Force control

A robot hand grabs a delicate object without crushing it.

Force control refers to the control of the force with which a machine or the manipulator of a robot acts on an object or its environment. By regulating the contact force, damage to the machine and the objects to be processed and injuries when handling people can be prevented. In production tasks, it can compensate for errors and reduce wear through a uniform contact force. The force control achieves more uniform results than the position control that is also used in the control of machines . Force control can be used as an alternative to the usual motion control, but is usually used in addition, in the form of hybrid control concepts. For the regulation, the acting force is usually measured using force transducers or estimated using the motor current.

Force control has been the subject of research for almost three decades and is increasingly opening up further areas of application thanks to advances in sensors and actuators as well as new control concepts . Force control is particularly suitable for contact tasks that serve the mechanical processing of workpieces, but is also used in telemedicine , service robotics and the scanning of surfaces.

Force sensors are used to measure force and can measure forces and torques in all three spatial directions. Alternatively, the forces can also be sensorless z. B. can be estimated based on the motor currents. The control concepts used are indirect force control by modeling the robot as a mechanical resistance (impedance) and direct force control in parallel or hybrid concepts. Adaptive approaches , fuzzy controllers and machine learning for force control are currently the subject of research.

General

Error in motion control (red) in contrast to force control (green).

The control of the contact force between a manipulator and its environment is an increasingly important task in the field of mechanical production, as well as industrial and service robotics . One motivation for using force control is safety for people and machines. For various reasons, the movements of the robot or machine parts can be blocked by obstacles while the program is running. In service robotics, these can be moving objects or people; in industrial robotics, problems can arise with cooperating robots, changing work environments or an imprecise environmental model. If the trajectory is misaligned with classic motion control and the programmed robot pose (s) cannot be approached , the motion control will increase the manipulated variable - usually the motor current - in order to correct the position error. Increasing the manipulated variable can have the following effects:

1. The obstacle is removed or damaged / destroyed.
2. The machine will be damaged or destroyed.
3. The manipulated variable restrictions are exceeded and the robot controller switches off.

A force control can prevent this by regulating the maximum force of the machine in these cases and thus avoiding damage or making collisions detectable at an early stage.

In mechanical production tasks, unevenness in the workpiece often leads to problems with movement control. As can be seen in the adjacent figure, unevenness in the surface leads to the tool penetrating too far into the surface ( ) or losing contact with the workpiece ( ) during position control (red ). As a result, for example, when grinding and polishing, the force acting on the workpiece and tool changes. A force control (green) makes sense here, as this ensures even material removal through constant contact with the workpiece. ${\displaystyle P'_{1}}$ ${\displaystyle P'_{2}}$

application

In the case of force control, a distinction can be made between applications with pronounced and applications with potential contact . From a pronounced contact occurs when the contact of the machine is a central part of the task with the environment or the workpiece and is explicitly regulated. These include, above all, mechanical deformation and surface processing tasks. For tasks with potential contact, the essential process variable is the positioning of the machine or its parts. Greater contact forces between the machine and the environment result from a dynamic environment or an imprecise environmental model. In this case, the machine should yield to the environment and avoid large contact forces.

Industrial robot bending sheet metal on the bending machine

The main applications of force control today are mechanical manufacturing work. This means in particular manufacturing tasks such as grinding , polishing and deburring as well as force-controlled processes such as the controlled joining, bending and pressing of bolts into prefabricated holes. Another common use of force control is scanning unknown surfaces. A constant contact pressure is set in the normal direction of the surface via the force control and the scanning head is moved in the surface direction via position control. The surface can then be described in Cartesian coordinates using direct kinematics .

Further applications of force control with potential contact can be found in medical technology and in cooperating robots. Robots that are used in telemedicine , i.e. robot-assisted medical operations, can prevent injuries more effectively using force control. In addition, the direct feedback of the measured contact forces to the operator by means of a force feedback operating device is of great interest here. Possible uses for this range up to internet-based teleoperations.

In principle, force control is also useful wherever machines and robots cooperate with each other or with people, as well as in environments in which the environment is not exactly described or is dynamic and cannot be precisely described. There, force control helps to be able to respond to obstacles and deviations in the environmental model and to avoid damage.

history

The first significant work on force control was published in 1980 by John Kenneth Salisbury at Stanford University . It describes a method for active stiffness control, a simple form of impedance control. However, the method does not yet allow a combination with a movement control, instead a force control takes place in all spatial directions. The position of the surface must therefore be known. Due to the poor performance of the robot controls at that time, the force control could only be carried out on mainframes. This achieved a controller cycle of ~ 100 ms.

In 1981 Raibert and Craig presented a work on hybrid force / position control that is still significant today. They describe a process in which, with the help of a matrix (separation matrix ), it is explicitly specified for all spatial directions whether a movement or force control is used. Raibert and Craig merely sketch the controller concepts and assume that they are feasible.

In 1989 Koivo introduced an expanded representation of the concepts of Raibert and Craig. Exact knowledge of the surface position is still necessary here as well, which is what the typical tasks of force control today, such as B. scanning surfaces, is still not allowed.

Force control has been the subject of intensive research over the past two decades and has made great progress thanks to the further development of sensors and control algorithms. For a number of years now, the large automation technology manufacturers have been offering software and hardware packages for their controls to allow force control. Modern machine controls are capable of real-time force control in one spatial direction with a cycle time of less than 10 ms.

Force measurement

→ Main article : Force measurement

In order to close the force control loop in terms of regulation , the instantaneous value of the contact force must be known. The contact force can either be measured directly or estimated.

Direct force measurement

Foil strain gauges

The trivial approach to force control is the direct measurement of the contact forces that occur using force / torque sensors on the end effector of the machine or on the wrist of the industrial robot. Force / torque sensors measure the forces that occur by measuring the deformation on the sensor. The most common way of measuring deformations is measurement using strain gauges .

In addition to the widespread strain gauges made from variable electrical resistances, there are also other designs that use piezoelectric , optical or capacitive principles for measurement. In practice, however, they are only used for special applications. For example, capacitive strain gauges can also be used in the high temperature range above 1000 ° C.

Strain gauges are designed in such a way that they have as linear a relationship as possible between expansion and electrical resistance within the working space. There are also several ways to reduce measurement errors and interference. In order to exclude temperature influences and to increase the measuring reliability, two strain gauges can be arranged complementarily.

Modern force / torque sensors measure both forces and torques in all three spatial directions and are available with almost any value range. The accuracy is usually in the per mille range of the maximum measured value. The sampling rates of the sensors are in the range of around 1 kHz. An extension of the 6-axis force / torque sensors are 12- and 18-axis sensors which, in addition to the six force or torque components, are also able to measure six speed and acceleration components.

Six-axis force / torque sensor

A force / torque sensor with measurement in three force and three torque components.

So-called six-axis force / torque sensors are often used in modern applications. These are mounted between the robot hand and the end effector and can record both forces and torques in all three spatial directions. For this purpose, they are equipped with six or more strain gauges (or strain gauges) that detect deformations in the micrometer range. These deformations are converted into three force and torque components using a calibration matrix.

Force / torque sensors contain a digital signal processor, which constantly records and filters the sensor data (strain) in parallel, calculates the measurement data (forces / moments) and makes it available via the communication interface of the sensor.

It should be noted that the measured values ​​correspond to the forces on the sensor and usually have to be converted into the forces and torques on the end effector or tool using a suitable transformation.

Since force / torque sensors are still relatively expensive (between € 4,000 and € 15,000) and very sensitive to overload and malfunctions, they - and thus also the force control - have so far been used hesitantly in industry. One solution is the indirect force measurement or estimation, which enables force control without costly and failure-prone force sensors.

Force estimation

A cost-saving alternative to direct force measurement is the estimation of the force (also known as “indirect force measurement”). This makes it possible to dispense with the use of force / torque sensors. In addition to cost savings, doing without them brings other advantages: Force sensors are usually the weakest link in the mechanical chain of the machine or the robot system, so doing without them brings greater stability and less susceptibility to mechanical failure. In addition, the elimination of force / torque sensors brings with it a higher level of safety, since no sensor cables need to be led out and protected directly on the wrist of the manipulator.

A common method for indirect force measurement or force estimation is the measurement of the motor currents that are applied to control movement. These are, with some restrictions, proportional to the torque applied on the driven robot axis. Adjusted for the effects of gravitation, inertia and friction, the motor currents are largely linear to the torques of the individual axes. The contact force at the end effector can be determined from the torques known with it.

Separate dynamic and static forces

When measuring and estimating the force, it may be necessary to filter the sensor signals. Numerous side effects and side forces can occur that do not correspond to the measurement of the contact force. This is especially true if a larger load mass is mounted on the manipulator. This disrupts the force measurement when the manipulator moves with high accelerations.

In order to be able to adjust the measurement for side effects, both an exact dynamic model of the machine and a model or an estimate of the load must be available. This estimate can be determined using reference movements (free movement without object contact). After estimating the load, the measurement or estimation of the forces can be adjusted for Coriolis , centripetal and centrifugal forces , gravitational and frictional effects and inertia . Adaptive approaches can also be used here in order to continuously adapt the estimate of the load.

Control concepts

Various control concepts are used for force control. Depending on the desired behavior of the system, a distinction is made between concepts of direct force control and indirect control by specifying the compliance or mechanical impedance. As a rule, force control is combined with motion control. Force control concepts must take into account the problem of coupling between force and position: If the manipulator is in contact with the environment, a change in position also means a change in contact force.

Impedance control

The impedance control or compliance control regulates the compliance of the system, i.e. the link between force and position in the event of object contact. Compliance is defined in the specialist literature as the "measure of the robot's ability to counteract the contact forces". There are passive and active approaches to this. The flexibility of the robot system is modeled as mechanical impedance, which describes the relationship between the force applied and the resulting speed. The machine or the manipulator of the robot is viewed as a mechanical resistance with position restrictions due to the environment. The causality of the mechanical impedance therefore describes that a movement of the robot results in a force. With mechanical admittance, on the other hand, a force exerted on the robot leads to a resulting movement.

Passive impedance control

Remote Center of Compliance yields to rotational and translational deviations during an insertion process.

No force measurement is required for passive compliance control , as there is no explicit force control. Instead, the manipulator and / or end effector is designed to be flexible in a way that can minimize contact forces occurring during the task to be performed. Typical applications are inserting and gripping processes. The end effector is designed in such a way that it allows translational and rotational deviations orthogonally to the gripping or insertion direction, but has a high degree of rigidity in the gripping or insertion direction. The figure opposite shows a so-called Remote Center of Compliance (RCC) that makes this possible. As an alternative to an RCC, the entire machine can also be designed to be structurally elastic.

Passive impedance control is a very good solution with regard to the system dynamics , since there are no dead times due to the control. Passive compliance control is, however, often limited in the task by the mechanical specification of the end effector and cannot be used without further ado for different and changing tasks or environmental conditions.

Active impedance control

Active compliance control describes the control of the manipulator due to a deviation of the end effector. This is particularly suitable for guiding robots by an operator, for example as part of a teach-in process.

The active compliance control is based on the idea of ​​mapping the system of machine and environment as a spring-damper-mass system. The force that occurs and the movement (position , speed and acceleration ) are directly related to the spring-damper-mass equation: ${\displaystyle F}$ ${\displaystyle x(t)\!\,}$ ${\displaystyle {\dot {x}}(t)}$ ${\displaystyle {\ddot {x}}(t)}$

${\displaystyle F(t)=c\cdot x(t)+d\cdot {\dot {x}}(t)+m\cdot {\ddot {x}}(t)}$

The flexibility or mechanical impedance of the system is determined by the rigidity , the damping and the inertia and can be influenced by these three variables. A mechanical target impedance , which is achieved by the machine control, is specified for the control via these three variables . ${\displaystyle c}$ ${\displaystyle d}$ ${\displaystyle m}$

Block diagram of the active impedance control with specification of the force ( ) and the position ( ). ${\displaystyle F_{soll}}$ ${\displaystyle X_{soll}}$

The figure shows the block diagram of a force-based impedance control. The impedance in the block diagram represents the components mentioned , and . A position-based impedance control can be designed analogously with internal position or movement control. ${\displaystyle c}$ ${\displaystyle d}$ ${\displaystyle m}$

Alternatively and analogously, the flexibility ( admittance ) can also be regulated instead of the resistance . In contrast to the impedance control, the admittance appears in the control law as the reciprocal of the impedance.

Direct force control

The concepts mentioned above involve what is known as indirect force control, since the contact force is not explicitly specified as a reference variable, but is determined indirectly via the controller parameters damping , stiffness and (virtual) mass . Direct force control is presented below .

Direct force control uses the desired force as a setpoint within a closed control loop . It is implemented as a parallel force / position control in the form of a cascade control or as a hybrid force / position control in which a switch is made between position and force control.

Parallel force / position control

One possibility of force control is the parallel force / position control. The control is designed as a cascade control and has an outer force control loop and an inner position control loop. As shown in the following figure, a corresponding infeed correction is calculated from the difference between the target and actual force. This infeed correction is offset against the nominal position values, with the merging of and the position specification of the force control ( ) having a higher priority, i.e. a position error is tolerated in favor of the correct force control. The calculated value is the input variable for the inner position control loop. ${\displaystyle X_{soll}}$ ${\displaystyle X_{korr}}$ ${\displaystyle X_{korr}}$

Block diagram of the parallel force / position control with specification of the force ( ) and the position ( ). ${\displaystyle F_{soll}}$ ${\displaystyle X_{soll}}$

Analogous to an internal position control, an internal speed control can also take place, which has a higher dynamic. It should be noted that in this case the inner control loop should be saturated, so as not to generate a (theoretically) arbitrarily increasing speed in the free movement until contact is made.

Hybrid force / position control

The hybrid force / position control, which works with two separate control systems and can also be used with hard, inflexible contact surfaces, offers an improvement over the concepts explained above. With hybrid force / position control, the space is divided into a restricted ( English : constrained ) and an unrestricted (English: unconstrained ) space. The restricted space contains restrictions in the form of obstacles and does not allow free movement; the unrestricted space allows free movement. Each dimension of space is either limited or unlimited.

Block diagram of the hybrid force / position control with separation matrix Σ .

With hybrid force control, force control is used for the restricted space, position control is used for the unrestricted space. The figure shows such a scheme. The matrix Σ indicates which spatial directions are restricted and is a diagonal matrix consisting of zeros and ones.

Which spatial direction is restricted and which is unrestricted can be statically specified, for example. Force and position control is then explicitly specified for each spatial direction; the matrix Σ is then static. Another possibility is to switch the matrix Σ dynamically using force measurement . When contact is made or a collision occurs, it is possible to switch from position control to force control for individual spatial directions. In the case of contact tasks, all spatial directions would be movement-controlled in the case of free movement; after contact was established, a switch would be made to force control in the contact direction by selecting the appropriate matrix Σ .

research

Research in recent years has increasingly focused on adaptive concepts, the use of fuzzy controllers and machine learning, as well as force-based whole-body control.

Adaptive force control

The previously mentioned, non-adaptive concepts are based on an exact knowledge of the dynamic process parameters. These are usually determined and adjusted through experiments and calibration. Problems can arise from measurement errors and variable loads. In adaptive power control position-dependent and thus time-varying parts of the system to be construed as a parameter fluctuations, and in the course of the control by adaptation adjusted constant.

It should be noted that due to the changing regulation no guarantee for dynamic stability of the system can be given. Adaptive control is therefore usually only used offline and the results are intensively tested in the simulation before use on the real system.

Fuzzy control and machine learning

An explicit system model is a prerequisite for the use of classic design methods. If this cannot be mapped or is difficult to map, fuzzy controllers or machine learning can be considered. By fuzzy logic acquired human knowledge can be translated into a control action in the form of fuzzy rule requirements. An explicit specification of the controller parameters is no longer necessary.

In addition, approaches with the help of machine learning no longer require humans to create the control behavior, but use machine learning as the basis for the control.

Whole body regulation

Due to the high complexity of modern robot systems, such as humanoid robots , a large number of actuated degrees of freedom must be regulated. In addition, such systems are increasingly used in the immediate human environment. Accordingly, concepts from force and impedance control are used in this area in a targeted manner to increase safety, as this enables flexible interaction of the robot with the environment and humans.

literature

• Bruno Siciliano , Luigi Villani: Robot Force Control . Springer, 2000, ISBN 0-7923-7733-8 .
• Wolfgang Weber: Industrial robots. Methods of control and regulation . Fachbuchverlag Leipzig, 2002, ISBN 3-446-21604-9 .
• Lorenzo Sciavicco , Bruno Siciliano: Modeling and Control of Robot Manipulators . Springer, 1999, ISBN 1-85233-221-2 .
• Klaus Richter: Force control of elastic robots . VDI-Verlag, 1991, ISBN 3-18-145908-9 .

Individual evidence

1. ↑ ^{a b} Vadym Rusin: Adaptive control of robot systems in contact tasks. ( Memento from January 9, 2016 in the Internet Archive ) (PDF; 4.5 MB). Otto von Guericke University Magdeburg, 2007.
2. ↑ John Kenneth Salisbury: Active Stiffness Control of a Manipulator in Cartesian Coordinates. 19th IEEE Conference on Decision and Control, December 1980.
3. ↑ ^{a b c d} Marcus Dapper: Force sensorless manipulator Force control for scanning unknown, hard surfaces. Rheinische Friedrich-Wilhelms-Universität Bonn, November 2003.
4. ^ MH Raibert, John Craig: Hybrid Position / Force Control of Manipulators. ASME Journal of Dynamic Systems, Measurement and Control, June 1981.
5. ↑ AJ Koivo: Fundamentals for Control of Robotic Manipulator. Wiley & Sons, New York, USA 1989.
6. ↑ Malik Cabaravdic: Contribution to optimizing the chip volume in industrial robot-assisted belt grinding of freely shaped workpieces. P. 110, Technical University of Dortmund, February 2008.
7. ↑ Eko Bono Suprijadi: Kinematic real-time control and power control of a four-legged walking machine based on simplified kinematics. University of Duisburg-Essen, May 2005.
8. ^ John Simpson, Zheng Li, Chris Cook: Sensorless Force Estimation for Robots with Friction. (PDF; 731 kB) November 2002.
9. ↑ D. Colombo, D. Dallefrate, L. Molinari Tosatti: PC Based Control Systems for Compliance Control and Intuitive Programming of Industrial Robots. (PDF; 816 kB) In: Proceedings of the Joint Conference on Robotics. May 2006.
10. ↑ Alexander Winkler: A contribution to force-based human-robot interaction.  ( Page no longer available , search in web archives )  Info: The link was automatically marked as defective. Please check the link according to the instructions and then remove this notice. (PDF; 6.3 MB) Chemnitz University of Technology, 2006.  
11. ↑ Lorenzo Sciavicco, Bruno Siciliano: Modeling and Control of Robot Manipulators. 2nd edition, Springer Verlag, 1999, ISBN 1-85233-221-2 .
12. ↑ Alexander Dietrich : Whole-Body Impedance Control of Wheeled Humanoid Robots , ISBN 978-3-319-40556-8 , Springer International Publishing, 2016.

This article was added to the list of excellent articles on September 19, 2009 in this version .