Industrial robot calibration

Industrial robotics took off in the USA in 1963 with the Unimate robot from Unimation. The first applications involved handling tasks in auto mechanics workshops, followed by spot welding in the assembly shops of major automakers. All these applications were labor-intensive, and were recognized for their arduous nature. Spot welding was therefore the first process application to be robotized. This application was also the driving force behind the introduction of off-line programming (OLP) in the industry.

Spot-welding programming has been carried out solely by apprenticeship for fifteen years. In this programming mode, the operator uses a console to manually move the robot using push-buttons, and to memorize the robot's position (when stationary) on specific fixtures. Over the years, this programming has been enriched with various robot-specific functions, as well as process-specific functions. These functions refined programming, facilitated the operator's task and improved the quality of the final result. They did, however, lengthen the time needed for programming, but as the number of robots installed remained low, users were content to make do with them. By the early 1980s, more and more robots were being installed on assembly lines, and the time needed to start them up and fine-tune them was increasing. It was therefore imperative to reduce these times.

There were two possibilities: the duplication (from one robot to another) of existing, successful programs, or the analytical creation of new programs (the forerunner of PHL).

Program duplication could be carried out in two ways, either from a robot in the installation, or from a so-called school station located off-line. The notion of offline programming was born.

This meant redesigning the assembly lines, which made program duplication virtually impossible.

However, for the sake of completeness, we should point out that painting robots (among the pioneers of robotics) used a particular programming method: the operator learned the robot's trajectory via manual guidance at real speed. The main shortcomings of this programming method lay in the mass to be handled (inertia problems). The quality and cycle time required were not easy to achieve, as program corrections were not possible, and meant that a large number of programs had to be created in order to retain only the best.

A variant of this programming approach appeared around 1978, using a lightweight mechanical structure with the same theoretical kinematic model as the robot. This structure was called a syntaxer (Renault later called such a system a "puppet"). At the time, we were in the context of off-line programming, and the problems encountered prefigured...