Christian Czarnecki - Robotic Process Automation: Management, Technology, Applications

In practice, a broad variety of robotic process automation (RPA) cases have been documented, such as the automation of core processes at Telefonica O2 ().

RPA is an innovative approach to transform the process execution without changing the underlying application systems (e.g., ).

This book follows a broad understanding of RPA, as an umbrella term for different automation approaches (e.g., .

On January 21, 1899, Scientific American overviewed Some Curious Automata. One of them was the automatic duck created by the French inventor Jacques de Vaucanson in the eighteenth century shown in ).

Figure 1 Blueprint of an automatic duck (, p.43).

Note that there is debate whether the (graphical) description of the automatic duck given above is historically correct. From our research background, we do not feel competent to contribute to this episode of the history of automation. However, we like to use this background as a setting for the following gedankenexperiment (thought experiment). For the sake of our argument, we do like to focus on the following thesis, which was already stated above in a slightly different form:

It is impossible to distinguish the automatic duck from a real, living one.

Hence, we like to ask the following question:

What are the necessary prerequisites that this statement impossible to distinguish the automatic duck from a real, living one is true?

Imagine an automatic duck in our time. What would be required to confirm the above statement? At least, the following assumptions have to be made:

Sensors: Our automatic duck needs to have some mechanical or digital sensors to realize the world around it. With respect to todays state-of-the-art technology, this might be, for example, a digital camera, to search for and identify seeds. Presumably, our automatic duck is equipped with further sensors like radar sensors, some GPS antenna or light detection, and ranging sensor (lidar).

Actuators: Our automatic duck has to have some mechanical beak, legs, and wings, which enable the duck to pick, to move, and to fly around. Those actuators are required to interact with the real world.

Functions: Our automatic duck must be able to perform some typical actions, which a real duck can perform, for instance, waddling, quaking, and splashing in the water.

To be sure, these are at least some minimal requirements to make this imagination of an automatic duck true. If this illusion will be perfect, then it is clear that the automatic duck is still a thing, a piece of human engineering art, but not a real, living duck. However, by definition, an ideal, automatic duck cannot be distinguished from a real duck. Therefore, we can call such an automatic duck functionally equivalent to a real duck.

A very interesting question is now, what is meant by functionally equivalent? In which contexts does it make sense to distinguish between a real and an automatic duck? For example, on the one hand, it can be argued that the automatic duck must be fed the same way as a real duck when it is hungry. On the other hand, it might be countered that it is not necessary, because the automatic duck is just a machine.

As long as we keep the postulation that our automatic duck should be an equivalent to a real duck, there should be no discussion about copying all aspects. However, this approach comes to a point where some aspects might be questionable or even inadequate. For example, would it really be advisable that our automatic duck has a lifetime comparable to that of a real duck? Even if this point is confirmed, how should the aging for an automatic duck be realized technically? In addition, from an environmental perspective, is it really reasonable to produce a machine that has a predetermined, man-made end of life? Obviously, there are other aspects of a real duck (e.g., reproduction) that would at least require intense discussions that go beyond the scope of this book. In conclusion, the equivalent between our automatic duck and a real duck would only be valid for a set of aspects.

Excluding the postulation of a full equivalent between real and automatic duck, the required functionalities can only be judged based on the usage context. For example, for an automatic duck that is used in an educational context for agricultural students, simulating an adequate duck housing, feeding is an essential functionality, whereas an automatic duck used in a surveillance context to deter potential housebreakers, quacking is the core functionality, while feeding might even be hindering in this case.

Furthermore, while developing an automatic duck further questions arise: Would it be a good idea to construct a duck which does not consume so many seeds? Or to construct a duck that is capable to fly much faster and higher than a real duck? Is such a duck still functionally equivalent to a real duck? Of course not; however, changing or optimizing functionalities could be recommendable, and might lead to an advantage of our automatic duck compared to a real duck. For example, in the usage scenario of the agricultural education the possibility of reusing seeds due to an intentionally omitted digestion function is clearly beneficial. Further developments could include useful functionalities for this context (e.g., evaluation of students performance) which have never been a part of a real duck.

In summary, our gedankenexperiment demonstrates three facets:

Although it might be interesting to have a full functionally equivalent duck, in most practical cases the functional scope is limited to a specific set of aspects, and does not imply copying all functions of a real duck.