A Positioning Method Based on Place Cells and Head-Direction Cells for Inertial/Visual Brain-Inspired Navigation...

Area of Science:

• Neuroscience
• Robotics
• Artificial Intelligence

Background:

• Mammalian navigation relies on integrated visual and self-motion cues.
• Accurate positioning without satellite signals remains a challenge for autonomous systems.
• Brain-inspired engineering offers novel approaches to spatial representation and navigation.

Purpose of the Study:

• To develop a brain-inspired positioning method using multi-sensor input for satellite-denied environments.
• To model head-direction and place cells using continuous attractor neural networks (CANNs).
• To decode navigation parameters from neural network activity.

Main Methods:

• Established head-direction and place cell models based on continuous attractor neural networks (CANNs).
• Encoded visual and inertial sensor data into the neural network models.
• Decoded directional and positional information from population neuron firing patterns.

Main Results:

• The brain-inspired navigation model successfully integrated multi-sensor information.
• The model produced more accurate and stable navigation parameters compared to existing methods.
• Generated realistic motion paths based on decoded spatial information.

Conclusions:

• The proposed model demonstrates the efficacy of fusing diverse sensor data for brain-inspired navigation.
• This research advances the development of autonomous navigation systems in challenging environments.
• The study highlights the potential of CANNs for encoding and decoding spatial information.