DE102016219128B4 - Method for performing wiping movements by a robotic arm - Google Patents

Abstract

A method for carrying out wiping movements by a robot arm (12), the method comprising the following steps: carrying out a wiping movement on a surface (14) to be processed by the robot arm (12), detecting a contact force between the end effector (16) of the Robot arm (12) and the surface to be processed (14) occurs during the wiping movement, creating a model of the effect of the wiping movement by the following steps: Defining a lower limit for the contact force, above which it is assumed that there is contact between the end effector (16) and the surface (14) to be machined, wherein portions of the surface (14) to be machined, in which the lower limit for the measured contact force is exceeded, are used to determine the orientation and / or the extent of the surface (14) to be machined determine, the position of the end effector (16) in the normal direction to the to be processed during the wiping movement en surface (14) is detected, in the sections of the surface to be processed (14) in which the contact force is above the lower limit, it is assumed that the wiping movement had the desired effect, provided that the position of the end effector (16) was in the normal direction lies within a defined range relative to the surface (14) to be machined, wherein if the measured contact force is in the desired range, i.e. above the lower limit for the contact force, the position of the end effector (16), however, does not lie within the defined range, it is assumed that the swiping movement was not performed properly.

Description

The invention relates to a method for performing wiping movements by a robot arm.

Autonomous robots are increasingly used in household scenarios, for professional service, e.g. in supermarkets, in public buildings (museums, schools, or community centers), in hospitals or in elderly care and in industrial production to treat everyday objects. Wiping movements are a fundamental skill for a variety of tasks, such as cleaning rooms, corridors, furniture, household objects or exhibits, washing people in need of care, and processing objects in manufacturing technology, such as polishing or Grinding occurs. The autonomous execution of the wiping movements must meet the high quality requirements of the various domains.

The prior art examines wiping movements mainly in the area of automatic cleaning. The planning and execution of wiping movements is predominantly examined. For this purpose, flexible manipulators are often used (such as the DLR / KUKA lightweight robot III) that adapt to the environment. These robots use torque sensors in the robot joints or force sensors on the end effector to measure and control forces.

US 8065127 B2 describes the use of particle simulation to simulate non-contact painting by a robot. The procedure does not describe a method for analyzing the quality of the work carried out.

DE 102 30 021 C1 describes a method for cleaning a component using a robot arm. A force sensor is used to detect contact between the cleaning head and the workpiece.

The methods described in the prior art for autonomous wiping with robots mainly deal with the planning and execution of wiping movements in the context of cleaning tasks. The disadvantage here is that the quality of the wiping movements (e.g. the degree of the expected cleaning) is only detected visually, if at all, and can therefore only be used with clearly visible media. The approaches described are accordingly difficult to apply in poor lighting conditions, small dust particles or transparent liquids. In addition, robots often cover large parts of the work area, which disrupts image processing algorithms.

So far, no models are known for qualitatively evaluating the result of a wiping movement carried out by a robot.

The object of the invention is to provide a method for performing wiping movements by a robot, by means of which the quality of the wiping movement can be assessed.

The object is achieved according to the invention by the features of claim 1.

• Es wird eine Wischbewegung auf einer zu bearbeitenden Oberfläche durch den Roboterarm durchgeführt. Bevorzugt handelt es sich hierbei um mindestens eine ebene Oberfläche. Zum Durchführen der Wischbewegung ist es bevorzugt, dass am Endeffektor des Roboters ein Wischwerkzeug angebracht ist. Hierbei kann es sich beispielsweise um einen Schwamm, eine Bürste oder ein anderes Werkzeug handeln, durch das Wischbewegung durchgeführt wird. Das zu verwendende Wischwerkzeug wird maßgeblich von dem Material und der Struktur der zu bearbeitenden Oberfläche und ferner von dem zu wischenden Medium abhängen (sofern denn ein solches existiert).

The method according to the invention for performing wiping movements by a robot arm comprises the following steps:

• A wiping movement is carried out on a surface to be processed by the robot arm. This is preferably at least one flat surface. To carry out the wiping movement, it is preferred that a wiping tool is attached to the end effector of the robot. This can be, for example, a sponge, a brush or some other tool by means of which a wiping movement is carried out. The wiping tool to be used will largely depend on the material and structure of the surface to be processed and also on the medium to be wiped (if one exists).

According to the invention, a contact force that occurs between the end effector of the robot arm or the wiping tool and the surface to be processed during the wiping movement is detected.

• Es wird eine Untergrenze für die Kontaktkraft definiert, oberhalb derer angenommen wird, dass ein Kontakt zwischen dem Endeffektor und der zu bearbeitenden Oberfläche besteht. Anders ausgedrückt wird angenommen, dass der Endeffektor bzw. das Wischwerkzeug Kontakt zu der zu bearbeitenden Oberfläche hat, wenn die gemessene Kontaktkraft oberhalb der genannten Untergrenze liegt. Beispielsweise kann angenommen werden, dass ein Kontakt vorliegt, wenn die gemessene Kontaktkraft größer als 90% der Zielkontaktkraft ist. Bei der Zielkontaktkraft kann es sich beispielsweise um die maximale Kraft handeln, die der Roboterarm aufbringen kann. Diese kann beispielsweise durch die Steuerung des Roboters limitiert werden.

Model the effect of the swiping motion by doing the following:

• A lower limit is defined for the contact force, above which it is assumed that there is contact between the end effector and the surface to be processed. In other words, it is assumed that the end effector or the wiping tool is in contact with the surface to be processed when the measured contact force is above the mentioned lower limit. For example, it can be assumed that there is contact if the measured contact force is greater than 90% of the target contact force. The target contact force can be, for example, the maximum force that the robot arm can apply. This can be limited, for example, by controlling the robot.

Sections on the surface to be processed, in which the lower limit for the measured contact force is exceeded, are used to determine the orientation and / or extent of the to be machined surface. Only these sections of the surface to be processed are preferably used, while other sections in which the lower limit for the measured contact force is not exceeded are not taken into account.

During the wiping movement, the position of the end effector is recorded in the normal direction to the surface to be processed. In the sections of the surface to be processed in which the contact force is above the lower limit, it is assumed that the wiping movement had the desired effect, provided that the position of the end effector in the normal direction relative to the surface to be processed is within the previously automatically determined orientation and extent of the Surface lies.

In other words, it is assumed that the wiping movement has been carried out properly when the contact force is in a certain range and also when the position of the end effector in the normal direction to the surface to be processed at the time when the contact force is in the desired range is also in a desired area. If this is not the case, only the contact force would be in the desired range, but not the position of the end effector. This could happen, for example, if the robot is not correctly aligned relative to the surface to be processed or if the robot arm has been disturbed by an obstacle, for example a person. In this case, the contact force could be in the target area (for example 95%), but at this point in time the end effector would not be in an area that corresponds to the extent of the surface to be processed. In this case, it can be assumed that the swiping movement was not performed properly in this section. Accordingly, the expected effect does not occur. In principle, it is also possible that the desired wiping movement has been carried out properly, even if the contact force is not in the desired range (i.e. below the lower limit). In this case, however, the probability that the wiping movement was carried out correctly is lower, so that a correct wiping movement is preferably only assumed if the lower limit for the contact force has been exceeded.

It is preferred that during the automatic determination of the alignment and / or extent of the surface to be processed, the extent of the surface in the X, Y and Z directions is automatically determined. This height in the Z direction can then preferably be taken into account for the simulation of the wiping effect. In other words, a high contact force, which is measured outside the determined height in the Z-direction of the surface to be processed and which exceeds the required lower limit for the contact force, would not be interpreted as a successful wiping movement because it is outside the determined spatial extent of the surface to be processed Surface takes place in the Z direction. In this case, it would be assumed that, for example, the high contact force was caused by a disturbance in the robot arm.

The method according to the invention is based on the procedure of the human body when performing a wiping movement. For example, if a person wanted to dust a piece of furniture with a feather duster, he would try to touch the entire surface of the piece of furniture with the feather duster. The behavior is based on the knowledge that dust particles are electrostatically attracted by the feather duster and furthermore on the assumption that the dust is evenly distributed on the surface of the piece of furniture. Thus, when the feather duster, that is, the tool, was in physical contact with the entire target surface, the human would assume that the dusting task is being performed properly.

However, in many cases the effect of dusting is not directly noticeable. The only reliable feedback is the haptic information that is available to the person during the swiping movements. The use of this haptic information is an important factor if a visual check is not possible. (See [5]: JJ Gibson, "Observations on active touch." Psychological review, Vol. 69, No. 6, p. 477, 1962 ) Haptic feedback thus enables a person to assess the effect of the wiping task carried out by comparing the desired contact forces with the actually felt contact forces. (See [6]: JR Flanagan, MC Bowman, and RS Johansson, "Control strategies in object manipulation tasks," Current opinion in neurobiology, Vol. 16, No. 6, pp. 650-659, 2006 )

If, for example, the person decides that poor contact has taken place, he could repeat the swiping movement in some areas. This combination of cognitive abilities and haptic perception is transferred within the scope of the method according to the invention to a robotic method for performing a wiping movement. Here, the possibilities of robots are used to measure contact forces with a surface to be processed, for example by means of torque sensors or force sensors.

Since the quality of the wiping task carried out depends largely on the type of wiping movement, the tool used, the medium and the depends on the surface to be machined, it is necessary to use suitable models for the parameters mentioned. Such models were presented in the following publication: [8] D. Leidner, W. Benjjani, A. Albu-Schaffer and M. Beetz, “Robotic agents representing, reasoning, and executing wiping tasks for daily household chores”, in Proc. of the International Conference on Autonomous Agents and Multiagent Systems (AAMAS), 2016 . In this publication, however, these models were only used to plan swiping movements and to simulate the effects in advance. However, the quality of the wiping movements actually carried out was not checked here. In the context of the method according to the invention, the models used in the last-mentioned publication can be used to simulate wiping movements.

The method described is thus based on the ability of compliant robots to automatically measure and regulate contact forces. The desired maximum contact force is preferably set in a task-specific manner. Deviations from the desired force are used to detect anomalies. An internal model of the wiping effect makes it possible to evaluate the quality of the execution and to identify bad or completely unprocessed areas.

It is thus possible to determine, depending on the location, at which points the robot has touched the surface to be processed and at which points it has not. Based on the recorded measured values, a contact simulation can be carried out, which simulates where and how the tool (e.g. sponge, wiper, grinding device, etc.) has touched the surface. Any complex CAD data can be used for the tool. The recorded contact forces are then used to analyze the effectiveness of the wiping movement. For example, it can be assumed that the lower the contact force, the lower the wiping effect.

A particle model can be used to simulate the wiping effect. The simulated tool simulates the particles in the event of contact. Here, not only simple cleaning (moving or absorbing particles) can be simulated, but various effects for any media (e.g. cleaning agents, paint, leaves, shards, wiping away chips, etc.) can be simulated, as well as the tasks of absorbing, moving, collecting, Applying, distributing, processing, scrubbing, removing and crushing. The particle model is used for the location-dependent determination of the wiping effect. This enables a robot to evaluate its own movements and automatically plan new movements in order to revise those areas which were previously insufficiently processed.

The properties of different tools, media and surfaces can be simulated using a probabilistic model. For example, it can be assumed that a brush will be less effective than a sponge at removing liquids.

In the prior art, it is always assumed that the planned wiping movements lead to contact with the surface. In the present invention, on the other hand, it is automatically identified when a contact situation is present without an assumption having been made beforehand as to when and where the contact will occur exactly. Quality control is thus possible using the method according to the invention.

In the prior art, methods are also described to distinguish collisions between the robot and the environment (for example with a person) from intentional contact with a workpiece to be machined. However, the effects of collisions or bad contact situations on the quality of the wiping result have not yet been described in the context of wiping movements.

The method according to the invention makes visual feedback superfluous. This is made possible by a combination of the haptic force feedback and the internal particle model.

It is preferred that a contact probability is determined based on the measured contact forces between the end effector and the surface to be processed.

Alternatively or additionally, it is possible to determine a wipe effect probability of occurrence. This is the likelihood that the wiping effect was performed properly. Here, not only the required contact force has to be taken into account, but also the interaction between the wiping tool and the medium to be wiped. For example, it is more likely that shards on the surface to be processed can be wiped away more easily with a broom than is the case with sand (if a broom is also used as a tool).

Furthermore, it is preferred that when creating the model of the wiping movement, a combination of the material of the surface to be processed, the wiping tool used and possibly the medium to be wiped away is simulated.

It is also preferred to take the type of wiping movement into account. A possible wiping movement would be, for example, the absorption of particles, as occurs, for example, when you want to suck up a liquid with a sponge. Another possible wiping movement would be the collection of particles at a common collection point, such as when particles are being swept together.

Another possible wiping movement would be the removal of particles, as occurs, for example, when a surface is sanded.

It is preferred that a Random Sample Consensus (RanSaC) algorithm is used to determine the orientation and / or extent of the surface to be processed.

It is preferred here that the RanSaC algorithm takes into account those contact points between the end effector and the surface to be processed, at which the contact force is above a defined lower limit and preferably below a defined upper limit.

Furthermore, it is preferred that the lower limit is 80%, 85% or 90% of the desired maximum contact force and the upper limit is 100% of the maximum contact force. Basically, values between 75% and 99% of the maximum contact force can be selected for the lower limit. It is also possible to choose the lower limit at 100% and the upper limit at 100%. It would also be possible to shift the corridor between the permitted lower limit and the permitted upper limit, for example to a minimum of 40% and a maximum of 50% in the event that 100% of the force cannot be achieved, but only 50% (e.g. because the maximum value is too is set high).

The advantage of the method according to the invention is that no prior knowledge of the orientation, position and alignment of the surface is required. These are recognized automatically.

In the following, preferred embodiments of the invention are explained with reference to figures.

• 1 einen Roboter bei der Durchführung des erfindungsgemäßen Verfahrens,
• 2 einen Verlauf der erfassten Kontaktkraft und der Endeffektor-Position,
• 3 eine Darstellung der Schätzung der Oberfläche und der Kraftvisualisierung,

Show it:

• 1 a robot performing the method according to the invention,
• 2 a course of the recorded contact force and the end effector position,
• 3 a representation of the estimation of the surface and the force visualization,

In 1 is a surface to be processed 14th , namely a wooden board on which there are a large number of particles, for example in the form of crumbs. The end effector 16 of the robot arm 12th is here a robot hand through which a tool 18th is held. The wiping movement shown can be, for example, a sponge through which the crumbs are removed from the wooden board 14th should be wiped away.

According to the invention, the contact force between the end effector is increased while the wiping movement is being carried out 16 or the tool 18th and the surface to be machined 14th measured. This is in the upper part of the 2 shown. At the same time, the course of the position of the end effector becomes 16 or the tool 18th recorded. The Tool Center Point (TCP) commonly used in robotics can be used for this purpose. If the contact force is within a certain range (e.g. 90% to 100% of the desired maximum force) and at the same time the position of the TCP is in the area in which it is assumed that the surface to be processed is located, it is assumed that the wiping task has been carried out properly. In the event that the contact force is in the desired range between 90% and 100%, but not the position of the end effector, it is assumed that the wiping movement was not carried out properly. For example, this can be due to the fact that the robot arm was touched by a foreign body, for example a person, or that the robot was not correctly aligned relative to the surface to be processed.

In the upper part of the 2 those areas in which the contact force is in the desired range are shown hatched. Those sections of these areas in which the position of the end effector is not in the desired range are shown hatched differently.

These areas correspond to in 3 the sections in which the course of the force visualization covers the area of the simulated surface to be machined 14th leaves (represented by circles). The orientation and / or extent of the surface to be machined 14th is simulated using the contact points mentioned above, which are in the desired areas. The RanSaC method can preferably be used here. Here, positions where there is a high probability of contact are used to estimate the orientation and extent of the surface to be processed. This is in 3 represented by a cuboid. All measured values that are located within this cuboid (shown by crosses) are used in the following to simulate the effect of the wiping movement. On the other hand, measured values that are outside this cuboid are not taken into account, since it must be assumed, for example, that there is a disturbance due to an external force. Such readings are in 3 shown as circles, whereby it can be seen that they leave the cuboid towards the top.

The method according to the invention is based on a particle model. For example, it can be assumed that in the case of solid substances, when a sponge comes into contact with the particles, the position of the particles is changed. If liquids are to be simulated, the particles are deleted when touched, i.e. absorbed. As already mentioned, other arbitrarily complex effects can also be simulated, for example application, removal, etc. The recorded contact force is used in this case to vary the extent of the effect on the particles. The particle distribution changes over time and can be evaluated for quality analysis. In the case of an undisturbed execution of the wiping task, it can be calculated, for example, that 97% of the particles were collected at the destination. In the event of poor contact with the surface to be processed or an external effect, it is also possible to predict the success of the wiping movement, for example success rates of 67% or 80% can be specified here. These simulated success rates were verified by the applicant on the basis of actual wiping tests, it being shown that the simulation was carried out very precisely.

Claims (7)

A method for performing wiping movements by a robot arm (12), the method comprising the following steps: Carrying out a wiping movement on a surface (14) to be processed by the robot arm (12), Detecting a contact force that occurs between the end effector (16) of the robot arm (12) and the surface to be processed (14) during the wiping movement, Create a model of the swipe effect by doing the following: Defining a lower limit for the contact force, above which it is assumed that there is contact between the end effector (16) and the surface to be processed (14), wherein sections of the surface (14) to be processed, in which the lower limit for the measured contact force is exceeded, are used to determine the orientation and / or the extent of the surface (14) to be processed, wherein the position of the end effector (16) in the normal direction to the surface (14) to be processed is detected during the wiping movement, wherein in the sections of the surface to be processed (14) in which the contact force is above the lower limit, it is assumed that the wiping movement had the desired effect, provided that the position of the end effector (16) in the normal direction relative to the surface to be processed (14 ) lies within a defined range, wherein if the measured contact force is in the desired range, i.e. above the lower limit for the contact force, but the position of the end effector (16) is not within the defined range, it is assumed that the wiping movement was not carried out properly. Procedure according to Claim 1 , characterized in that, based on the measured contact forces between the end effector (16) and the surface (14) to be processed, a contact probability and / or a wiping effect probability is determined. Procedure according to Claim 1 or 2 , characterized in that when creating the model of the wiping movement, a combination of the material of the surface (14) to be processed, the wiping tool used and the medium to be wiped away is simulated. Method according to one of the Claims 1 until 3 , characterized in that a RanSaC (Random Sample Consensus) algorithm is used to determine the orientation and / or extent of the surface (14) to be processed. Procedure according to Claim 4 , characterized in that the RanSaC algorithm only takes into account those contact points between the end effector (16) and the surface to be processed (14) at which the contact force is above a defined lower limit and preferably below a defined upper limit. Procedure according to Claim 5 , characterized in that the lower limit is 80%, 85% or 90% of the maximum contact force and the upper limit is 100% of the maximum contact force. Method according to one of the Claims 1 until 6th , characterized in that the alignment and / or extension of a plurality of two-dimensional surfaces (14) are recorded, so that a three-dimensional surface is recorded.

Images