Universal Method for Enhancing Dynamics in Neural Networks via Memristor and Application in IoT-Based Robot...

Area of Science:

• Robotics
• Artificial Intelligence
• Neuroscience

Background:

• Mobile robots require advanced navigation and mapping for extreme environments.
• Memristive neural networks (MNNs) offer chaotic dynamics for robot control.
• Existing MNNs lack expandability for diverse robotic applications.

Purpose of the Study:

• To propose a universal method for enhancing neural network dynamics.
• To generate diverse MNNs with rich dynamics for IoT robot navigation and security.
• To explore expandable MNNs for varied application scenarios.

Main Methods:

• Enhancing neural network dynamics by increasing memristive elements, neurons, and integration.
• Deriving multiple memristive central cyclic neural networks (MCCNNs) from a central cyclic neural network.
• Numerically investigating MCCNN dynamics: bifurcation, multistability, and amplitude control.
• Verifying MCCNN feasibility through analog circuit and digital hardware implementation.

Main Results:

• Successfully derived and investigated various MCCNN dynamics.
• Demonstrated the physical existence and feasibility of MCCNNs.
• Applied MCCNNs to drive an IoT-based mobile robot.
• Validated superior performance in area coverage and obstacle avoidance.

Conclusions:

• The proposed method effectively enhances neural network dynamics for robotics.
• MCCNNs provide a versatile and reliable solution for IoT robot navigation and security.
• Experimental results confirm the practical feasibility and effectiveness of MCCNNs in real-world applications.