A system-on-chip development of a neuro-fuzzy embedded agent for ambient-intelligence environments

Area of Science:

• Computer Science
• Artificial Intelligence
• Embedded Systems

Background:

• Ambient intelligence environments require intelligent agents for real-time control and adaptation.
• Existing agent-based approaches often demand significant computational resources.

Purpose of the Study:

• To develop and implement a computationally efficient neuro-fuzzy agent on a system-on-chip (SoC) for ambient intelligence.
• To achieve personalized and adaptive real-time control of environments.

Main Methods:

• Development of a hardware/software (HW/SW) architecture on a field-programmable gate array (FPGA).
• Implementation of an adaptive neuro-fuzzy inference system with piecewise multilinear behavior.
• Integration of a MicroBlaze processor (SW) and parallel IP cores for neuro-fuzzy modeling (HW).

Main Results:

• The developed SoC demonstrates high-performance, life-long control and adaptation in real-world ubiquitous computing environments.
• The neuro-fuzzy agent exhibits computational efficiency, scalability, and universal approximation capabilities.
• The system effectively anticipates the needs and desires of inhabitants in an ambient intelligence setting.

Conclusions:

• The SoC-based neuro-fuzzy agent provides an efficient and scalable solution for real-time environmental control in ambient intelligence.
• This approach retains the modeling capabilities of larger systems while offering superior performance and adaptability.
• The developed architecture is suitable for autonomous electronic devices in personalized and adaptive environments.