About this Research Topic
Social robots should be able to operate in highly challenging environments populated with complex objects and in social settings involving humans, animals, and other robots. Despite these challenges, we expect these robots to be mindful towards us while executing their tasks, demonstrating adaptive behaviors.
Conventional machine learning approaches do not scale well with the dynamic nature of such real-world interactions as they require samples from stationary data distributions. The real-world is not stationary, it changes continuously. In such contexts, training data and learning objectives may also change rapidly. Lifelong learning is able to address this problem by learning incrementally and facilitating the learning of new concepts, situations, and abilities. Consider a robot that is vacuuming while a person is reading a newspaper. When given a negative feedback in this situation, the robot should be able to identify this as a new context, and thereon adapt its behaviors accordingly in similar spatial or social contexts - e.g., when people are watching TV, the robot should be able to link this situation to the previously experienced one and avoid vacuuming.
The Research Topic seeks answers to the following challenges:
RC1: What can we learn from Social Sciences?
Humans excel in developing their skills using new experiences, without suffering the adverse effects of catastrophic forgetting. Understanding the mechanisms enabling this crucial capability is one of the important goals.
RC2: What can we learn from Machine Learning?
For an embodied agent, lifelong learning brings new perspectives and addresses many critical issues while introducing other challenges. Tackling this critical problem requires more established research agenda which can significantly benefit from the growing literature of established methods in machine learning.
RC3: What issues should be addressed while adopting lifelong learning methods for HRI and social robotics?
Lifelong learning leads to new challenges for the robotics community, e.g.: What cues can be used for making decisions about a new experience from which a concept can be learned? (How can old knowledge/skills be kept while learning new ones? How can seemingly different situations be represented so that they are considered similar? How can lifelong learning be integrated into a fully-fledged cognitive architecture?
RC4: What other alternative approaches can we adopt to foster successful long-term HRI?
For example, program repair and synthesis is a promising alternative adapting to user context in the long term.
We invite regular papers or position papers from a wide range of theoretical, experimental, and methodological approaches, from traditional quantitative, qualitative and mixed methods, through to ethnomethodology and new/emergent methods for studying long-term/lifelong and longitudinal human-robot interaction.
Suggested topics include, but are not limited to:
● Lifelong personalization and/or adaptation
● Modelling user(s) and/or user behavior(s) in multi-session/long-term HRI
● Modelling robot behavior in multi-session/long-term HRI
● Modelling context in multi-session/long-term HRI
● Agent/robot architectures for personalization / adaptation
● Lifelong (long-term) human-agent interactions
● Lifelong (long-term) multimodal interaction
● Lifelong (long-term) multi-user/multi-agent interaction
● Continual/lifelong machine learning
● Alternative approaches (e.g. interactive program repair)
● Development concerns, including deployment, scalability and complexity
● Tools and testbeds for evaluation of multi-session/long-term HRI
● Methodological challenges for achieving successful long-term HRI
● Data analysis and statistical techniques for long-term/lifelong HRI
● Metrics for evaluating long-term/lifelong HRI
● Deployed and/or emerging applications for long-term HRI
● User studies
● Philosophical, legal and ethical considerations
----------------------------------------
Dr. Parisi is Director of Applied AI at McD Tech Labs, where he develops technology for voice-activated drive-thru. All other Topic Editors declare no competing interests with regards to the Research Topic subject.
Conventional machine learning approaches do not scale well with the dynamic nature of such real-world interactions as they require samples from stationary data distributions. The real-world is not stationary, it changes continuously. In such contexts, training data and learning objectives may also change rapidly. Lifelong learning is able to address this problem by learning incrementally and facilitating the learning of new concepts, situations, and abilities. Consider a robot that is vacuuming while a person is reading a newspaper. When given a negative feedback in this situation, the robot should be able to identify this as a new context, and thereon adapt its behaviors accordingly in similar spatial or social contexts - e.g., when people are watching TV, the robot should be able to link this situation to the previously experienced one and avoid vacuuming.
The Research Topic seeks answers to the following challenges:
RC1: What can we learn from Social Sciences?
Humans excel in developing their skills using new experiences, without suffering the adverse effects of catastrophic forgetting. Understanding the mechanisms enabling this crucial capability is one of the important goals.
RC2: What can we learn from Machine Learning?
For an embodied agent, lifelong learning brings new perspectives and addresses many critical issues while introducing other challenges. Tackling this critical problem requires more established research agenda which can significantly benefit from the growing literature of established methods in machine learning.
RC3: What issues should be addressed while adopting lifelong learning methods for HRI and social robotics?
Lifelong learning leads to new challenges for the robotics community, e.g.: What cues can be used for making decisions about a new experience from which a concept can be learned? (How can old knowledge/skills be kept while learning new ones? How can seemingly different situations be represented so that they are considered similar? How can lifelong learning be integrated into a fully-fledged cognitive architecture?
RC4: What other alternative approaches can we adopt to foster successful long-term HRI?
For example, program repair and synthesis is a promising alternative adapting to user context in the long term.
We invite regular papers or position papers from a wide range of theoretical, experimental, and methodological approaches, from traditional quantitative, qualitative and mixed methods, through to ethnomethodology and new/emergent methods for studying long-term/lifelong and longitudinal human-robot interaction.
Suggested topics include, but are not limited to:
● Lifelong personalization and/or adaptation
● Modelling user(s) and/or user behavior(s) in multi-session/long-term HRI
● Modelling robot behavior in multi-session/long-term HRI
● Modelling context in multi-session/long-term HRI
● Agent/robot architectures for personalization / adaptation
● Lifelong (long-term) human-agent interactions
● Lifelong (long-term) multimodal interaction
● Lifelong (long-term) multi-user/multi-agent interaction
● Continual/lifelong machine learning
● Alternative approaches (e.g. interactive program repair)
● Development concerns, including deployment, scalability and complexity
● Tools and testbeds for evaluation of multi-session/long-term HRI
● Methodological challenges for achieving successful long-term HRI
● Data analysis and statistical techniques for long-term/lifelong HRI
● Metrics for evaluating long-term/lifelong HRI
● Deployed and/or emerging applications for long-term HRI
● User studies
● Philosophical, legal and ethical considerations
----------------------------------------
Dr. Parisi is Director of Applied AI at McD Tech Labs, where he develops technology for voice-activated drive-thru. All other Topic Editors declare no competing interests with regards to the Research Topic subject.
Keywords: Human-Robot Interaction, Life-Long Learning, Long-Term Interaction, Adaptation, Personalisation
Important Note: All contributions to this Research Topic must be within the scope of the section and journal to which they are submitted, as defined in their mission statements. Frontiers reserves the right to guide an out-of-scope manuscript to a more suitable section or journal at any stage of peer review.
