RatInABox, a toolkit for modelling locomotion and neuronal activity in continuous environments

Area of Science:

• Computational neuroscience
• Systems neuroscience
• Robotics

Background:

• Generating synthetic data for studying brain function in spatial navigation is complex and lacks standardization.
• Existing methods are time-consuming and hinder reproducibility and comparison across studies.

Purpose of the Study:

• To introduce RatInABox, an open-source Python toolkit for modeling rodent locomotion and generating synthetic neural data.
• To provide a unified framework for creating realistic spatial navigation datasets for computational research.

Main Methods:

• Developed a Python toolkit with capabilities for environment construction (1D/2D, barriers, cues).
• Implemented a physically realistic random motion model fitted to experimental data.
• Enabled rapid online calculation of neural data for hippocampal cell types (place, grid, boundary vector, head direction cells).
• Provided a framework for custom cell types, network models, and motion trajectories.

Main Results:

• The toolkit generates spatially and temporally continuous, topographically sensitive motion and neural data.
• Default settings replicate rodent foraging behaviors, including velocity statistics and wall-following tendencies.
• Demonstrated utility in decoding position from neural data and solving reinforcement learning tasks.

Conclusions:

• RatInABox offers a standardized and efficient framework for generating synthetic data in spatial navigation research.
• The toolkit facilitates computational modeling, data analysis, and reproducibility in neuroscience.
• It is expected to significantly accelerate research into the neural basis of navigation.