Real-Time Task Assignment Approach Leveraging Reinforcement Learning with Evolution Strategies for Long-Term Latency...

Area of Science:

• Computer Science
• Artificial Intelligence
• Distributed Computing

Background:

• Fog computing offers ultralow latency crucial for time-sensitive Internet of Things (IoT) applications.
• Minimizing both service delivery and computation latency is essential for fog infrastructure.
• Computation latency is an internal factor manageable by the fog infrastructure.

Purpose of the Study:

• To propose a reinforcement learning approach for real-time task assignment in fog servers.
• To minimize the total computation latency in fog computing environments over the long term.

Main Methods:

• Utilizing evolution strategies within a reinforcement learning framework.
• Developing a real-time task assignment mechanism for fog servers.
• Implementing and evaluating the approach on heterogeneous computing platforms.

Main Results:

• Achieved an approximate 16.1% reduction in computation latency compared to existing methods.
• Demonstrated low computational complexity and effective parallel operation of the learning algorithm.

Conclusions:

• The proposed reinforcement learning approach effectively minimizes computation latency in fog computing.
• The algorithm's efficiency and parallelizability make it suitable for modern heterogeneous computing platforms.
• This method enhances the performance of time-sensitive IoT services by optimizing fog server task allocation.