The main innovation brought forth by the European Project RoCKIn is the definition, implementation and application to an actual robot competition of the novel paradigm of benchmarking through competitions. By doing so, RoCKIn set in motion an evolutionary process to transform robot competitions from successful showcases with limited scientific impact into benchmarking tools for the consistent and objective evaluation of the performance of autonomous robot systems. Our work began by revisiting, in the light of the features and limitations of a competition setting, the very foundations of the scientific method; then we built on these by designing a novel type of competitions where the concepts of benchmark and objective performance metrics are the key points; finally, we arrived to the implementation of such concepts in the form of a real-world robot competition. This chapter describes the above process, explaining how each of its several aspects (theoretical, technical, procedural) has been tackled by RoCKIn. Special attention will be devoted to the problems of defining performance metrics and of capturing the ground truth needed to reliably assess robot perceptions and actions.
Part of the book: RoCKIn