Building egocentric models of local space from retinal input.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Cognitive Science

Background:

• Object localization is crucial for real-world interaction and motor control.
• The brain must convert visual input from a retinocentric (eye-centered) frame to an egocentric (body-centered) frame to guide actions.
• This transformation is vital for tasks like navigating complex environments and avoiding obstacles.

Purpose of the Study:

• To review the anatomical, physiological, and computational mechanisms of retinocentric-to-egocentric reference frame transformations.
• To explore how distance estimation contributes to 3D spatial representations.
• To compare implementations across biological and artificial neural networks and discuss the maintenance of internal spatial models.

Main Methods:

• Review of existing literature on reference frames in diverse organisms (insects to primates).
• Analysis of computational models and physiological studies of neural circuits.
• Comparison of findings across different nervous systems and behaviors.

Main Results:

• Evidence for both retinocentric and egocentric reference frames across species.
• The role of distance estimation in building 3D spatial maps.
• Proposed neural network implementations for reference frame transformation.
• Discussion on the persistence of egocentric models independent of sensory input.

Conclusions:

• Reference frame transformation is a fundamental neural computation for spatial cognition and goal-directed behavior.
• Investigating this process across various systems offers insights into canonical neural computations.
• Further research is needed to fully elucidate the neural basis of maintaining internal spatial models.