The central nervous system enables autonomous agents to dynamically interact with their environment. The brain does thus not operate in a vacuum, but through its sensory-motor apparatus, constituting a closed-loop with the surroundings in which it is embedded. This has important implications for the study of action and perception and their interrelatedness with the ever-changing environment, as one can only be fully understood in context of the others. For example, due to the sharp drop-off in visual acuity with eccentricity, eye-movements are required for object recognition while at the same time the perceptual system needs to correct for displacement of images caused by these very eye-movements and disentangle them from object motion. However, such brain-body-environment interactions have largely been ignored in computational neuroscience. This is partially due to a focus on biological realism which, though ultimately essential, mainly limited research to the use of detailed neuro-models tackling toy problems in simplified environments. At the same time, few attempts have been made in neuroscience to embody functionally (rather than biologically) complex models in challenging environments and evaluate their performance on realistic sensory-motor tasks. A solution to the former problem might come in the form of deep (reinforcement) learning which offers a more teleological, performance-oriented, approach to study action, perception and their interaction. The latter problem can be resolved using recently developed virtual environments (such as the Human Brain Project’s neurorobotics platform) that jointly simulate brain systems, their embodiment, and naturalistic surroundings.
With this Research Topic, we would like to highlight how the fields of neurocomputational modeling, deep learning, and robotics can enrich neuroscientific research on functionally relevant sensory-motor interactions in dynamic surroundings.
Relevant topics
• Artificial neural networks & deep learning
• (neuro-) Robotics
• Control theory
• Dynamical systems theory
• Sensory-motor integration mechanisms
• Action selection
• Visual stability
• Cross-saccadic information integration (for object recognition)
The central nervous system enables autonomous agents to dynamically interact with their environment. The brain does thus not operate in a vacuum, but through its sensory-motor apparatus, constituting a closed-loop with the surroundings in which it is embedded. This has important implications for the study of action and perception and their interrelatedness with the ever-changing environment, as one can only be fully understood in context of the others. For example, due to the sharp drop-off in visual acuity with eccentricity, eye-movements are required for object recognition while at the same time the perceptual system needs to correct for displacement of images caused by these very eye-movements and disentangle them from object motion. However, such brain-body-environment interactions have largely been ignored in computational neuroscience. This is partially due to a focus on biological realism which, though ultimately essential, mainly limited research to the use of detailed neuro-models tackling toy problems in simplified environments. At the same time, few attempts have been made in neuroscience to embody functionally (rather than biologically) complex models in challenging environments and evaluate their performance on realistic sensory-motor tasks. A solution to the former problem might come in the form of deep (reinforcement) learning which offers a more teleological, performance-oriented, approach to study action, perception and their interaction. The latter problem can be resolved using recently developed virtual environments (such as the Human Brain Project’s neurorobotics platform) that jointly simulate brain systems, their embodiment, and naturalistic surroundings.
With this Research Topic, we would like to highlight how the fields of neurocomputational modeling, deep learning, and robotics can enrich neuroscientific research on functionally relevant sensory-motor interactions in dynamic surroundings.
Relevant topics
• Artificial neural networks & deep learning
• (neuro-) Robotics
• Control theory
• Dynamical systems theory
• Sensory-motor integration mechanisms
• Action selection
• Visual stability
• Cross-saccadic information integration (for object recognition)
