A Memristive Spiking Neural Network Circuit for Bio-Inspired Navigation Based on Spatial Cognitive Mechanisms

Area of Science:

• Neuromorphic Engineering
• Computational Neuroscience
• Robotics

Background:

• Cognitive navigation is vital for survival, enabling autonomous exploration.
• Current bio-inspired navigation often lacks neuromorphic implementation.
• Efficient, low-power navigation solutions are critical for autonomous systems.

Purpose of the Study:

• To propose a bio-inspired memristive spiking neural network (SNN) circuit for goal-oriented navigation.
• To enable online decision-making using reward-based learning in a neuromorphic system.
• To demonstrate a practical, energy-efficient solution for robotic navigation.

Main Methods:

• Developed a memristive SNN circuit with place cell, action cell, and reward-based learning modules.
• Implemented real-time spatial encoding using Poisson spiking.
• Quantized the SNN and deployed it on a memristive hardware platform.

Main Results:

• Achieved goal-oriented navigation with online decision-making capabilities.
• Demonstrated effective reward-based learning adaptable to delayed and sparse rewards.
• Reported approximately a 21x reduction in energy consumption during forward computation compared to digital systems.

Conclusions:

• The proposed bio-inspired memristive SNN offers a viable neuromorphic solution for robotic navigation.
• This approach significantly enhances energy efficiency in low-power navigation applications.
• The work provides a practical implementation for advanced autonomous systems.