A gradual neural-network approach for frequency assignment in satellite communication systems

Area of Science:

• Artificial Intelligence
• Computer Science
• Telecommunications Engineering

Background:

• Combinatorial optimization problems require simultaneous constraint satisfaction and goal function optimization.
• The frequency assignment problem in satellite communication is an NP-complete problem crucial for managing increasing demand.
• Minimizing co-channel interference is essential for efficient satellite communication system operation.

Purpose of the Study:

• To introduce a novel neural-network approach, the gradual neural network (GNN), for solving combinatorial optimization problems.
• To demonstrate the effectiveness of GNN in addressing the frequency assignment problem in satellite communication systems.
• To achieve both constraint satisfaction and goal function optimization simultaneously.

Main Methods:

• The proposed gradual neural network (GNN) utilizes NxM binary neurons for N-carrier-M-segment systems.
• A gradual expansion scheme of activated neurons is employed for optimization.
• Heuristic methods are integrated to ensure constraint satisfaction within the binary neural network.

Main Results:

• GNN successfully solved the frequency assignment problem, a typical combinatorial optimization challenge.
• The approach demonstrated efficient solutions for practical-sized satellite communication systems.
• GNN significantly outperformed existing algorithms in finding better solutions for 15 tested instances.

Conclusions:

• The gradual neural network (GNN) offers a powerful and efficient method for solving complex combinatorial optimization problems.
• GNN provides a superior approach to frequency assignment in satellite communications, enhancing system performance.
• The GNN's capability in finding optimal solutions surpasses current algorithms, paving the way for improved satellite network management.