Purposive learning from humans - Hamlyn Symposium for Medical Robotics

Recent successes in purposive learning have been realised by overcoming some key challenges: i) learning object classifications and their relationships (knowledge gathered on the objects and relation of all possible known actions in the form of ontology); ii) activity classifications; and iii) a graph of possible transitions between activities. In this talk, I will highlight works I have done in robot imitation learning from human observations, and demonstrate how reasoning about the meanings of human activities can be a powerful way for robots to learn from humans.     «

Recent successes in purposive learning have been realised by overcoming some key challenges: i) learning object classifications and their relationships (knowledge gathered on the objects and relation of all possible known actions in the form of ontology); ii) activity classifications; and iii) a graph of possible transitions between activities. In this talk, I will highlight works I have done in robot imitation learning from human observations, and demonstrate how reasoning about the meanings...     »