The ION architecture for robot learning

P. Qualtrough · 2002

It is claimed that one of the main reasons why the development of intelligent robots has been slower than expected is that machine learning has been seen as an "add on" feature-one to be placed in the higher and later-developed levels of robot architectures. A case is made for incorporating machine learning at the earliest possible stage, and relying on it as the primary method of developing robot controllers. An architecture is proposed to support this approach.

Read the paper · More papers on PaperTik