Bio-Inspired Autonomous Learning Algorithm With Application to Mobile Robot Obstacle Avoidance.

Area of Science:

• Robotics
• Computational Neuroscience
• Artificial Intelligence

Background:

• Spiking Neural Networks (SNNs), the third generation of Artificial Neural Networks (ANNs), offer high processing power and mimic biological neural networks.
• Challenges remain in applying SNNs to real-world applications, particularly in robot control.
• Autonomous learning in robots is crucial for adaptability and complex task execution.

Purpose of the Study:

• To propose a biologically inspired autonomous learning algorithm for Spiking Neural Networks (SNNs).
• To address challenges in applying SNNs to robot control, specifically for obstacle avoidance.
• To introduce a novel reward generation mechanism for enhanced learning and decision-making.

Main Methods:

• Developed a bio-inspired autonomous learning algorithm utilizing reward-modulated spike-timing-dependent plasticity (STDP).
• Implemented a novel reward generation mechanism to provide signals for learning and decision-making.
• Evaluated the algorithm using a mobile robot obstacle avoidance task.

Main Results:

• The mobile robot demonstrated significant learning ability in obstacle avoidance.
• Experimental results confirmed successful obstacle avoidance after a period of learning trials.
• The proposed algorithm proved effective in a typical robotic task scenario.

Conclusions:

• The proposed reward-modulated STDP algorithm offers a viable method for autonomous learning in bio-inspired robots.
• This approach provides an alternative for designing robots with enhanced learning capabilities for practical applications.
• The study highlights the potential of SNNs in advancing robot control and autonomous systems.