Home » Research » Publications » Perspectives of Robot Action Generalisation: Challenges and Achievements
Perspectives of Robot Action Generalisation: Challenges and Achievements
Title
Perspectives of Robot Action Generalisation: Challenges and Achievements
Publication Type
Conference Papers
Year of Publication
2014
Authors
Veskos, P, Pastra, K
Conference Name
International IEEE Workshop on Autonomous Cognitive Robotics
Conference Location
Stirling, UK.
Abstract
Making robots more accessible necessitates simplifying their skill acquisition processes. The traditional approach of programmatically specifying each skill that a robot should have is too low level and repetitive to be conveniently used by non expert users. The concept underlying this requirement for higher level communication and teaching is economy. Economy in learning is a highly desirable human skill for robots to possess: it can maximise the utility of behaviours already learnt, by reusing them in modified forms, hence minimising the overhead of training new ones.
Making robots more accessible necessitates simplifying their skill acquisition processes. The traditional approach of programmatically specifying each skill that a robot should have is too low level and repetitive to be conveniently used by non expert users. The concept underlying this requirement for higher level communication and teaching is economy. Economy in learning is a highly desirable human skill for robots to possess: it can maximise the utility of behaviours already learnt, by reusing them in modified forms, hence minimising the overhead of training new ones.