haptic support in the remaining DoFs.

[0061] The degree of shared control (DoSC) describes a gradual adjustment of the degree of assistance. Strong assistance ensures very precise guidance in direction s by defining a small desired radial dead zone r. dand a small desired rotational dead zone α d . With low support, x remains within a region of interest (ROI) defined by a larger r d and α d is defined, while the user can perform fine manipulation.

[0062] In some preferred embodiments of the present invention, the DoSC is determined by varying r d and α d continuously adjusted. If the estimators provide context information C with a high uncertainty U, the DoSC can be chosen small, and in particular, it can be chosen as follows rd=krU, where k r is a weighting factor.

[0063] If r d is changed, the current radius r should follow this change. Here too, a force orthogonal to s allows fr⊥ a smooth and safe change of r. A virtual mass point with mass M r and the damping factor D r introduced, whose position r represents. Mrr¨+Drr˙=fr⊥.

[0064] If fr⊥ is chosen so that it is subjected to a constant force fo⊥ This allows the decrease and increase of r by fr⊥{>0if r>rd<0if r <rd

[0065] The force exerted by the user on the input device, as well as the resistance forces composed of other VFs, friction, and damping, are included in f_, which acts on r. It causes the deflection of the SDS between x and x p , which leads to f, with fa⊥ as its component orthogonal to s. The choice of fr⊥ follows according to fr⊥={fo⊥+fa⊥if r≠rd0otherwise.

[0066] If r = r_, fr⊥ set to 0 to ensure the desired haptic guidance. The decrease of r is in Fig. 3c. The same method is used for x d applied depending on J.

[0067] If the user moves the input device along s, the corresponding forces go into the component of f e along s and thus at / = 0 in fa‖ According to equation (4), fa‖ in ff‖ integrated, resulting in an update of p a as in Eq. (3).

[0068] Overall, in the Fig. 3 a ffVF shown at x a attached to s see Fig. (a), and adapted for L2 (see Fig. (b)) and L1 (see Fig. (c)). r and I are marked in the figures. s is shown as a curve in the figures above and as a horizontal line in the figures below. f e(light arrows) combines the sum of the user forces on the input device and other opposing forces, i.e. VF and other dynamic effects of the input device, such as inertia and friction. fe steers the SDS between x p and x. This creates a force f, which causes xx p follows. The dark circles represent the mass points. Fig. shows how x p remains on the cylinder wall when f e causes x to leave the cylinder and f and f a generated. r and I determine the length and radius of the cylinder. Fig. shows that r = l = 0 in L2 to x a = x p leads. ff‖, that of fe‖ through ƒa|| is influenced, affects x a and causes its movement. Finally, the Fig. that in L1 DoSC r d determined. If r d ≠ r, acts ƒr⊥ to a damped mass point at x r , which decreases r. x p stays on r. 3) L0 - no autonomy:

[0069] During L0, the user can freely telemanipulate the TCP instrument within its working range. By continuously adjusting the desired dead zones r d and α d at x and thus also r and α, so that x always remains within the cylinder, the desired behavior is achieved. l = 0 causes an update of the motion of p a , as explained in the previous paragraph. The continuous updating of r d , α d , r, α and p a enables continuous switching to L1 and L2. D. Transitions between the LoAs Table I P loa for various LoAs Typ L0 L1 L2 Beschreibung ƒc|| 0 0 > 0 Kraft entlang s r > 0 r d 0 radiale Totzone α > 0 α d 0 Rotations-Totzone l 0 0 0 axiale Totzone

[0070] Table I provides an overview of the required P loa, which are described in detail in Section III-C. In L1, r is varied according to Eq. (6) until the desired radial dead zone r d is reached. α is treated accordingly.

[0071] The presented ffVF encodes L0, L1 and L2 by introducing a feed-forward behavior that is ƒƒ|| is shown, and the variation of the dead zones r d and α n . The variation of ƒƒ||, r d and α d also enables transitions between different LoAs. Table I gives an overview of these parameters for the different LoAs. For transitions from L0 to L1, α d and r d adjusted according to the current DoSC and equation (5). At α d ≠ α or r d≠ r, the corresponding deadband can be varied as explained in Section III-C.2. This provides the user with as much guidance as the uncertainty of the estimators allows. The transition from L1 to L0 is achieved by continuously adjusting α d and r d at x and thus also of r and α, so that x remains within the cylindrical VF.

[0072] For a transition from L1 to L2, minimal dead zones are required, which leads to r d = α d = 0. r d and α d are achieved by applying ƒr⊥ according to Eq. (6). ƒc||>0N leads to a gradually increasing acceleration of x a , which leads to an autonomous movement of x along s. For the transition from L2 to L1, ƒc||=0 no autonomous movement of x a more. r d and α dare adjusted according to the current DoSC. IV. Evaluation

[0073] This section presents experiments to evaluate the proposed methods. They are carried out on the DLR research platform for surgical robots MiroSurge, which is located in Fig. 1. In the Fig. Figure 1 shows the surgeon's console with a haptic input device (a). This controls the surgical instrument and the endoscope, each attached to the MIRO robotic arms shown in figure (b). Figure (c) shows the endoscope image, which provides insight into the evaluation task performed in Section IV-C. A 7-DoF lightweight robotic arm (MIRO) is mounted on the left side rail of an operating table. A MICA drive unit is attached to the MIRO, which actuates a wrist-worn surgical instrument with an additional 2 DoFs. A Force Dimension sigma.7 controls the pose of the instrument (TCP x). The sigma.7 is a 6 DoF input device with an additional grasping DoF. It allows the representation of wrenches in 7 DoFs. A second MIRO, carrying an endoscope, is mounted on the operating table.A software layer uses a position controller to map the motion of up to two input devices to x and the endoscope TCP of several avatar robots, using inverse kinematics that maintains the motion constraints at the incision point (see

[31] and

[32] ). In all experiments, M = 0.01 kg and D = 10 Ns / m are set. The experiments described in Section IV-A and Section IV-B use a simulated MICA and MIROs as described in Chapter 8 of

[33] . The experiment presented in Section IV-C is performed with the real MICA and MIROs. In these experiments, a haptic input device from sigma.7 is used. A. Interactive autonomy at task level L2

[0074] In this experiment, the compliant behavior of ffVF compared to other VFs is evaluated. For this purpose, a virtual wall is implemented to avoid the joint limits of the sigma.7, as described in

[30] . Reaching the virtual wall triggers the counter-rotating virtual wrenches w w = [f w , τ w ] that are projected onto the input device. w w influences the movement of ffVF. For this experience, x follows a predefined spline without user interaction on the input device. The uncertainties of the estimators are preset to ensure L2.

[0075] Hypothesis: Before x reaches the wall, it follows x p . When x comes into contact with the virtual wall, the movement of x p = x a slowed down and finally stopped according to the movement of x.

[0076] Experimental setup: A single-state SCT is used, which defines a linear spline s between p ref , the position of O ref , and p tar , the position of O tar , L2 is defined throughout the experiment with ƒc||=2N and r d = a d = 0. There are no user interactions with the sigma.7.

[0077] Result: Fig. 4 shows that the movement of x is slowed down when it reaches the virtual wall. p slowed down and remains in equilibrium of ƒa|| and ƒc||, ie ƒƒ||=0 stand. Fig. shows the positions of x and x p in the axial direction to s (see second curve from the top), as well as their distance d (see first curve from the top). In addition, in (a) the temporal development of ƒc|| (third curve from the top), (bottom curve) and ƒƒ|| (second curve from the bottom). At t = 0s, x and x p to move, as in Fig. is slightly disturbed by the dynamic effects of the input device. At t = 2.2s, x hits the Fig. virtual wall shown. With increasing ƒw|| the movement of x slows down and its distance to x p increases (see Fig. This increases ƒa||. Accordingly, the movement of x p slowed down until it is in equilibrium of ƒa|| and ƒc|| comes to a standstill, ie ƒƒ||=0 (see Fig. .

[0078] Discussion: This experiment demonstrates how interactive autonomy at the task level can be achieved through ffVF. The interactive behavior in the presence of a repulsive VF demonstrates that ffVF cooperates with other VFs. The same approach allows user interactions such as acceleration or deceleration during L2. B. Transitions between the LoAs

[0079] In the following experiment, the transitions between L2 and L1 are investigated in detail by artificially varying U while x is shifted along s to x p follows.

[0080] Hypothesis: When U is varied, the appropriate LoA changes accordingly, leading to transitions between L2 and L1 and continuous adjustments of DoSC. These transitions and adjustments lead to smooth trajectories of x.

[0081] Experimental setup: The same SCT as in Section IV-A generates s. Artificial changes in U are made according to Fig. 5b. During L1, the user telemanipulates the DoF along s, while all other DoFs are constrained by ffVF. In L2, there is no user interaction with the sigma.7.

[0082] Result: Fig. Figure 5 shows the movement of x (see curve along (A), (C), and (D)) during L1 and L2. The user experiences both unrestricted movements within the radial dead zone and distortions when x leaves the dead zone. Varying U leads to variations of r. d and thus to an increase or decrease of r. x follows the movement of the cylinder with a smooth trajectory. As in the Fig. 5, x (curve along (A), (B), (C) and (D)) moves along s (linear curve), as shown in the Fig. is shown. Fig. shows the variation of U, which is a variation of LoA, f c and r d (A) The handover from L2 to L1 causes ƒc||=0N, therefore x moves p not autonomously along s. The user takes over and moves x freely within the cylinder. Since x is inside the cylinder, there are no virtual wrenches generated by the ffVF. (B) U is further increased, resulting in an increase of r d leads. (C) r d is gradually decreased according to U. The user moves along the spline while the ffVF controls the remaining Cartesian DoF. (D) U = 0, L2 takes over, the user removes their hand from the input device.

[0083] Discussion: The evaluated concept of ffVF and ASCOPE enables transitions between the displayed LoA while ensuring smooth trajectories of x. If x remains within the cylindrical ffVF at L1, the user can manipulate without a virtual wrench, resulting in a movement similar to that at L0. When reaching the lateral region of the cylinder, support is provided in the form of haptic guidance. C. Switching between different SCTs

[0084] According to Nagy et al., basic principles required in surgery are developed through peg transfer tasks [5]. Therefore, in this experiment, a peg transfer task was implemented on the real MICA and MIROs, which Fig. 1. The SCT image (shown in Fig. 2b), the transition and placement are executed sequentially. Control over the opening and closing of the gripper DoF always remains with the user. Hypothesis: If SCT Pick, Transition, and Place are executed sequentially, the ring will be successfully moved from x to x, taking advantage of the transitions between L2 and L1 and vice versa. s after x g transferred.

[0085] Experimental setup: The SCT pick, transition and place work with O ref = x. pick sets O tar = x s , while transition and place x g as O tar As soon as the state of the currently active SCT is reached, the state stateo of the subsequent SCT is activated. During the experiment, U is artificially varied, taking into account an occlusion of the tracking module's line of sight. The variation of U according to Fig. 5b leads to different LoA, which leads to a variation of r d leads, while αd = 0. During L1, the user controls the DoF along the spline and movement within the dead zones. During L2, the user either removes the hand from the input device or interacts with L2 to adjust the speed of task execution.

[0086] Result: The ring is successfully removed from x s after x g transferred. Fig. Figure 6 shows the movement of x and the different r during the experiment. As in the Fig. 6, x (curves inside the cylinders that run around the given motion path) moves along different s (given motion path inside the cylinders) to reach the ring at x s and at x g U is placed according to Fig. modulated, which also contains the legend.

[0087] Discussion: Changes in U cause changes in DoSC. During L2, the user interacts with the input device. Therefore, this experiment demonstrates the combination of the experiments in Section IV-A and Section IV-B, which leads to the successful completion of a meaningful task in surgeon training. V. Discussion and Conclusion

[0088] This work presents a novel method that can be used in particular for teleoperation applications. The method can provide the user with haptic augmentation during L1 and L2 and simultaneously enable continuous transitions between L0, L1, L2, and vice versa. Furthermore, according to some preferred embodiments, the method according to the invention also enables user interactions in L2. Estimators can preferably provide context information as well as an uncertainty parameter determined from the context information, on the basis of which an appropriate LoA for the current situation can be selected. During the execution of an SCT, ASCOPE can parameterize the VFs based on the adequate LoA and the current context information. The presented ffVF enables both haptic augmentation during L1 and autonomous movements during L2.The design of the ffVF takes into account the forces applied to the input device, enabling user interactions such as acceleration and deceleration in L2. Furthermore, it ensures that the ffVF obeys other VFs, such as the rVF system presented in

[30] . Furthermore, the ffVF follows splines, allowing for generalization to various tasks. Further preferred embodiments of the invention

[0089] Further preferred embodiments of the invention are summarized below

[0090] The present invention may preferably provide the following features: 0.1 A robot system that can be controlled either hands-on (torque- or Cartesian impedance-controlled (lightweight) robot) or by telemanipulation. 0.2 In the case of telemanipulation, a haptic input device where forces and torques can be set in at least one, ideally all, degrees of freedom. 0.3 Using virtual fixtures with dead zones. 1.1 The force can be determined in such a way that user interactions with the input device can influence the position of the VF at any time. This makes the VF interactive for the user at all times. 1.2 In addition, the force can be defined in such a way that other VFs can be complied with, which represent, for example, the system limits of the robot. 1.3 This force acts on a virtual mass point that is introduced at the origin of the VF. 1.4 This mass point receives a damping. 1.5 The force acts in a specific direction, e.g., along a spline. Thus, the VF can be moved along a spline. 2 The dead zone (translational and rotational) can be increased or decreased at runtime. 2.1 The enlargement / reduction of the dead zone is triggered by a force. 2.2 The force can be determined in such a way that user interactions with the input device always influence the position of the VF. This makes the VF interactive for the user at all times. 2.3 In addition, the force can be defined in such a way that other VFs are respected, which represent, for example, the system limits of the robot. 2.4 This force is applied to a virtual mass point that is introduced at the position of the current dead zone. 2.5 The mass point receives a damping. 2.6 The enlargement / reduction of the dead zone influences the precision of the haptic augmentation. 3 Using feature groups 1 and 2, different levels of autonomy can be realized. 3.1 Task-level autonomy: The force according to feature group 1 is defined along a trajectory. This causes the IP (Interest Point) to follow this trajectory autonomously. 3.2 Shared control: 3.2.1 The force according to feature group 1 can be set in such a way that no autonomous movements of the IP result. 3.2.2 The dead zones can be adjusted using feature group 2 so that the inaccuracies of the system can be adjusted by the user. 3.3 Direct control (HandsOn, telemanipulation): The dead zones can be adjusted so that the user does not feel any haptic augmentation. 3.4 For the transition between task-level autonomy and shared control, the force can be adjusted according to feature group 1. 3.5 For the transition between shared control and remote manipulation, the dead zones may be adjusted in accordance with feature group 2. Bibliography [1] T. Haidegger, „Autonomy for Surgical Robots: Concepts and Paradigms,“ IEEE Trans. Med. Robot. Bionics, vol. 1, no. 2, pp. 65-76, Mai 2019. [2] A. Attanasio, B. Scaglioni, M. Leonetti, A. F. Frangi, W. Cross, C. S. Biyani, und P. Valdastri, „Autonomous Tissue Retraction in Robotic Assisted Minimally Invasive Surgery - A Feasibility Study,“ IEEE Robot. Autom. Lett., Vol. 5, No. 4, pp. 6528-6535, Oktober 2020. [3] P. Fiorini, K. Y. Goldberg, Y. Liu, und R. H. Taylor, „Concepts and Trends in Autonomy for Robot-Assisted Surgery,“ Proc. IEEE, Vol. 110, No. 7, pp. 993-1011, Juli 2022. [4] T. D. Nagy und T. Haidegger, „Performance and Capability Assessment in Surgical Subtask Automation,“ Sensors, Vol. 22, No. 7, p. 2501, März 2022, Number: 7. [5] T. D. Nagy und T. P. Haidegger, „Towards Standard Approaches for the Evaluation of Autonomous Surgical Subtask Execution,“ in 2021 IEEE 25th International Conference on Intelligent Engineering Systems (INES). Budapest, Hungary: IEEE, Juli 2021, pp. 000 067-000 074. [6] M. Stoiber, M. Sundermeyer, und R. Triebel, „Iterative corresponding geometry: Fusing region and depth for highly efficient 3d tracking of textureless objects,“ in Proceedings of the IEEE / CVF Conference on Computer Vision and Pattern Recognition (CVPR), Juni 2022, pp. 6855-6865. [7] M. Hwang, J. Ichnowski, B. Thananjeyan, D. Seita, S. Paradis, D. Fer, T. Low, und K. Goldberg, „Automating Surgical Peg Transfer: Calibration With Deep Learning Can Exceed Speed, Accuracy, and Consistency of Humans,“ IEEE Trans. Automat. Sci. Eng., pp. 1-14, 2022. [8] J. Rosen und J. Ma, „Autonomous Operation in Surgical Robotics,“ Mechanical Engineering, Vol. 137, No. 09, pp. S15-S18, September 2015. [9] M. M. Rahman, M. V. Balakuntala, G. Gonzalez, M. Agarwal, U. Kaur, V. L. N. Venkatesh, N. Sanchez-Tamayo, Y. Xue, R. M. Voyles, V. Aggarwal, und J. Wachs, „SARTRES: a semi-autonomous robot teleoperation environment for surgery,“ Computer Methods in Biomechanics and Biomedical Engineering: Imaging & Visualization, Vol. 9, No. 4, pp. 376-383, Juli 2021.

[10] S. Sen, A. Garg, D. V. Gealy, S. McKinley, Y. Jen, und K. Goldberg, „Automating multi-throw multilateral surgical suturing with a mechanical needle guide and sequential convex optimization,“ in 2016 IEEE International Conference on Robotics and Automation (ICRA). Stockholm, Schweden: IEEE, Mai 2016, pp. 4178-4185.

[11] Hyosig Kang und J. Wen, „Autonomous suturing using minimally invasive surgical robots,“ in Proceedings of the 2000. IEEE International Conference on Control Applications. Conference Proceedings (Cat. No.00CH37162). Anchorage, AK, USA: IEEE, 2000, pp. 742-747.

[12] S. A. Pedram, C. Shin, P. W. Ferguson, J. Ma, E. P. Dutson, und J. Rosen, „Autonomous Suturing Framework and Quantification Using a Cable-Driven Surgical Robot,“ IEEE Trans. Robot., vol. 37, no. 2, pp. 404-417, April 2021.

[13] M. Kam, H. Saeidi, M. H. Hsieh, J. U. Kang, und A. Krieger, „A Confidence-Based Supervised-Autonomous Control Strategy for Robotic Vaginal Cuff Closure,“ in 2021 IEEE International Conference on Robotics and Automation (ICRA). Xi'an, China: IEEE, Mai 2021, pp. 12 261-12 267.

[14] H. Saeidi, J. D. Opfermann, M. Kam, S. Wei, S. Leonard, M. H. Hsieh, J. U. Kang, und A. Krieger, „Autonomous robotic laparoscopic surgery for intestinal anastomosis,“ Sci. Robot., vol. 7, no. 62, p. eabj2908, Januar 2022.

[15] D. Zhang, B. Xiao, B. Huang, L. Zhang, J. Liu, und G.-Z. Yang, „A Self-Adaptive Motion Scaling Framework for Surgical Robot Remote Control,“ IEEE Robot. Autom. Lett., vol. 4, no. 2, pp. 359-366, April 2019.

[16] L. Rosenberg, „Virtual fixtures: Perceptual tools for telerobotic manipulation,“ in Proceedings of IEEE Virtual Reality Annual International Symposium. Seattle, WA, USA: IEEE, 1993, pp. 76-82.

[17] J. J. Abbott, P. Marayong, und A. M. Okamura, „Haptic Virtual Fixtures for Robot-Assisted Manipulation,“ in Robotics Research, S. Thrun, R. Brooks, and H. Durrant-Whyte, Eds. Berlin, Heidelberg: Springer Berlin Heidelberg, 2007, vol. 28, pp. 49-64, series Title: Springer Tracts in Advanced Robotics.

[18] S. A. Bowyer, B. L. Davies, und F. Rodriguez y Baena, „Active Constraints / Virtual Fixtures: A Survey,“ IEEE Trans. Robot., vol. 30, no. 1, pp. 138-157, Februar 2014.

[19] M. Selvaggio, G. A. Fontanelli, F. Ficuciello, L. Villani, und B. Siciliano, „Passive Virtual Fixtures Adaptation in Minimally Invasive Robotic Surgery,“ IEEE Robot. Autom. Lett., vol. 3, no. 4, pp. 3129-3136, Oktober 2018, Nummer: 4.

[20] Z. Li, A. Gordon, T. Looi, J. Drake, C. Forrest, und R. H. Taylor, „Anatomical Mesh-Based Virtual Fixtures for Surgical Robots,“ arXiv:2006.02415 [cs, eess], Juli 2020, arXiv: 2006.02415.

[21] M. Muhlbauer, F. Steinmetz, F. Stulp, T. Hulin, und A. Albu-Schaffer, „Multi-Phase Multi-Modal Haptic Teleoperation,“ in 2022 IEEE / RSJ International Conference on Intelligent Robots and Systems (IROS). Kyoto, Japan: IEEE, Oktober 2022, pp. 7734-7741.

[22] M. Muhlbauer, F. Stulp, A. Albu-Schaffer, und J. Silverio, „Mixture of experts on Riemannian manifolds for visual-servoing fixtures.“

[23] G. A. Fontanelli, G.-Z. Yang, und B. Siciliano, „A Comparison of Assistive Methods for Suturing in MIRS,“ in 2018 IEEE / RSJ International Conference on Intelligent Robots and Systems (IROS). Madrid: IEEE, Oct. 2018, pp. 4389-4395.

[24] H. Saeidi, J. D. Opfermann, M. Kam, S. Raghunathan, S. Leonard, und A. Krieger, „A Confidence-Based Shared Control Strategy for the Smart Tissue Autonomous Robot (STAR),“ in 2018 IEEE / RSJ International Conference on Intelligent Robots and Systems (IROS), Oktober 2018, pp. 1268-1275, iSSN: 2153-0866.

[25] F. Ferraguti, N. Preda, G. De Rossi, M. Bonfe, R. Muradore, P. Fiorini, und C. Secchi, „A two-layer approach for shared control in semiautonomous robotic surgery,“ in 2015 European Control Conference (ECC). Linz, Austria: IEEE, Jul. 2015, pp. 747-752.

[26] K. Shamaei, Y. Che, A. Murali, S. Sen, S. Patil, K. Goldberg, und A. M. Okamura, „A paced shared-control teleoperated architecture for supervised automation of multilateral surgical tasks,“ in 2015 IEEE / RSJ International Conference on Intelligent Robots and Systems (IROS). Hamburg, Germany: IEEE, September 2015, pp. 1434-1439.

[27] R. Balachandran, H. Mishra, M. Cappelli, B. Weber, C. Secchi, C. Ott, und A. Albu-Schaeffer, „Adaptive authority allocation in shared control of robots using bayesian filters,“ in 2021 IEEE International Conference on Robotics and Automation (ICRA), 2020, pp. 11 298-11 304.

[28] S. Bustamante, G. Quere, K. Hagmann, X. Wu, P. Schmaus, J. Vogel, F. Stulp, und D. Leidner, „Toward Seamless Transitions Between Shared Control and Supervised Autonomy in Robotic Assistance,“ IEEE Robot. Autom. Lett., vol. 6, no. 2, pp. 3833-3840, April 2021.

[29] G. Quere, A. Hagengruber, M. Iskandar, S. Bustamante, D. Leidner, F. Stulp, and J. Vogel, „Shared Control Templates for Assistive Robotics,“ in 2020 IEEE International Conference on Robotics and Automation (ICRA), Mai 2020, pp. 1956-1962, iSSN: 2577-087X.

[30] K. Hagmann, A. Hellings-Kuß, J. Klodmann, R. Richter, F. Stulp, and D. Leidner, „A Digital Twin Approach for Contextual Assistance for Surgeons During Surgical Robotics Training,“ Front. Robot. AI, vol. 8, p. 735566, September 2021.

[31] C. Schlenk, T. Bahls, S. Tarassenko, J. Klodmann, M. Bihler, and T. Wuesthoff, „Robot Integrated User Interface for Physical Interaction with the DLR MIRO in Versatile Medical Procedures,“ J. Med. Robot. Res., vol. 03, no. 02, p. 1840006, Juni 2018, Nummer: 02.

[32] U. Seibold, K. Bernhard, T. Bahls, R. Haslinger, und F. Steidle, „THE DLR MIROSURGE SURGICAL ROBOTIC DEMONSTRATOR,“ in The Encyclopedia of Medical Robotics. Patel R, 2018, vol. Volume 1: Minimally Invasive Surgical RoboticsVolume, pp. 111-142.

[33] L. Grespan, P. Fiorini, und G. Colucci, The Route to Patient Safety in Robotic Surgery, ser. Springer Tracts in Advanced Robotics. Cham: Springer International Publishing, 2019, vol. 126.

[34] G. Quere, S. Bustamante, A. Hagengruber, J. Vogel, F. Steinmetz, and F. Stulp, “Learning and Interactive Design of Shared Control Templates,” in 2021 IEEE / RSJ International Conference on Intelligent Robots and Systems (IROS). Prague, Czech Republic: IEEE, September 2021, pp. 1887-1894. QUOTES CONTAINED IN THE DESCRIPTION

[0000] This list of documents submitted by the applicant was generated automatically and is included solely for the convenience of the reader. This list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions. Cited non-patent literature

[0000] T. Haidegger, “Autonomy for Surgical Robots: Concepts and Paradigms,” IEEE Trans. Med. Robot. Bionics, vol. 1, no. 2, pp. 65-76, May 2019

[0090] A. Attanasio, B. Scaglioni, M. Leonetti, A. F. Frangi, W. Cross, C. S. Biyani, und P. Valdastri, „Autonomous Tissue Retraction in Robotic Assisted Minimally Invasive Surgery - A Feasibility Study,“ IEEE Robot. Autom. Lett., Vol. 5, No. 4, pp. 6528-6535, Oktober 2020

[0090] P. Fiorini, K. Y. Goldberg, Y. Liu, und R. H. Taylor, „Concepts and Trends in Autonomy for Robot-Assisted Surgery,“ Proc. IEEE, Vol. 110, No. 7, pp. 993-1011, Juli 2022

[0090] T. D. Nagy und T. Haidegger, „Performance and Capability Assessment in Surgical Subtask Automation,“ Sensors, Vol. 22, No. 7, p. 2501, März 2022, Number: 7

[0090] T. D. Nagy und T. P. Haidegger, „Towards Standard Approaches for the Evaluation of Autonomous Surgical Subtask Execution,“ in 2021 IEEE 25th International Conference on Intelligent Engineering Systems (INES). Budapest, Hungary: IEEE, Juli 2021, pp. 000 067-000 074

[0090] M. Stoiber, M. Sundermeyer, und R. Triebel, „Iterative corresponding geometry: Fusing region and depth for highly efficient 3d tracking of textureless objects,“ in Proceedings of the IEEE / CVF Conference on Computer Vision and Pattern Recognition (CVPR), Juni 2022, pp. 6855-6865

[0090] M. Hwang, J. Ichnowski, B. Thananjeyan, D. Seita, S. Paradis, D. Fer, T. Low, und K. Goldberg, „Automating Surgical Peg Transfer: Calibration With Deep Learning Can Exceed Speed, Accuracy, and Consistency of Humans,“ IEEE Trans. Automat. Sci. Eng., pp. 1-14, 2022

[0090] J. Rosen und J. Ma, „Autonomous Operation in Surgical Robotics,“ Mechanical Engineering, Vol. 137, No. 09, pp. S15-S18, September 2015

[0090] M. M. Rahman, M. V. Balakuntala, G. Gonzalez, M. Agarwal, U. Kaur, V. L. N. Venkatesh, N. Sanchez-Tamayo, Y. Xue, R. M. Voyles, V. Aggarwal, und J. Wachs, „SARTRES: a semi-autonomous robot teleoperation environment for surgery,“ Computer Methods in Biomechanics and Biomedical Engineering: Imaging & Visualization, Vol. 9, No. 4, pp. 376-383, Juli 2021

[0090] S. Sen, A. Garg, D. V. Gealy, S. McKinley, Y. Jen, und K. Goldberg, „Automating multi-throw multilateral surgical suturing with a mechanical needle guide and sequential convex optimization,“ in 2016 IEEE International Conference on Robotics and Automation (ICRA). Stockholm, Schweden: IEEE, Mai 2016, pp. 4178-4185

[0090] Hyosig Kang und J. Wen, „Autonomous suturing using minimally invasive surgical robots,“ in Proceedings of the 2000. IEEE International Conference on Control Applications. Conference Proceedings (Cat. No.00CH37162). Anchorage, AK, USA: IEEE, 2000, pp. 742-747

[0090] S. A. Pedram, C. Shin, P. W. Ferguson, J. Ma, E. P. Dutson, und J. Rosen, „Autonomous Suturing Framework and Quantification Using a Cable-Driven Surgical Robot,“ IEEE Trans. Robot., vol. 37, no. 2, pp. 404-417, April 2021

[0090] M. Kam, H. Saeidi, M. H. Hsieh, J. U. Kang, und A. Krieger, „A Confidence-Based Supervised-Autonomous Control Strategy for Robotic Vaginal Cuff Closure,“ in 2021 IEEE International Conference on Robotics and Automation (ICRA). Xi'an, China: IEEE, Mai 2021, pp. 12 261-12 267

[0090] H. Saeidi, J. D. Opfermann, M. Kam, S. Wei, S. Leonard, M. H. Hsieh, J. U. Kang, und A. Krieger, „Autonomous robotic laparoscopic surgery for intestinal anastomosis,“ Sci. Robot., vol. 7, no. 62, p. eabj2908, Januar 2022

[0090] D. Zhang, B. Xiao, B. Huang, L. Zhang, J. Liu, und G.-Z. Yang, „A Self-Adaptive Motion Scaling Framework for Surgical Robot Remote Control,“ IEEE Robot. Autom. Lett., vol. 4, no. 2, pp. 359-366, April 2019

[0090] L. Rosenberg, „Virtual fixtures: Perceptual tools for telerobotic manipulation,“ in Proceedings of IEEE Virtual Reality Annual International Symposium. Seattle, WA, USA: IEEE, 1993, pp. 76-82

[0090] J. J. Abbott, P. Marayong, und A. M. Okamura, „Haptic Virtual Fixtures for Robot-Assisted Manipulation,“ in Robotics Research, S. Thrun, R. Brooks, and H. Durrant-Whyte, Eds. Berlin, Heidelberg: Springer Berlin Heidelberg, 2007, vol. 28, pp. 49-64, series Title: Springer Tracts in Advanced Robotics

[0090] S. A. Bowyer, B. L. Davies, und F. Rodriguez y Baena, „Active Constraints / Virtual Fixtures: A Survey,“ IEEE Trans. Robot., vol. 30, no. 1, pp. 138-157, Februar 2014

[0090] M. Selvaggio, G. A. Fontanelli, F. Ficuciello, L. Villani, und B. Siciliano, „Passive Virtual Fixtures Adaptation in Minimally Invasive Robotic Surgery,“ IEEE Robot. Autom. Lett., vol. 3, no. 4, pp. 3129-3136, Oktober 2018, Nummer: 4

[0090] Z. Li, A. Gordon, T. Looi, J. Drake, C. Forrest, und R. H. Taylor, „Anatomical Mesh-Based Virtual Fixtures for Surgical Robots,“ arXiv:2006.02415 [cs, eess], Juli 2020, arXiv: 2006.02415

[0090] M. Muhlbauer, F. Steinmetz, F. Stulp, T. Hulin, und A. Albu-Schaffer, „Multi-Phase Multi-Modal Haptic Teleoperation,“ in 2022 IEEE / RSJ International Conference on Intelligent Robots and Systems (IROS). Kyoto, Japan: IEEE, Oktober 2022, pp. 7734-7741

[0090] M. Muhlbauer, F. Stulp, A. Albu-Schaffer, und J. Silverio, „Mixture of experts on Riemannian manifolds for visual-servoing fixtures.

[0090] G. A. Fontanelli, G.-Z. Yang, und B. Siciliano, „A Comparison of Assistive Methods for Suturing in MIRS,“ in 2018 IEEE / RSJ International Conference on Intelligent Robots and Systems (IROS). Madrid: IEEE, Oct. 2018, pp. 4389-4395

[0090] H. Saeidi, J. D. Opfermann, M. Kam, S. Raghunathan, S. Leonard, und A. Krieger, „A Confidence-Based Shared Control Strategy for the Smart Tissue Autonomous Robot (STAR),“ in 2018 IEEE / RSJ International Conference on Intelligent Robots and Systems (IROS), Oktober 2018, pp. 1268-1275, iSSN: 2153-0866

[0090] F. Ferraguti, N. Preda, G. De Rossi, M. Bonfe, R. Muradore, P. Fiorini, und C. Secchi, „A two-layer approach for shared control in semiautonomous robotic surgery,“ in 2015 European Control Conference (ECC). Linz, Austria: IEEE, Jul. 2015, pp. 747-752

[0090] K. Shamaei, Y. Che, A. Murali, S. Sen, S. Patil, K. Goldberg, und A. M. Okamura, „A paced shared-control teleoperated architecture for supervised automation of multilateral surgical tasks,“ in 2015 IEEE / RSJ International Conference on Intelligent Robots and Systems (IROS). Hamburg, Germany: IEEE, September 2015, pp. 1434-1439

[0090] R. Balachandran, H. Mishra, M. Cappelli, B. Weber, C. Secchi, C. Ott, und A. Albu-Schaeffer, „Adaptive authority allocation in shared control of robots using bayesian filters,“ in 2021 IEEE International Conference on Robotics and Automation (ICRA), 2020, pp. 11 298-11 304

[0090] S. Bustamante, G. Quere, K. Hagmann, X. Wu, P. Schmaus, J. Vogel, F. Stulp, und D. Leidner, „Toward Seamless Transitions Between Shared Control and Supervised Autonomy in Robotic Assistance,“ IEEE Robot. Autom. Lett., vol. 6, no. 2, pp. 3833-3840, April 2021

[0090] G. Quere, A. Hagengruber, M. Iskandar, S. Bustamante, D. Leidner, F. Stulp, and J. Vogel, „Shared Control Templates for Assistive Robotics,“ in 2020 IEEE International Conference on Robotics and Automation (ICRA), Mai 2020, pp. 1956-1962, iSSN: 2577-087X

[0090] K. Hagmann, A. Hellings-Kuß, J. Klodmann, R. Richter, F. Stulp, and D. Leidner, „A Digital Twin Approach for Contextual Assistance for Surgeons During Surgical Robotics Training,“ Front. Robot. AI, vol. 8, p. 735566, September 2021

[0090] C. Schlenk, T. Bahls, S. Tarassenko, J. Klodmann, M. Bihler, and T. Wuesthoff, „Robot Integrated User Interface for Physical Interaction with the DLR MIRO in Versatile Medical Procedures,“ J. Med. Robot. Res., vol. 03, no. 02, p. 1840006, Juni 2018, Nummer: 02

[0090] U. Seibold, K. Bernhard, T. Bahls, R. Haslinger, und F. Steidle, „THE DLR MIROSURGE SURGICAL ROBOTIC DEMONSTRATOR,“ in The Encyclopedia of Medical Robotics. Patel R, 2018, vol. Volume 1: Minimally Invasive Surgical RoboticsVolume, pp. 111-142

[0090] L. Grespan, P. Fiorini, und G. Colucci, The Route to Patient Safety in Robotic Surgery, ser. Springer Tracts in Advanced Robotics. Cham: Springer International Publishing, 2019, vol. 126

[0090] G. Quere, S. Bustamante, A. Hagengruber, J. Vogel, F. Steinmetz, und F. Stulp, „Learning and Interactive Design of Shared Control Templates,“ in 2021 IEEE / RSJ International Conference on Intelligent Robots and Systems (IROS). Prague, Czech Republic: IEEE, September 2021, pp. 1887-1894

[0090]

Claims

[1] Robot system comprising at least one robot arm and a control unit, wherein the control unit is configured to: - to operate the robot system in a first operating mode, a second operating mode and a third operating mode and to switch between the individual operating modes; - to determine an autonomy parameter that defines a gradually adjustable degree of autonomy for the control of the robot system within the second operating mode, - to control the robot system in the first operating mode in response to a user input; - to control the robot system in the second operating mode depending on a user input, a specified movement sequence and the autonomy parameter; and - to control the robot system in the third operating mode depending on a predetermined movement sequence; where - the control unit is configured to set the degree of autonomy of the control defined by the autonomy parameter in the second operating mode using a virtual mount and by varying the parameters of the virtual mount. [2] Robot system according to claim 1, characterized by that the virtual holder is defined by a geometric body and by geometric parameters that describe the geometric body, wherein a dead zone is defined in the interior of the geometric body, within which inputs from a user to control the robot system are implemented without restriction, wherein the virtual holder is preferably defined by a cylinder, an ellipsoid or a cuboid. [3] Robot system according to claim 1 or 2, characterized by that the control unit is configured to switch the operating mode and / or the level of autonomy depending on a user input. [4] Robot system according to claim 1 or 2, characterized by that the control unit is configured to set the operating mode and / or the degree of autonomy automatically and depending on previously determined context information. [5] Robot system according to claim 4, characterized by that the control unit is configured to determine an uncertainty parameter U for automatically setting the operating mode and / or the degree of autonomy, wherein the uncertainty parameter U represents the unreliability of measurement data recorded within the robot system. [6] Robot system according to claim 5, characterized by that the control device is designed to determine the operating mode as a function of the determined uncertainty parameter U, wherein the control device is designed to: - to select the first operating mode if the uncertainty parameter U exceeds a first threshold U1; - to select the second operating mode, provided that the uncertainty parameter U is between the first threshold and a second threshold U2, where U1 is greater than U2; and - to select the third operating mode if the determined uncertainty parameter U falls below the second threshold value U2. [7] Robot system according to one of claims 5 to 6, characterized by that the Virtual Mount is defined by a cylinder having a cylinder radius r d where the cylinder radius r d in the second operating mode according to rd=krU is set as a function of a previously determined uncertainty parameter U, where k r describes a weighting factor. [8] Robot system according to one of claims 1 to 7, characterized bythat the control unit is configured to switch from the third operating mode to the second operating mode if the control unit detects a force exerted by the user via an input device that exceeds a predetermined force threshold. [9] Robot system according to one of claims 1 to 8, characterized by in that the control unit is configured to control the robot system in a third operating mode such that a component of the robot system is moved according to a predetermined movement sequence, wherein the speed of the movement is adjustable depending on a user input. [10] Robot system according to one of claims 1 to 9, characterized by an input device configured to detect a user input and, in response to the user input, generate an input signal corresponding to the input and transmit the input signal to the control unit. [11] Method for controlling a robot system comprising a robot arm and a control unit, the method comprising the following steps: - controlling the robot system in a first operating mode, a second operating mode and a third operating mode, wherein switching between the operating modes is possible; - Determining an autonomy parameter that is gradually adjustable and defines the degree of autonomy of the control of the robot system during the second operating mode; - controlling the robot system in the first operating mode in response to a user input, provided the first operating mode is selected; - controlling the robot system in the second operating mode depending on a user input, a predetermined movement sequence and the autonomy parameter, if the second operating mode is selected; and - controlling the robot system in the third operating mode depending on a predetermined movement sequence, provided that the third operating mode is selected; - wherein the degree of autonomy of the control of the robot system in the second operating mode is adjusted using a virtual holder and by varying the parameters of the virtual holder.

Images