Collaborative handling robot with mixed force control laws providing high effector sensitivity and allowing interaction with the robot's body - Patents.com

Abstract

A collaborative process robot is disclosed having a mixed force control law that provides high effector sensitivity and allows interaction with the robot's body. The invention essentially consists of carefully positioning a multi-axis force sensor between the end member (flange) of an industrial collaborative process robot and the implement supported thereby, and modifying the increased force control law implemented in the robot controller by introducing a saturation function.

Description

[Technical Field]

[0001] The present invention relates to the field of robotics, and more particularly to physical human-robot interaction (pHRI) utilized by collaborative processing robots.

[0002] Generally, pHRI is an operation that allows a human operator to enter the working area of ​​a robot, allowing one to directly and physically interact with the other.

[0003] The present invention more particularly relates to a force increasing control law for a collaborative processing robot. [Background technology]

[0004] In the field of robotics, there are various systems that allow an operator to be assisted in an action.

[0005] To remotely manipulate objects and perform laborious tasks, there are first systems called teleoperated systems, which generally consist of a control arm and a controlled arm coupled to each other.

[0006] However, these are systems that are complex in both their design and use. As such, they prove costly and difficult to master. Generally, the manufacturability achieved with the systems is inferior to that achieved by operating directly on the components with bare hands or through tools to perform the work.

[0007] To assist operators in performing complex and / or effortful tasks while simultaneously maintaining systems that are simpler than teleoperated systems, systems known as collaborative processing systems have been developed. These systems are generally made up of collaborative processing robots or co-robots that perform tasks accomplished using tools and that are equipped with guides that allow a human operator to control the movement of the collaborative processing robot via the guides.

[0008] Collaborative processing thus allows the tool to be operated together by the robot and the operator so that the operator is assisted in carrying out the task thus accomplished.

[0009] More generally, this mode of interaction between humans and robots can provide several functions to assist movement, such as by compensating for the weight of the implement, adding programmable mechanical constraints, and multiplying forces.

[0010] Thus, through collaborative processing, various learning / programming functions are performed in situ by guiding the robot through the points, movements and / or forces involved in a task, which task is then performed automatically by the same robot.

[0011] The suitability of a robot for collaborative processing depends primarily on its sensitivity to the forces applied by the operator and its environment relative to its end element or member with the tool or gripper. The quality of collaborative processing as perceived by the operator is very strongly linked to the idea of ​​mechanical transparency achieved by the robot system (robot mechanisms, sensors, and actuators subject to control laws).

[0012] The transparency of a robotic system describes its ability to move in an unconstrained direction while minimizing the forces of interaction with the operator and / or its environment.

[0013] A system that exhibits complete transparency is able to follow movements made to the implement by the operator in the constrained space without offering any resistance to such movements, and obstacles caused by the robot are not perceived by the operator.

[0014] Robotic systems that can enable collaborative processing can be divided into three distinct categories.

[0015] The first concerns mechanically transparent coupled systems. The mechanical design of the actuators allows for excellent force transmission both from the motors to the segments and end pieces, and from the segments and end pieces to the motors. Besides any constraints that may be on the displacement, the control law only needs to compensate for the weight of the robot and the instrument as experienced by the joints, to enable transparent collaborative processing, as disclosed in publication [1] and patent document 1.

[0016] The second category is systems with insufficiently transparent joints: the mechanical friction within the joints, including reduction gears and motors, transmitted to the point of interaction with the operator during the movement process exceeds the difficulty threshold.

[0017] Finally, there are systems with irreversible joints. At zero motor force, these joints are stuck and no movement is possible, no matter what force is applied to the robot by the operator. This is especially the case for joints created with worm / wheel type mechanical systems with high reduction ratios.

[0018] The present invention relates exclusively to the category of systems with insufficiently transparent and / or irreversible joints, which currently covers the majority of industrial robots on the market. In particular, industrial robots have been designed first and foremost to optimize positioning accuracy / repeatability at the expense of the ability to interact with a human operator.

[0019] The inventors therefore sought to improve the collaborative processing of existing industrial robots with joints that are insufficiently mechanically transparent or irreversible.

[0020] The inventor has analyzed the various drawbacks / limitations of these existing robots and itemized all the currently available solutions.

[0021] First, the cooperative processing of the instruments needs to be performed with better transparency than is inherent in the machine, especially with greatly reduced joint friction. To overcome the mechanical limitation on transparency in a poorly transparent system, a first solution consists of modeling the frictional forces of the joints in order to add compensation for this to the actuator control law.

[0022] This solution is almost entirely unsatisfactory for current industrial robots, since the friction model depends on highly variable parameters such as joint lubrication, temperature, and wear. Also, the variation in friction is steep and extremely non-linear around zero velocity. Compensation for friction is therefore inoperable for precise or slow movements, as is evident from the publication [2].

[0023] Also, the implemented robotic system has less sensitivity than it needs to have in the instrument, but sensitivity needs to remain throughout the robot body.

[0024] In systems with insufficiently transparent or irreversible joints, one solution is to install force sensors between the end segments or members and the tool, capable of measuring the six components of the loading pattern (three force components and three torques). The predicted weight of the tool is pre-subtracted from the force measurement. A closed-loop feedback control law is then implemented in the robot control, so that the actuators keep this force measurement at zero. This allows the robot system to move continuously in each case to counteract the force of the operator's interaction with the robot. The internal friction forces of the joint are therefore rejected, whatever their nature, without the need for any form of predictive calculation: [3].

[0025] This solution therefore allows the complete rejection of frictional forces, according to the implementation of the control law, whereas it is impossible to overcome the influence of the robot's mechanical inertia: [4].

[0026] In particular, a control law associated with its gain adjustment is described as passive if the controlled robotic system is stable in any mechanically passive environment and in its interaction with a human operator (which may itself be considered passive [6]).[5] It has then been theoretically shown that with simple correction devices, adjustments that overcompensate for the robot's inertia are no longer passive and are unstable, especially in contact with highly stiff environments or implements with high inertia.

[0027] Furthermore, the main limitation of this solution of measuring the force on the terminal element is to avoid any inertia between the robot body and its environment. In particular, the feedback control indiscriminately rejects friction and all interaction forces upstream of the force measurement sensor, since they are not measured. This can inadvertently lead to situations where the robot body exerts significant and potentially destructive forces on elements present in the workspace or on other operators, when the operator is primarily focused on moving the implement.

[0028] Another limitation of this solution is that it is difficult to master the robot's reconfigurable movements close to its unique configuration, since the force projection measured on certain joints is zero or very low. For the same reason, in the case of robots with redundant kinematics, internal movements of the robot that do not result in any displacement of the end members (end segments) cannot be controlled by instrument-only collaborative manipulation. In an anthropomorphic arm with seven degrees of freedom, it is impossible for the operator to control the elbow position during collaborative manipulation tasks without interacting with the robot body.

[0029] Furthermore, robotic systems with mechanically insufficiently transparent or irreversible joints need to remain sensitive over the entire body of the robot, even if this means lower sensitivity at the instruments.

[0030] To achieve this, one solution is to cover the surface of the robot body segments with a force-sensitive layer.

[0031] US Patent No. 5,649,999 proposes a solution to this by covering it with skin, so that if the robot body comes into contact with an element of the workspace above the detection threshold of the skin, a signal is transmitted to the robot to stop movement until the contact is removed, which generally requires some other means to detach the robot, such as a pendant.

[0032] Particularly more highly evolved skin, such as that described in US Pat. No. 6,239,693, allows determining the location of contact and its intensity, so that only those components of movement in the direction of contact are stopped, while others are left free, thereby allowing better continuity of cooperative processing.

[0033] A drawback of using skins is that they must, by definition, be adapted to the specific geometry of each robot. Another drawback is that it does not allow the operator to, for example, manipulate the robot body with their second hand, even if it would be practical to do so to control the configuration of the robot in its vicinity or during collaborative manipulation through the kinematic unity of the robot.

[0034] An alternative solution to have good sensitivity throughout the robot body is to fit force sensors that measure the six components of the loading pattern between the robot base and its plinths (three force components and three torques). Thus, the forces applied to the robot body and the tool are actually measured by sensors in the base. Since a full gravity model may be sufficient for slow movements, as described in publication [7] or in patent document 4, it is sufficient to reuse the same law for force sensors positioned on the end members, with the difference being that a prediction of the robot's dynamic loading pattern, reduced to sensors in the base, must be subtracted from the measurements.

[0035] The drawback with this solution is that the weight of the robot is in fact much greater than the force applied by the operator, thus requiring a load sensor with a much larger inner diameter with much stricter requirements on the measurement accuracy (noise, linearity and drift) required of the force sensor on the end member of the robot.

[0036] Another drawback of positioning the force sensor between the base of the robot and its prince is that the seating of the robot needs to be particularly well controlled when the robot is installed, as otherwise errors will be introduced into the prediction of the gravity loading pattern, and these will lead to significant forces that will disrupt the cooperative process.

[0037] A final drawback affects the cooperative operation of two hands on separate robot segments, where the internal force components of the robot's kinematic chain do not have a total resultant force that can be measured by a force sensor at the base. This situation may occur especially when the operator wants to bend the robot elbow by coordinating the arm and forearm together. This leads to obstacles in certain cooperative operation configurations.

[0038] Another alternative solution is to place joint torque sensors at the output of each robot joint, with one component per joint. This solution makes it possible to make the system fully sensitive to the forces applied to all segments of the robot, even in the case of two-handed cooperative processing: [8], [9].

[0039] The drawback with this solution is that it requires a new design of the robot joint, something that cannot be added to an already existing industrial robot.

[0040] Finally, another alternative solution is to implement a force-increasing control law in the robot controller. Such a control law utilizes force measurements (indirect measurements) τ from sensors positioned to measure the forces of cooperating mechanical interface elements on the implement, such as a handle, and from setpoints or motor force measurements (indirect measurements) τ of actuators on the robot. m, for example, by combining measurements of motor current or actuator pressure.

[0041] Such a control is described in particular in US Pat. No. 5,629,999.

[0042] As explained in the publication

[10] , the principle of force multiplication control is as follows: The loading pattern of measurements from sensors positioned to measure the force of a cooperating mechanical interface element (e.g., a handle) on the instrument is calculated as F h and its projection in motor space is τ h The vector of setpoints, or measurements of motor forces (indirect measurements) for each actuator on the robot, is given by τ m Then, the two measured / controlled parameters F h and τ m The gravitational component seen from is assumed to have already been compensated for. The workspace forces on the tool and robot body (operator force F h (excluding F t and its projection on the same motor space is τ t and the torque of mechanical friction of the joint is τ f is set as

[0043] At equilibrium, the following mechanical relations can be written: τ m +τ t +τ h +τ f =0 (1)

[0044] The force increase gain is then g f τ is set as >1. ∈ is then defined as the error torque in motor space, satisfying the following relationship: τ ∈ =-(τ m +τ h ) / g f +τ h .

[0045] In the same way as the force control described earlier, closed-loop feedback control is then performed on the robot actuator setpoint to keep the error torque at zero. At equilibrium: 0=-(τ m +τ h ) / g f +τ h (2)

[0046] Substituting (1) into (2), we get the following:

[0047]

number

[0048] This equation shows that: On the one hand, the force multiplication objective is actually met: the forces exerted by the instrument on the workspace, excluding joint friction, do not actually correspond to the forces exerted by the operator on the collaborative processing interface multiplied by the force multiplication gain. - On the other hand, the instrument is not in contact (τ t = 0), the robot joint friction experienced by the operator is divided by this same gain.

[0049] Since there is no difference between the forces on the instrument and on the robot body, such control therefore offers the advantage of giving specific sensitivity on the robot body and increasing transparency at the collaborative processing interface.

[0050] The drawback with such a solution is the compromise that has to be reached regarding the force multiplication gain: this needs to be high enough to reduce the amount of friction experienced, but if it is too high, the sensitivity on the robot body is too low.

[0051] Another drawback is that the operator may interact with the robot body or the collaborative processing interface, but not both simultaneously. The robot's displacement then does not correspond to the resultant of the two interaction forces, something that is unintuitive and difficult for the operator to control.

[0052] There is therefore a need to improve the cooperative processing of existing industrial robots with mechanically insufficiently transparent or irreversible joints, in particular to alleviate the above drawbacks, and more particularly those of the force increasing control law. [Prior art documents] [Patent documents]

[0053] [Patent Document 1] International Publication No. 2014 / 161796 Brochure [Patent Document 2] International Publication No. 2016 / 000005 Brochure [Patent Document 3] International Publication No. 2010 / 097459 Brochure [Patent Document 4] U.S. Patent Application Publication No. US2015 / 0290809A1 [Patent Document 5] International Publication No. 2015 / 197333 Brochure Summary of the Invention [Problem to be solved by the invention]

[0054] It is an object of the present invention to at least partially meet this need. [Means for solving the problem]

[0055] To do this, one aspect of the present invention comprises: - a kinematic chain of mechanical elements with a proximal end element and a distal end element forming the base of the robot, the various elements being mounted with the ability to move relative to one another so that the distal end element is able to move with respect to the proximal end element; - an instrument and / or gripper intended to be operated by a human operator, the instrument and / or gripper being connected to the distal end element so as to have the same degrees of freedom as the distal end element; - means for controlling at least a part of a first chain of elements, - actuators arranged on the chain to effect all of the movement of and / or the application of forces between the various elements of the chain relative to one another; means for measuring the displacement of the elements relative to one another; Where appropriate, means for measuring the force exerted by the actuator; a single multi-axis force sensor positioned between the distal end element and the instrument and / or gripper to measure forces applied thereto; - controlling means comprising a controller for controlling the actuator based on measurements obtained by the means for measuring displacement and, where appropriate, the means for measuring the force applied by the actuator and measurements from the multi-axis force sensor, in accordance with a control law implemented in the controller; and a collaborative processing robot comprising: - a force multiplication loop configured to amplify at the robot joints forces applied by an operator to the tool and measured by the multi-axis force sensor, the measurements being for at least some of the degrees of freedom of the distal end, the force multiplication loop comprising a comparator for subtracting products of anti-windup gains Kaw from products of integral gains Ki of the loop, and an integrator receiving the results from the comparator to provide setpoint velocities for the various elements of the chain; an inner velocity loop with proportional gain Kv receiving a velocity setpoint from the force multiplication loop to provide a non-saturating reference torque for the various actuators; - saturation term τ sat is the vector τ of the dry friction coefficient of the actuator f0 an inner velocity loop saturation function selected to be greater than or equal to an anti-windup component obtained as the product of saturation and the force correction applied by the gain Kaw, which is fed back to the input of the integrator of the force multiplication loop so that the integrator of the force multiplication loop cuts off its integration as soon as saturation is established.

[0056] Relationship Kaw=Kv -1 is preferably set.

[0057] More preferably, the saturation term τ sat is the vector τ f0 plus twice the sum of the uncertainty values ​​above it.

[0058] The term "controller" is used herein and in the context of the present invention in its usual broad sense, i.e., a combination of hardware and software for programming and controlling a robot.

[0059] According to a first configuration, if the actuator can be controlled by a controller directly with a force, the saturation function is applied directly at the output of the inner velocity loop.

[0060] According to a second configuration, if the actuator cannot be directly controlled by force, but is controlled, for example, by a closed position or speed controller, the force τ exerted by the actuator m is then measured and considered in the saturation calculation.

[0061] Preferably, the means for measuring the displacement of the elements relative to one another comprise an absolute position sensor, or even an absolute multi-swivel sensor if placed directly at the output of the motor before the deceleration stage.

[0062] The controller may be configured to implement additional control laws selected from, for example, control with programmable virtual mechanical constraints, control with Cartesian or joint velocity limits, control with workspace constraints, teleoperated control with or without force feedback. For control with programmable virtual mechanical constraints, Cartesian or joint velocity limits, or workspace constraints, reference may be made to the teachings of US Pat. No. 6,239,999.

[0063] Teleoperated controls with or without force feedback may implement the principles described in

[11] or

[12] .

[0064] Thus, the present invention essentially consists of the intelligent positioning of a multi-axis force sensor between the end member (flange) and the holding implement of a collaborative processing industrial robot, and a modification of the force increase control law implemented in the robot's controller by adding a saturation function.

[0065] The present invention is therefore able to mitigate the drawbacks of the prior art force increasing laws, thereby allowing for increased sensitivity on the robot body and for human operator interaction with the robot body and cooperating handling interfaces such as tools supported by the robot's end members.

[0066] That is, the control according to the present invention combines the sensitive force measurements of the multi-axis sensors with the force measurements or set points for the actuators on the other hand, which ensures an improvement, albeit a lower sensitivity to physical interactions with the entire robot body

[10] , making it possible to displace the robot by directly manipulating the implement with greatly reduced forces (high level of transparency, such that the mechanical friction of the robot joints is hidden).

[0067] Therefore, an operator can easily manipulate an implement supported by the robot without worrying about the robot body exerting high forces if it inadvertently collides with an obstacle in the environment.

[0068] When this happens, the control respects the "natural" balance of forces; the opposing forces cancel each other out and the robot comes to a stop without exerting a greater force on the obstacle than the operator is applying to the implement.

[0069] The present invention provides a number of advantages in addition to collaborative processing capabilities on industrial robots, - a significant reduction in the joint mechanical friction experienced by a human operator when directly handling the instrument; - the sensitivity of the robot's body to the forces of interaction with the operator and / or the workspace, - a "natural" balance of forces that is respected in the case of multiple joint interactions between the operator, the workspace, the robot body and the implement, even in the presence of internal force components, - passivity of the interaction between the robot and its working environment as a result of the stability obtained in contact with any passive environment; - no changes are required to the mechanical architecture of the robot or its actuators, - there is no need to cover the robot body with touch-sensitive elements; - There are no restrictions on use in proximity to or through the mechanical unity of the robot; - the possibility of combining the control laws with other additional control laws useful for collaborative processing (virtual constraints, speed and workspace limitations, teleoperation, etc.); Examples include:

[0070] None of the solutions proposed in the prior art for controlling industrial collaborative processing robots is able to provide all of these advantages simultaneously.

[0071] Another subject of the invention is the use of an industrial collaborative handling robot such as that described above as a robot to assist in surgical interventions, or as a robot for assembling or handling heavy loads, or as a lead-through programming robot.

[0072] Further advantages and properties of the present invention will become better apparent upon reading the detailed description of exemplary embodiments of the invention, given by way of non-limiting example, and with reference to the following drawings, in which: [Brief explanation of the drawings]

[0073] [Figure 1] 1 is a schematic diagram of an example of an industrial collaborative processing robot with its controller utilized as a system for multiplying the force applied by a human operator to an implement carried by the robot. [Figure 2] FIG. 2 is a diagram summarizing all of the forces applied to the system of FIG. 1. [Figure 3] FIG. 2 illustrates a control law according to the present invention as implemented by the controller of the robot shown in FIG. 1. [Figure 4] FIG. 10 illustrates a modification of the control law according to the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0074] FIG. 1 shows an industrial collaborative processing robot 1 according to the invention, used as a system to assist in the manipulation of tools.

[0075] In the illustrated example, the instrument is a needle used by a surgeon for surgical intervention in an operating room that constitutes the working environment of the robot 1. In this illustrated example, the collaborative processing robot 1 is controlled by a mixed force control law, described in detail below, that allows the human operator (surgeon) to manipulate the instrument together with the robot, while at the same time remaining sensitive to interactions with the robot body and compensating for the weight and friction of the instrument in the robot joints.

[0076] This law is combined with an additional control law that imposes virtual guidance constraints on the implement (which will not be described in detail here in the context of the present invention).

[0077] The collaborative processing robot 1 is a robot having a manipulator arm with six degrees of freedom.

[0078] The robot 1 thus comprises a kinematic chain of interconnected elements, with a proximal end element 2 forming the base of the robot, and a distal end element 3 forming a flange, and in addition two elements 4, 5 or segments interconnected between the base 2 and the distal end element 3.

[0079] The robot 1 additionally comprises an instrument, which in the illustrated example is a needle 6, fixed to a handle 7 that is manipulated by a human operator. The needle 6 and the handle 7 are connected to the distal end element 3 (flange) so as to have the same degrees of freedom as that element, i.e. six degrees of freedom relative to the base 2.

[0080] Thus, the needle 6 can be moved in all directions in space when translating and rotating relative to the base 2 .

[0081] The operator may interact with the instruments 6, 7 and / or the body of the robot, in particular its elements 5, within the interaction area IZ.

[0082] The workspace may also interact with the tool 6 and / or the body of the robot, in particular with its elements 4, in case of intentional or unintentional (undesired) contact with the operator.

[0083] That is, there are two interaction ports installed directly on the robot, one via the robot body, in particular its elements 4, 5, and the other via the instruments 6, 7.

[0084] The robot further comprises means for controlling the chain of elements and therefore instruments 6, 7 connected to the distal end element 3.

[0085] The control means firstly comprise a controller 10 which executes a program for controlling the main chain of elements of the robot in order to maintain coordination between the robot body and the implements 6,7.

[0086] The control means also comprise actuators, not shown, arranged at each one of the joints of the chain of elements, so as to enable one of the elements to be moved relative to the adjacent element of the respective joint or to apply a force between these elements. The controller 10 controls the various actuators, as indicated diagrammatically by the arrows 11 in Figure 1, so as to enable the instruments 6, 7 to be moved in a coordinated manner relative to the base 2.

[0087] Furthermore, a multi-axis force sensor 8, preferably a six-axis sensor, is disposed between the distal end element 3 and the instruments 6, 7 so as to enable the controller 10 to generate signals indicative of the forces exerted by the instruments 6, 7 on the distal end element 3, as indicated by arrows 12 in Figure 1. The instruments 6, 7 are thus connected to the distal end element 3 via the multi-axis force sensor 8.

[0088] The control means further comprise means for measuring the displacement of the various elements, here a number of position sensors, not shown, each arranged at one of the joints of the main chain of elements, so as to enable the controller 10 to generate a signal indicative of the relative position of the two elements making up that joint, as indicated by the arrows 13 in Figure 1. The position sensors are absolute position sensors.

[0089] The absolute position sensor and multi-axis force sensor 8 thus enable the controller 10 to measure the movements of the instruments 6, 7 and of the robot body relative to the base 2 at any moment, and the forces applied to the instruments 6, 7 at any moment.

[0090] In accordance with the present invention, the controller 10 executes a program according to control laws, described in detail below, that allow increased sensitivity to forces applied to the instruments 6, 7 by an operator or their environment.

[0091] This control law may be combined with other additional control laws not described in detail herein, which in particular make it possible to apply virtual constraints 14 to the displacement of the instrument, in the case shown in Figure 1, so as to guide the insertion of the needle 6 into the body of patient B towards the area of ​​interest.

[0092] The load pattern for the force applied to the distal end element 3 of the robot as measured by the force sensor 8, minus the known weights of the instruments 6, 7 and reduced to the center point S of the sensor frame of reference, is shown as [Equation 2], with the components shown at the base of the sensor frame of reference S.

[0093]

number

[0094]

number

[0095] In equation (3), equation (4) is the resultant force and equation (5) is the torque at the measured force point S, which are expressed in the frame of reference S.

[0096]

number

[0097]

number

[0098] The Cartesian position of the frame of reference S, expressed in the frame of reference E associated with the end member of the robot, is X S,E ∈SE(3). Since the force sensor 8 is rigidly connected to the distal end element 3, X S,E is a constant and can be decomposed as [6], i.e., the rotation matrix relative to the frame of reference S in the frame of reference E, and [7], i.e., the original position of the frame of reference S shown in the frame of reference E.

[0099]

number

[0100]

number

[0101] This results in the following equation:

[0102]

number

[0103] Reduced to the center of the frame of reference E and shown therein is the loading pattern for the force exerted by the sensor 8 on the distal end element 3 of the robot.

[0104] The vector of robot joint position measurements is given by: where N is the number of joints.

[0105]

number

[0106] The vector of position measurements of the robot actuator is given by: where P≦N is the number of robot joints.

[0107]

number

[0108] At a given joint position q, the Cartesian position in the frame of reference E associated with the end member of the robot, shown in the frame of reference B of the robot base, is X = X E,B (q)∈SE(3). This position can be decomposed as

[11] , a rotation matrix for the rotation of the frame of reference E in the frame of reference B, and

[12] , the original position of the frame of reference E shown in the frame of reference B.

[0109]

number

[0110]

number

[0111] This results in the following equation:

[0112]

number

[0113] The load pattern applied to the end member of the robot by the sensor shown in the frame of reference B reduced to the center of the frame of reference E.

[0114] Jacobian matrix X in application examples E,B (q) is shown as [Equation 14], where [Equation 15] is the kinematic load pattern on the robot effector reduced to the center of the frame of reference E, shown in the frame of reference B, and can be decomposed as [Equation 17], i.e., the velocity of the origin of the frame of reference E shown in the frame of reference B, and [Equation 18], i.e., the rotational velocity vector relative to the frame of reference E shown in the frame of reference B.

[0115]

number

[0116]

number

[0117]

number

[0118]

number

[0119]

number

[0120] The matrix for the reduction ratio from the actuator space to the joint space is given as follows:

[0121]

number

[0122] This is then expressed as [Equation 20], i.e., the load pattern W for the force applied to the end member by the sensor. S into the actuator space, τ S =G T .J T .W S where,

[0123]

number

[0124] All of the forces applied to the system are then considered in actuator space and these are summarized in the diagram in Figure 2.

[0125] The force vectors applied by the operator and workspace to the robot body projected into actuator space are given as follows:

[0126]

number

[0127] The force vector produced by the actuator (measured or setpoint force by implementation) is given as:

[0128]

number

[0129] The joint friction force vector projected into actuator space is given as follows:

[0130]

number

[0131] The gravitational force vector projected into actuator space for a robotic instrument system is given as:

[0132]

number

[0133] The centrifugal and Coriolis force vectors for a robotic instrument system, projected into actuator space, are given as follows:

[0134]

number

[0135] The inertial matrix of the robotic instrument system projected into actuator space is given as follows:

[0136]

number

[0137] At equilibrium and low speed [Equation 27], the following relationship applies: τ s +τ b +τ g +τ f +τ m =0 (3)

[0138]

number

[0139] The mixing force control law according to the present invention is shown in the diagram of FIG.

[0140] The following loop is implemented in the controller 10: - Force increase loop 100, an inner speed loop 101 that receives speed set points from the force multiplication loop to provide set point torques to the various actuators; - inner velocity loop saturation function 102, - Forecast for compensation of Model 103.

[0141] The various interactions between the loops are described in detail below with reference to the robot actuator space.

[0142] The vector of the gravitational force model of the robotic instrument system projected into actuator space is given as:

[0143]

number

[0144] The vector of the joint friction force model projected into the actuator space is expressed as follows:

[0145]

number

[0146] An example of a friction model in actuator space only may be (30), where (31) denotes the vector of dry friction coefficients for the actuator, (32) denotes the nominal speed for the application of the model, and (33) denotes the damping in the actuator.

[0147]

number

[0148]

number

[0149]

number

[0150]

number

[0151] The following notation is introduced to simplify the connection with the force amplification control as described in

[10] .

[0152] τ h =τ s shows the projected forces into the actuator space of the interaction ports whose sensitivity increases, in our example, of the instruments attached to the force sensors. Equation 34 shows the projections of the forces into the actuator space of the other interaction ports whose sensitivity does not increase, in this example, all predictions of the external forces applied to the robot body.

[0153]

number

[0154] Therefore, the following equation is obtained: where g f >1 indicates the amplification factor of the force multiplication loop.

[0155]

number

[0156] [Number 36] is the force increaseis the integral gain of the loop. Ki is inversely proportional to the apparent inertia of the controlled system. The theoretical limit for setting this gain, and therefore the achievable apparent inertia of a robot controlled according to a passive criterion, i.e., the unconditional stability of a robot interacting with any passive environment, is approximately the mechanical inertia inherent to the robot: [4].

[0157]

number

[0158] The setpoint force vector derived from any additional control law not detailed here can be expressed as [Equation 37], which may be a virtual constraint law, a speed limit law, a workspace limiting law, or a remote control law.

[0159]

number

[0160] At equilibrium and outside the saturation range, the integrator input is zero and the force increase The nature of the loop is therefore actually [Equation 38] to Face it.

[0161]

number

[0162] By substituting (3) and (4) into (5), we get:

[0163]

number

[0164]

number

[0165] Therefore, the friction not compensated for by the model [Eq. 41] is in fact τ b =0, τ ref = 0, and it can be seen that the force is reduced by a force amplification factor when the operator is simply using the equipment, which corresponds to [Equation 42].

[0166]

number

[0167]

number

[0168] Equation 43 shows the proportional gain of the inner velocity loop 101. The purpose of the inner velocity loop is to linearize the system, especially for dry friction, so that the force increase The accumulation required in the loop integrator is reduced, improving friction rejection, especially at the point of change as a sign of joint velocity.

[0169]

number

[0170] Equation 44 is the saturation term for the velocity loop 101. The saturation function makes it possible to limit the contribution of the force amplification just in time with the reduction of dry friction, allowing for a "natural" balance of forces when there is joint interaction on the implement and on the body of the robot.

[0171]

number

[0172] To do this, according to the present invention, τ sat is the order τ f0 is chosen to be, and therefore |τ h |>>|τ f0 In |, control is always saturated at the equilibrium.

[0173] Equation (3) then becomes [Equation 45], where [Equation 46].

[0174]

number

[0175]

number

[0176] The "natural" balance of forces is thus in fact restored, τ h ≒-τ b This becomes:

[0177] Equation (47) shows the anti-windup gain of the integrator of the force multiplication loop 100, which is necessary to make the integrator stop integrating as soon as saturation occurs, and is advantageously chosen to be Equation (48).

[0178]

number

[0179]

number

[0180] Finally, [Equation 49] and [Equation 50] are the force increase By adjusting these, the force can be adjusted as specified in the publication

[10] . increase It becomes possible to optimize the loop stability and passband.

[0181]

number

[0182]

number

[0183] force set point τ m is the prediction period τ of [Eq. (51) and [Eq. (52)], which is any additional control law, i.e., modeling friction and gravity. ref is obtained by adding

[0184]

number

[0185]

number

[0186] The inventors have implemented the force control law with the saturation function described so far in industrial robot controllers from the Staubli range TX2_90 and TX2_60L. Collaborative performance has been demonstrated. Functionality requiring virtual constraints, speed limits, and limited workspaces has been successfully combined without loss of performance.

[0187] The modifications to the control law according to the invention are shown in the diagram of Figure 4. Such modifications are applied to actuators that cannot be directly controlled by forces. These may be, for example, hydraulic or pneumatic actuators fitted with servo valves.

[0188] Here, the force increase At the output of loop 100, the speed setpoint [Equation 53] is then given by: m It must have the ability to measure the

[0189]

number

[0190] For hydraulic actuators, the set point may be that of servo-controlled hydraulic flow at measured actuator pressure.

[0191] As shown diagrammatically in Figure 4, the force τ m The measurements of τ are therefore taken on the robot body on the one hand, and on the saturation and anti-windup τ aw In the calculations, it is used to calculate the force estimate [Equation 54].

[0192]

number

[0193] The invention is not limited to the examples described so far, and features from the examples shown may in particular be combined with one another in variants not shown.

[0194] Other variations and modifications may be envisaged without departing in any way from the scope of the present invention.

[0195] The actuator may advantageously comprise a servomotor. Typically, the actuator may comprise an ironless DC motor, a brushless motor, a conventional DC motor, a shape memory alloy, a piezoelectric actuator, an activated polymer, a pneumatic or hydraulic actuator. The actuator may have brakes on one or more elements of the robot body. These brakes may thus be disc brakes, powder brakes, or magnetorheological or electrorheological fluid brakes. The actuator may also comprise a hydraulic actuator equipped with both a motor and a brake or a reverse actuation device and / or a variable stiffness device. If the actuator comprises, for example, a reduction gear associated with the motor, the reduction gear may be of any type, such as a simple gearing or planetary gearing reduction motor, or, in one or more stages, a "Harmonic Drive" (registered trademark) type reduction gear, a ball screw reduction gear, or a cable winch reduction gear. Instead of a reversible reduction gear, it is possible to have a non-reversible reduction gear, such as a worm and wheel reduction gear.

[0196] (References) [1] B. Rooks, “The harmonious robot”, Industrial Robot: An International Journal, vol. 33, n° %12, pp. 125-130, 2006. [2] P. Hamon, M. Gautier and P. Garrec, “New dry friction model with load- and velocity-dependence and dynamic identification of multi-DOF robots”, IEEE International Conference on Robotics and Automation, pp. 1077-1084, 2011. [3] W. S. Newman and Y. Zhang, “Stable interaction control and coulomb friction compensation using natural admittance control”, Journal of Robotic Systems, vol. 11, n° %11, pp. 3-11, 1994. [4] W. S. Newman, “Stability and performance limits of interaction controllers”, Journal of Dynamic Systems, Measurement, and Control, vol. 114, n° %14, pp. 563-570, 1992. [5] J. E. Colgate, “The Control of Dynamically interacting Systems”, PhD Thesis, Massachusetts Institute of Technology, 1988. [6] N. Hogan, “Controlling impedance at the man / machine interface”, IEEE International Conference on Robotics and Automation, Proceedings, vol. 3, pp. 1626-1631, 1989. [7] F. Geffard et al., “On the use of a base force / torque sensor in teleoperation”, Proceedings 2000 ICRA. Millennium Conference. IEEE International Conference on Robotics and Automation. Symposia Proceedings, vol. 3, pp. 2677-2683, 2000. [8] A. Albu-Schaeffer and C. Ott, “A Unified Passivity Based Control Framework for Position, Torque and Impedance Control of Flexible Joint Robots”, The International Journal of Robotics Research, vol. 26, 2007. [9] R. B. et al., “The KUKA-DLR Lightweight Robot arm - a new reference platform for robotics research and manufacturing”, ISR 2010 (41st International Symposium on Robotics) and ROBOTIK 2010 (6th German Conference on Robotics), pp. 1-8, 2010.

[10] X. Lamy et al., “Human force amplification with industrial robot: study of dynamic limitations”, IEEE / RSJ International Conference on Intelligent Robots and Systems (IROS)., 2010.

[11] K. Hashtrudi-Zaad and S. E. Salcudean, “Analysis of Control Architectures for Teleoperation Systems with Impedance / Admittance Master and Slave Manipulators”, The International Journal of Robotic Research, vol. 20, n° %16, pp. 419-445, 2001.

[12] A. Micaelli, “Teleoperation et telerobotique, chapitre 6, asservissement et lois de couplage en teleoperation [Remote operation and remote robotics, chapter 6, feedback control and coupling laws for remote operation]”, Hermes science, 2002. [Explanation of symbols]

[0197] 1. Industrial collaborative processing robot 2 Near-end elements 3 Distal end element 4 Elements 5 Elements 6 needles 7 Handle 8 Multi-axis force sensor 10 Controller 11 Arrow 12 Arrows 13 Arrow 100 Power Increase Loop 101 Inner Speed ​​Loop 102 Inner velocity loop saturation function 103 model

Claims

1. a kinematic chain of mechanical elements (2 to 5) with a proximal end element forming the base (2) of the robot, and a distal end element (3), the mechanical elements of said kinematic chain being mounted with the ability to move relative to each other such that the distal end element can move with respect to the proximal end element; an instrument (6, 7) and / or gripper intended to be operated by a human operator, said instrument (6, 7) and / or gripper being connected to said distal end element so as to have the same degrees of freedom as said distal end element; means for controlling at least a part of the kinematic chain of said mechanical element, - actuators arranged on the kinematic chain of mechanical elements to effect all of the movements of and / or apply forces between the mechanical elements of the kinematic chain relative to one another; means for measuring the displacement of said mechanical elements relative to one another; Where appropriate, means for measuring the force exerted by the actuator; a single multi-axis force sensor (8) arranged between the distal end element and the instrument and / or the gripper to measure the force applied to the distal end element and / or the instrument and / or the gripper; - a controller (10) for controlling the actuator based on measurements obtained by the means for measuring displacement and, where appropriate, the means for measuring the force exerted by the actuator and measurements from the multi-axis force sensor, according to a control law implemented in the controller (10); A collaborative processing robot (1) comprising: The control law is: a force multiplication loop (100) configured to amplify at a robot joint a force applied to the tool or the gripper and measured by the multi-axis force sensor, the measurement being for at least some of the degrees of freedom of the distal end element, and an anti-windup gain K ａｗ a force multiplication loop (100) comprising a comparator for subtracting the product of Kj from the product of an integral gain Ki of said force multiplication loop, and an integrator receiving the result from said comparator to provide a set point velocity for said mechanical element of said kinematic chain; an inner velocity loop (101) with a proportional gain Kv receiving the velocity set point from the force multiplication loop to provide a non-saturating reference torque to an actuator located on the kinematic chain of the mechanical element; Saturation term τ ｓａｔ is the vector τ of the dry friction coefficient of the actuator ｆ０ an inner velocity loop saturation function (102) selected to be greater than or equal to The inner velocity loop saturation function and the anti-windup gain K are fed back to the input of the integrator of the force multiplication loop so that the integrator of the force multiplication loop cuts off its integration as soon as saturation occurs. ａｗ and an anti-windup component obtained as the product of the force correction applied by A collaborative processing robot (1) comprising:

2. The anti-windup gain K ａｗ The product of these is Kv －１ The collaborative processing robot of claim 1 , wherein

3. The saturation term τ ｓａｔ is the vector τ ｆ０ 2. The collaborative processing robot of claim 1, wherein the uncertainty is equal to the sum of the uncertainty values ​​above plus two times the uncertainty value above.

4. The collaborative processing robot of claim 1 , wherein the actuator is directly controllable by the controller with a force, and the inner velocity loop saturation function is applied directly at the output of the inner velocity loop.

5. The actuator cannot be directly controlled by a force, and the force τ applied by the actuator ｍ is then measured and considered to select the saturation term of the inner velocity loop saturation function.

6. The collaborative processing robot of claim 1 , wherein the actuator comprises a servo motor.

7. 10. The collaborative processing robot of claim 1, wherein the means for measuring the displacement of the mechanical elements relative to one another comprises absolute position sensors.

8. 10. The collaborative processing robot of claim 1, wherein the controller is configured to implement at least one additional control law selected from control with programmable virtual mechanical constraints, control with Cartesian and / or joint velocity limits, control with workspace constraints, and teleoperated control with or without force feedback.

9. 2. The collaborative processing robot of claim 1, wherein the multi-axis force sensor (8) is positioned between the handle (7) of the tool and the tool (6) so as to measure only the force applied to the handle.

10. 10. Use of an industrial collaborative handling robot according to claim 1 as a robot for assisting surgical interventions, or as a robot for assembling or handling heavy loads, or as a lead-through programming robot.

Images