Research Topic

Scaling up Robotic Intelligence by Combining Symbolic AI with Deep Learning

  • Submission closed.

About this Research Topic

The recent wave of impressive advancements in deep learning cannot hide that their scope is still limited. Neural networks are impressive in classifying objects or to handle a limited set of well-defined control tasks. Deep neural network approaches shine in learning complex patterns in input signals via hierarchical feature extraction, but they are extremely hungry in resources such as labelled data or computational power that are often very scarce in robotic application domains. The challenges when deploying these networks from simulation to the real world are indicative of the problems of scaling up neural network-based AI to realistic, dynamic and unpredictable environments.

An alternative strand of research in robotic AI is focused on symbolic principles, where the world is casted in logical predicates, rules and conditional relationships to plan actions or to reason on context and experience. This knowledge and the associated rules must be hand-coded by experts, which makes symbolic AI better interpretable and predictable than “black-box” deep neural network components, which can be very necessary properties for robotic applications, e.g. in order to avoid physical damage to nearby humans. Drawbacks of symbolic AI are how to ground sensor input to symbols and the complexity of providing formal descriptions of complex, realistic environments.

With this Research Topic, we call for papers that will help in understanding the added value of combining principles from symbolic AI with the most recent advances in deep neural networks. We seek to understand where is the added value, and how synergies should be achieved between both branches in artificial intelligence.

Topics of interest include, but are not limited to:
• Deep symbolic reinforcement learning
• Symbol grounding of raw robotic sensor input
• Task representation, classification and transfer
• Explainability of neural networks for robotic applications
• Robotic planning and control algorithms
• Gestures and natural language processing and generation in social robotics
• Representation and storage of knowledge bases and experience pools
• Goal and intention recognition
• Intrinsically motivated and active learning


Keywords: Symbolic AI, Deep Learning, Robot Planning and Control, Adaptive Robots, Explainable AI


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.

The recent wave of impressive advancements in deep learning cannot hide that their scope is still limited. Neural networks are impressive in classifying objects or to handle a limited set of well-defined control tasks. Deep neural network approaches shine in learning complex patterns in input signals via hierarchical feature extraction, but they are extremely hungry in resources such as labelled data or computational power that are often very scarce in robotic application domains. The challenges when deploying these networks from simulation to the real world are indicative of the problems of scaling up neural network-based AI to realistic, dynamic and unpredictable environments.

An alternative strand of research in robotic AI is focused on symbolic principles, where the world is casted in logical predicates, rules and conditional relationships to plan actions or to reason on context and experience. This knowledge and the associated rules must be hand-coded by experts, which makes symbolic AI better interpretable and predictable than “black-box” deep neural network components, which can be very necessary properties for robotic applications, e.g. in order to avoid physical damage to nearby humans. Drawbacks of symbolic AI are how to ground sensor input to symbols and the complexity of providing formal descriptions of complex, realistic environments.

With this Research Topic, we call for papers that will help in understanding the added value of combining principles from symbolic AI with the most recent advances in deep neural networks. We seek to understand where is the added value, and how synergies should be achieved between both branches in artificial intelligence.

Topics of interest include, but are not limited to:
• Deep symbolic reinforcement learning
• Symbol grounding of raw robotic sensor input
• Task representation, classification and transfer
• Explainability of neural networks for robotic applications
• Robotic planning and control algorithms
• Gestures and natural language processing and generation in social robotics
• Representation and storage of knowledge bases and experience pools
• Goal and intention recognition
• Intrinsically motivated and active learning


Keywords: Symbolic AI, Deep Learning, Robot Planning and Control, Adaptive Robots, Explainable AI


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.

About Frontiers Research Topics

With their unique mixes of varied contributions from Original Research to Review Articles, Research Topics unify the most influential researchers, the latest key findings and historical advances in a hot research area! Find out more on how to host your own Frontiers Research Topic or contribute to one as an author.

Topic Editors

Loading..

Submission Deadlines

Submission closed.

Participating Journals

Loading..

Topic Editors

Loading..

Submission Deadlines

Submission closed.

Participating Journals

Loading..
Loading..

total views article views article downloads topic views

}
 
Top countries
Top referring sites
Loading..

Comments

Loading..

Add a comment

Add comment
Back to top
);