From Laboratory to Robot Competition

• In Event: Sensing Robots

Abstract

Engineers and scientists create robots using laboratories. Much has been written about how science is done within these laboratories. Laboratory methods mainly focus on cause-and-effect relations between independent and dependent variables. The usual strategy is to reduce variables by reducing complexity. Reducing complexity worked out for decades robots while were limited to factory shop floors, where the environment could be controlled to fit the needs of the robots. However, recently robots enter real homes and the method has to change accordingly. Researchers who are conducting studies in real homes have much less control over the variables. Especially the sensory overload of the massively complex everyday worlds overexerts the capabilities of recent robotics in its fundamental research state. Nevertheless it seems that currently robot research almost entirely focuses on application scenarios for everyday worlds.
I want to show how competitions are used to fill the gap between over-fitted laboratories and the overcomplex real worlds by creating an arena, which complexity exceeds the laboratory’s, but still bears less contingency than the everyday worlds. Along the line of laboratory research I want to show how the practice of utilizing these competitions to get more sensory feedback from a controllable yet open environment becomes an epistemic tool of building robots. I will draw from empirical examples as the RoboCup@Home which was initiated when field test in real (closed) supermarkets failed drastically due to complexity overload.

Author

• Arne Maibaum, TU Berlin