About this Research Topic
Humans can learn new skills and recognize new things quickly from a small number of data points. This could be attributed to our ability to generalize concepts and transfer from one task to another. For instance, using affordance perception, humans can easily recognize if a cuboid can be sat on even if they have never seen it or used it before. Similarly, humans can precisely predict the trajectory of a moving ball by perceiving and predicting the physical laws. Can robots use similarly layered cognitive systems to learn efficiently? Recent progress has been made in this area, but there are still many unsolved problems for efficient robot cognition and learning.
The aim of this Research Topic is to collect comprehensive updates and high-quality practices regarding machine-learning-based robot cognition. A generalist agent needs to combine high-level affordances, intermediate-level human or robot language, and low-level prediction and recognition of physical equations (matching or learning an observed phenomenon with known physical laws) in order to perform in a large variety of tasks and environments. The goal is to improve the state-of-the-art in the domains: language integration, affordances, and physics-based inductive biases and representations or their combination. Machine learning provides a common framework for jointly learning missing parameters from real-world data. An affordance allows collecting action possibilities, enabling fast discovery and learning the environment, often from one or a small number of observations. In addition, natural language provides a simple and promising approach for robotic communication and cognition tasks. Recent publications about the combination of large language models and robotics showed that positive transfer is possible when additional modalities or tasks are included in the data sets. Language models can also be used to generate policy code, and act as a translator between human language and robotics related languages. Alternatively, robotic learning combined with physic-informed information allows robots to extrapolate from a small number of samples.
The Research Topic seeks contributions of Original Research, Systematic Review, Data Report, Methods, and Review. Areas of interest include, but are not limited to:
- Affordance learning and perception in robotics
- Large-language-model-based robot learning
- Understanding manipulation and tool-discovery affordance
- Learning natural language interfaces of human-robot interaction
- Self-supervised learning for robot affordance and language representation
- Robot language processing
- Physics-informed robot learning
- Machine-learning models combining affordance representation, language processing, or physics-informed priors
- Learning models multi-modal observations with transformers
- Foundation Models for Robotics
Dr. Chen, Dr. Karl, and Elie Aljalbout are affiliated with Volkswagen; Dr. Zeng is affiliated with Google; Dr. Mayol-Cuevas is affiliated with Amazon. All other Topic Editors declare no competing interests with regard to the Research Topic subject.
The aim of this Research Topic is to collect comprehensive updates and high-quality practices regarding machine-learning-based robot cognition. A generalist agent needs to combine high-level affordances, intermediate-level human or robot language, and low-level prediction and recognition of physical equations (matching or learning an observed phenomenon with known physical laws) in order to perform in a large variety of tasks and environments. The goal is to improve the state-of-the-art in the domains: language integration, affordances, and physics-based inductive biases and representations or their combination. Machine learning provides a common framework for jointly learning missing parameters from real-world data. An affordance allows collecting action possibilities, enabling fast discovery and learning the environment, often from one or a small number of observations. In addition, natural language provides a simple and promising approach for robotic communication and cognition tasks. Recent publications about the combination of large language models and robotics showed that positive transfer is possible when additional modalities or tasks are included in the data sets. Language models can also be used to generate policy code, and act as a translator between human language and robotics related languages. Alternatively, robotic learning combined with physic-informed information allows robots to extrapolate from a small number of samples.
The Research Topic seeks contributions of Original Research, Systematic Review, Data Report, Methods, and Review. Areas of interest include, but are not limited to:
- Affordance learning and perception in robotics
- Large-language-model-based robot learning
- Understanding manipulation and tool-discovery affordance
- Learning natural language interfaces of human-robot interaction
- Self-supervised learning for robot affordance and language representation
- Robot language processing
- Physics-informed robot learning
- Machine-learning models combining affordance representation, language processing, or physics-informed priors
- Learning models multi-modal observations with transformers
- Foundation Models for Robotics
Dr. Chen, Dr. Karl, and Elie Aljalbout are affiliated with Volkswagen; Dr. Zeng is affiliated with Google; Dr. Mayol-Cuevas is affiliated with Amazon. All other Topic Editors declare no competing interests with regard to the Research Topic subject.
Keywords: Large language models, Robot learning, Affordance learning, Cognitive robotics, Language Acquisition, Physics-Informed learning, Transformers, Language Grounding
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.
