Solving navigational uncertainty using grid cells on robots.

Area of Science:

• Neuroscience
• Computational Biology
• Robotics

Background:

• Mammalian navigation relies on path integration and landmark calibration, both prone to uncertainty.
• Existing models struggle to explain how animals form coherent spatial representations amid environmental ambiguity.

Purpose of the Study:

• To investigate the role of conjunctive grid cells in resolving navigational uncertainty.
• To explore how multiple pose estimates are maintained and propagated in spatial representation.

Main Methods:

• Utilized the RatSLAM (Robot AnTicipation SLAM) computational model, inspired by rodent spatial cells and engineering principles.
• Replicated seminal experiments on rodent grid cells (Moser experiments).
• Applied the model to a novel paradigm simulating perceptual ambiguity in navigation.

Main Results:

• Demonstrated that conjunctive grid cells can maintain and propagate multiple pose estimates.
• Observed short-term population coding of multiple location hypotheses via computational modeling.
• Showed that correct pose estimation can be achieved over time despite ambiguous sensory cues.

Conclusions:

• Conjunctive grid cells play a critical role in filtering sensory uncertainty during navigation.
• The RatSLAM model provides a framework for understanding how the brain handles ambiguity in spatial representation.
• This research opens avenues for experimental testing of navigation in ambiguous environments.