Objectives

The main objective of this first workshop is to bring together a multidisciplinary group of researchers to identify and address key challenges for studying long-term / lifelong learning and its relevant aspects for social robotics in both lab and field. More specifically, the workshop aims

• to bring forth existing efforts and major accomplishments in long-term / lifelong learning that can potentially be used for HRI and social robotics,

• while encouraging the design of novel models, datasets, tools and applications in the context of long-term human-robot interaction and adaptation, and

• to focus on current trends and future directions in the field.

Research Challenges

Our workshop will focus on the following specific research challenges:

• RC1: What can we learn from studies from Social Sciences?

Humans excel in extending their skills with new experience without suffering adverse effects of catastrophic forgetting. The findings that shed light on the mechanisms enabling this crucial capability will be one of the important highlights of our workshop.

• RC2: What can we learn from studies from Machine Learning?

Lifelong learning, despite being very popular in the machine learning community, has not received as much attention from the robotics community. For an embodied agent, lifelong learning brings new perspectives and addresses many critical issues while introducing other new challenges. Tackling this critical open problem requires more established research agenda which can significantly benefit from the growing literature of established methods in the machine learning community.

• RC3: What issues should be addressed while adopting lifelong learning methods for HRI and social robotics?

Lifelong learning allows a social robot to extend itself with new experiences so that it can adapt to new situations and challenges. While addressing the important problem of adaptability, lifelong learning leads to several challenges, which need to be addressed by the robotics community. These challenges include for example: (i) Cues: What cues can be used for deciding a new experience as including something novel from which learn a new concept? (ii) Catastrophic forgetting: How can the robot keep the old knowledge/skills while learning new ones? (iii) Representation: How can seemingly different situations be represented so that they are considered similar and a robot can adapt its behaviors by transferring its experiences from a similar situation? (iv) Architecture and Modeling: How can a lifelong learning “module” can be modeled and integrated into a full-fledged cognitive architecture for robots.