Accueil > Robotics > Robots in industry

Robots in industry


Robots in industry

What fundamentally characterizes an industrial robot?

  • Its performance?
  • Its programming resources?

If in fact these two concepts are inseparable to characterize an industrial robot, the application remains the key element in any robotic thinking. And for any industrial project, the application is the starting point for the study of the robotic engineer.

While industrial robotics was born in 1963, it was not until the 1980s to attempt through standardization to clarify what was hidden behind these words. The most important of these is undoubtedly performance. It was the latter that imposed the most thought on standard setters. To fully understand the constraints of Offline Programming (PHL), it is necessary to review all the elements included in this term “performance”, and in particular the notions of repeatability and precision, source of much confusion.

But a successful installation is an installation whose performance is known to be maintained over time. Good maintainability is prepared at the same time as the general design of the installation, in the same way as, for example, the maintenance operations of an automobile, which are now planned from the study of the vehicle.

We must therefore never lose sight of the fact that all the constraints that will have to be taken into account during the study of the installation come directly from the application. Robotics has developed from applications. The greatest changes have taken place around these, and in particular those whose parameters were best dominated, or for which intense research has been carried out to better discipline them.

I. Reminder on the origins of PHL.


From the birth of industrial robotics in 1963 until the early 1980s, robot programming was done primarily through learning. The robot designers of that time, already aware of the difficulties of programming and the consequent heavy losses of time, tried to adapt the means of programming to the applications. The programming board integrated into the front of the control bay has replaced the mobile console for handling type applications (spot welding was at that time equated with handling). For a process application like painting, human body language was the natural movement to be reproduced. So the programming was done by manually moving the mechanical structure of the robot or a lightweight structure called a syntaxer (or puppet) by the man himself.

Note anecdotally, that in the case of the use of a syntaxer, we are in an offline programming context; and therefore, as early as 1975, users of these systems became aware of the importance of respecting the dimensions of the various constituent elements of an installation.

When it became necessary to tackle applications such as arc welding, the situation worsened. Not only the performance of the robots was insufficient, but also the programming functions. By making the latter evolve, the desks became less and less manageable monsters, the time required for programming increased.

In addition, the safety aspects were often forgotten, not to say ignored. The concept of offline programming therefore arose naturally. In summary, Offline Programming arose for:

  • Make the facility available to its production function more quickly.
  • Reduce the overall programming time (program creation and fine-tuning).
  • As a result, improve safety by reducing the presence of humans in the machine’s range.
  • Provide oneself with means of simulation, beyond the initial purely analytical character.


The hopes placed in the PHL very quickly vanished. Robotic CAD resources have generally met expectations as a design aid, as an aid in determining cycle times and in determining intrinsic feasibility at the level of known processes. The translation of simulation programs into runtime programs has turned out to be a different matter altogether.


The parameters of the geometric models taken into account by the simulation system (theoretical parameters) and those of the site (real parameters) differ sufficiently so that the trajectories determined in simulation are not directly transposable. On the other hand, anyone who has attempted to transport a robot program from one site to another (considered similar) has encountered serious difficulties. The same is true for robot swaps. Even a simple maintenance operation (during which some mechanical disassembly takes place), can take a catastrophic turn on restart.

Various attempts at improvement have been found to be insufficient, such as zero encoder settings, stampings, etc. because other factors, hitherto unsuspected, due to a lack of means of measurement, disrupted too incomplete analyzes.

The final finding is clear: The overall time required to complete offline programming is equal to or sometimes greater than that of traditional programming. However, the downtime of the installation is significantly reduced, and the resulting financial gain significant.

PHL is a natural evolution of traditional programming. It can be done in different forms, by programming from a syntax machine, from another robot, in pure analytics, or from a computer simulation system. The experience of the pioneers of the PHL showed that additional precautions or means had to be taken.

II. Conditions to be fulfilled to pass a PHL?

Ideally, the installation should be identical to the simulation model. This therefore means that:

  • All real components are strictly identical to their models.
  • All the components are perfectly positioned in space as in the theoretical simulation model.

Since these ideal conditions can only be observed exceptionally, a rigorous process must be followed.


As experience has shown that the use of PHL cannot be improvised, it should be taken into consideration when launching a new project. The approach should be based on the application considered and the results expected from it.
We must ask ourselves the following questions:

  • Under what conditions is the application suitable? (Essentially in terms of static and dynamic precision).
  • Is the PHL justified in this case? (Mainly for financial reasons).
  • Can we derive other benefits from it? (For example in operator safety, maintenance monitoring, etc.).

From this initial analysis, we can deduce the specific constraints linked to the project, and therefore the means to be implemented.

The first act is the initial analysis. Above all, you have to be able to appreciate the real performance necessary for the smooth running of the application. In general, the application imposes static and / or dynamic repeatability constraints and sometimes tool center speed constraints (in precision compared to a requested speed). In the context of the PHL, only the precision in static behavior and dynamic is significant.

Then, it is necessary to define the work volume (s) in which the robot will have to evolve. As the precision characteristics of a robot deteriorate particularly when approaching the limits of evolution of the axes, care must be taken to choose a robot whose work envelope is clearly greater than the actual volume required by the application.

Particular care will be taken in the study and construction of the support floors for the robots and their environmental accessories. A movement of the robot base of 0.1mm can result in an error of 0.5mm or more in the tool. Etc.

Knowing the conditions to be met for the employment of a PHL is the best guarantee of success. These conditions make it possible to decide not only what will be the general design of the site, the precautions to be taken during its assembly but also if a calibration is necessary.

The care taken in preparing for calibration is also a success factor. Above all, as we will show later, it will make it possible to limit on-site intervention times, and therefore its downtime. We should note that outside of certain material handling applications, in most cases calibration is required.

III. The contribution of dynamic control of the characteristics of an industrial robot.


These are defined, as well as the conditions under which the measurements must take place by the ISO 9283 standards.
These characteristics are generally the following:

  • Static repeatability.
  • Static precision of installation.
  • Multidirectional variation of the static installation precision.
  • Cornering error.
  • Drift of the pose.
  • Precision on a trajectory.
  • Trajectory repeatability.
  • Speed accuracy on a trajectory.
  • Speed repeatability on a trajectory.
  • Static compliance.


These characteristics can now be verified by metrological means based on various sensors and appropriate software. Software allows testing as recommended and described in ISO 9283 standards.

However, experience has shown that ISO testing is sometimes insufficient to fully characterize a robot against an application. In fact, for a given application, the roboticist must face particular trajectories, speeds and specific loads. Sometimes also, stopping points can present particularities by their position in space, by the trajectory and the speed of approach, the authorized or unauthorized passing, the stabilization time. Finally, it is often important to know how to determine the true cycle time, which cannot always be given by robotic simulation systems. Complementary software has therefore been created to respond to these scenarios. They give the operator total choice in his trajectories and test procedures so as to meet his real needs.

Among these additional tests, we will also cite the Hysteresis test. It is essential for measuring mechanical clearances. Accompanied by an essay on the overshoot error, it shows the behavior of a robot according to the state of its mechanics at an instant “t”. It is therefore one of the necessary tools in maintenance monitoring.
The main types of sensor used to date are:

  • Digitized tablets.
  • Laser proximity sensors.
  • Cameras associated with two or three.
  • Laser tracking interferometers.
  • Laser transmitters associated with receivers.
  • Wire potentiometers.

All these systems allow working in dynamics. Some are limited to a plan like tablets; laser transceivers simply check a movement on a straight line, the same for proximity sensors. The other solutions allow large displacement in space, the use of three associated cameras allows a measurement in 6 dimensions, that is to say that the measurement will include not only the spatial position, but also the orientation. Wire systems have the disadvantage of requiring contact with the robot and therefore in certain circumstances influencing the measurement results.

We now have a range of non-contact static and dynamic metrology products which allows a robot to be fully qualified in all its typical behaviors:

  • Tracking a trajectory.
  • Stop on a point.
  • The speed of movement.

Non-contact dynamic metrology takes on its full dimension, for example when measuring mechanical clearances, because the measuring system itself does not interfere with the result.

These means open up new perspectives, some of which have already proven their value. They are described in the following paragraphs.

IV. Robotic calibration and trajectory recalculation.

Depending on the application, the design of the installation, the means used and the desired goal, robotic calibration can be viewed from different angles. Sometimes it is the robot that will be treated, sometimes its environment, but more and more, it is the entire installation that is taken into account, thanks to the new equipment mentioned above, and to the development of adapted software and procedures. So it is the latter case, the most general that we will deal with, both
precedents being considered as special cases.


The robotic calibration procedure consists of 3 main phases:

  • The preparation.
  • Measures.
  • The cheking process.


Preparation is an essential phase, because by the care taken, the time spent on the site will be reduced.
In addition to a preliminary analysis of the characteristics of the robot if it is not yet known, the preparation consists in determining:

  • The location of the measurement cameras.
  • The mounting of target (s) on the robot.
  • Additional measurement points in the robot’s environment.
  • The preliminary verification trajectories of the robot’s behavior.
  • The robot’s calibration trajectories.
  • The verification and qualification trajectories of the measurements carried out.

All these simple trajectories can be transferred conventionally to the robot’s control bay. In the case of new installations, it is also possible to provide on its robotic CAD the accesses and location of the cameras, the fixings of the infrared targets on the tool mounted on the robot as well as the measurement points on the environment. . By doing this, we can fully simulate the calibration procedure as well as the simulation of the work done by the robot.


They take place in 2 stages:

  • The behavioral verification of the robot.
  • Identification measures (robot then environment).

Behavioral diagnosis is essential for several reasons. It makes it possible to avoid continuing operations if a serious anomaly is detected, for example excessive play, an incorrect gear ratio. We will also seek to be as close as possible to the actual operating conditions of the machine, depending on the application (volume of work, load, etc.). This check also makes it possible to choose the rest of the procedure. The simplest and most common is based on the method of analyzing axes independently of each other.

This method is applicable to any robot model with “healthy” behavior, that is to say one whose mechanical flexibilities are limited and mostly predictable. In the event that the robot leaves this context, the measurements are carried out from false matrices of points which are determined from the diagnosis and the expected result.

For the robot, pure calibration measurements are carried out in the same way as checks, but from specific trajectories. The environment is treated with a portable self-calibrating probe. This probe works in 6D, which allows the operator to take measurements quickly with little precaution. In this way, the robotic performance control and calibration system becomes a measuring machine with measuring accuracy of about 2 to 3/10 mm at a measuring distance of about 2.5 m.


After a calibration operation, the measurements carried out are checked and qualified. This control is carried out from a predetermined number of points arranged in the working era. The positions before and after calibration are compared and the positioning accuracy is measured with respect to the programmed theoretical points.


Calibration is responsible for correcting errors. These can be distinguished as geometric errors and non-geometric errors, identifiable errors and unidentifiable errors, correctable errors and uncorrectable errors. These are not always the same depending on the methodology chosen and the recalculation means used.

All these errors are not of the same weight, this weight varies from one robot to another, it also varies depending on the design of the installation and the means chosen in its realization. The consideration of errors rather than others also depends on the desired result.

The main parameters concerned are:

  • Zeros encoders.
  • The inter-axis distances (or overall length of the arms).
  • The interaxis angular positions.
  • The flexibility of the arms due to a load.
  • The flexibility of the arms due to a transmission effort.
  • The flexibility of transmissions.
  • Transmission reports.
  • The eccentricity of the transmissions.
  • The flexibility of the soil.
  • The position of the tool center.
  • The position of the equipment constituting the environment of the robot (for example a conveyor).
  • The position of the tools located on the previous equipment.


The trajectory recalculation filters depend directly on the chosen calibration methodology. In the most frequent case, the geometric model of the robot is redefined by the method of analyzing the axes independently of each other. This model is completed by determining the true position of the robot’s environment in relation to the robot’s coordinate system. The filter that is deduced from this makes it possible to take into account all the parameters described, including some of the consequences due to an on-board load.

When the flexibility is excessive, and especially when the geometric model is “unstable”, the filter is then based on a matrix of points whose definition varies according to the volume of work and the claims in terms of result.

Today, the means and procedures for robotic calibration in industry are a reality. It is again the application that is the guide in these actions. An answer exists for almost all classic cases. As an indication, we can quote the following figures:

In a working volume of 1 m3, for a robot whose original precision is between 5 to 10 mm, after calibration the average precision is reduced to better than 5/10 mm (for 3 s), without ever being beyond 1 mm.
For a working volume of 4 to 5 m3, for a robot whose original precision is between 10 and 20 mm, the average precision will be of the order of 8/10 mm (for 3 s), and a maximum less than 2 mm.

The time required to complete this work varies greatly. Apart from anything that may have already been cited as a constraint, it also depends on local conditions and the experience of the operator. In the best case, and for repetitive installations, with proper preparation, the installation downtime can drop to 1/2 hour. For a classic installation, but to discover, it is necessary to envisage 1/2 day to 1 day. For complex cases, with “unstable” geometric models, high volume of work and multiple workstations, the week may be necessary.

V. Maintaining performance.


Curative maintenance is still currently in use and still has its defenders. If it is inexpensive to maintain, it can have dramatic consequences for production costs, and for the company’s branding if the failure occurs during the fulfillment of a large order.

Preventive maintenance is overwhelmingly employed. Its efficiency and intrinsic costs improve over the years through the experience gained in the knowledge of materials. However, we find that very often parts are changed long before they reach the end of their life. The frequency of interventions is higher than necessary. A decrease in this frequency is risky as long as there are no reliable means of evaluation.


Predictive maintenance requires the availability of means that make it possible to judge the behavioral evolution of production tools. We have just described all of these means. If we associate objective physical measurements with:

  • Preventive maintenance experience
  • Information from quality control,

we then have all the ingredients to start predictive maintenance. As with preventive maintenance, it is over the years that people will acquire perfect mastery of this new concept.

The results acquired in this field in recent years, as well as the additional studies carried out on this subject, prove the value of this type of maintenance. It requires, because of its existence, better monitoring of the equipment, therefore better knowledge of its condition, and better reactivity.

If we look at the measurements described during a calibration operation, we find that we have all the information that defines the “zero” state of the hardware. By regularly resuming maintenance measurements, the basic monitoring necessary for predictive maintenance is established. Naturally, these measurements can be carried out on installations without any consideration of Offline Programming.

By installing a robot drift control system, the user protects himself against a serious incident, the causes of which may sometimes be external to the robotic installation. By implementing predictive maintenance, the user increases the intrinsic efficiency of his means of production.

The dynamic control systems for the performance of industrial robots have made it possible to develop industrial means of robotic calibration, and by extension to have the information necessary for a preventive maintenance approach. Using simple measuring tools, it is also possible to guard against unwelcome robot positioning drifts, thus improving the overall operating safety of robotic installations.

VI. Conclusion. 
The dynamic performance control systems of industrial robots make it possible to monitor the life of an industrial robot. If thanks to these tools, we can calibrate a robotic installation, help predictive maintenance, it is also possible, always in the same context, and therefore simultaneously, to ensure an installation recipe. In addition, as an extension of predictive maintenance, it becomes easy to meet the conditions imposed by ISO9000 quality assurance procedures.

It may interest you

Leave a Reply

Your email address will not be published.

Solve : *
23 + 26 =