DRIVSCO

Another related project is the DRIVSCO project. The topic of the DRIVSCO project looks fairly similar to the DIPLECS project, in particular the intended application area of car driver assistance, the strategy to ”start with hard coded functionalities and remove assumptions successively”, to use perception-action learning, and to correlate eye-movements to car sensor measurements.

In contrast to the DRIVSCO project, DIPLECS does not aim at building a complete system but has a much stronger focus on studying architectural principles, learning methods, and limits of learn-ability. Potential results of the DIPLECS project will have a more general character with impact to many other application areas. Furthermore, cognitive bootstrapping by action-precedes-perception learning and strong feedback cycles are unique to the COSPAL philosophy which is to be continued in DIPLECS. Furthermore, and in contrast to DRIVSCO, the COSPAL philosophy assumes that no sensor input (including vision) is completely reliable and measurements are contaminated by noise, outliers, and total failures due to wrong model assumptions.