A plea for a proper focus in robotics research
conference paper
Classical research in first generation robotics has been mainly directed towards industrial robotics using very basic methods in sensing (e.g. in vision). This is not surprising because robots were initialy primarily considered to replace humans in work that was either dìrty, dangerous, or debilitating. With the ongoing increase in afordable
computhg power, it was possible with simple methods to record the movements of factory workers and a robot was
then "trained" to mimic these movements. In terms of speed and product quality control, this approach has given a
considerable boost to the eficiency of industríal production and thus to economy. From the point of view of robotics research however this approach is very restricted. Today one still notices a considerable gap between what is shown in e.g. science fiction movies and the actual implementation of robots. There are several factors contributing to this phenomenon, e,g. the poor perfomance of classical AI methods, the lack of powerful system concepts, the importance attributed to emulation and the underestimation of experimenting, and finally the predominance of high-precision digital approaches, as opposed to hierarchies of dífferent granularities and analog solutions. By starting from system concepts we try to identify in this article the areas in robotics that are relatively underexposed, yet scientifically present great challenges in fundamental research. We argue that softcomputing may be viewed as one of the key enabling technologies for developing the next generation of robots.
computhg power, it was possible with simple methods to record the movements of factory workers and a robot was
then "trained" to mimic these movements. In terms of speed and product quality control, this approach has given a
considerable boost to the eficiency of industríal production and thus to economy. From the point of view of robotics research however this approach is very restricted. Today one still notices a considerable gap between what is shown in e.g. science fiction movies and the actual implementation of robots. There are several factors contributing to this phenomenon, e,g. the poor perfomance of classical AI methods, the lack of powerful system concepts, the importance attributed to emulation and the underestimation of experimenting, and finally the predominance of high-precision digital approaches, as opposed to hierarchies of dífferent granularities and analog solutions. By starting from system concepts we try to identify in this article the areas in robotics that are relatively underexposed, yet scientifically present great challenges in fundamental research. We argue that softcomputing may be viewed as one of the key enabling technologies for developing the next generation of robots.
TNO Identifier
95148
Publisher
Association for Intelligent Machinery
Source title
Proceedings JCIS '98, October 23-28, 1998, Research Triangle Park, NC, USA
Editor(s)
Wang, P.P.
Place of publication
Research Triangle Park, NC, USA
Pages
76-79
Files
To receive the publication files, please send an e-mail request to TNO Repository.