Objectives

It has been a long-standing question whether robots can reach human level of intelligence that understands the essence of observed actions and imitates them even under different circumstances. Contemporary research in robotics and machine learning has attempted to solve this question from two different perspectives: One in a bottom-up manner by, for instance, solely relying on perceived continuous sensory data, whereas the other by approaching rather from the symbolic level in a top-down fashion. Although there have been shown encouraging results in both flows, understanding and imitation of actions have yet to be fully solved.

Action semantics stands as a potential glue for bridging the gap between a symbolic action representation and its corresponding continuous signal level description. Semantic representation provides a tool for capturing the essence of action by revealing the inherent characteristics. Thus, semantic features help robots to understand, learn, and generate policies to imitate actions even in various styles with different objects. Thus, more descriptive semantics yields robots with greater capability and autonomy. In this full-day workshop, we aim at answering two major questions.

1. What have we learned from action semantics? In recent years, there has been a substantial contribution in semantic policy and action representation in the fields of robotics, computer vision, and machine learning. In this respect, we would like to invite experts in academia and motivate them to comment on the recent advances in semantic reasoning by addressing the problem of linking continuous sensory experiences and symbolic constructions to couple perception and execution of actions. This is of fundamental importance to ease the symbol grounding problem in robotics.
2. How much of semantic policy and action representation have been transferred from controlled lab setups to industrial environments? We would like to invite researchers from industry and initiate a discussion between academic and industrial communities. Such a provocative discussion catalyzes the interaction between the two communities by addressing the scalability and generalization problems which still remain unsolved. In this respect, we would like to discuss how to transfer our current knowledge and experience about semantic policies to new domains, for instance, industrial assembly tasks, with very little human intervention.

This workshop focuses on new technologies that allow robots to learn generic semantic models for different tasks. In this workshop, we will bring together researchers from diverse fields, including robotics, computer vision, and machine learning in order to overview the most recent scientific achievements and the next break-through topics, and also to propose new directions in the field.

Important Dates

• Workshops paper submission deadline: September 27th, 2019. (Extended): October 11th, 2019
• Workshops paper notification: October 20th, 2019.
• Workshops camera ready submission: October 31st, 2019.
• Workshop date: November 8th, 2019.

Acknowledgement