Self-Regulated Particle Swarm Multi-Task Optimization.

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Evolutionary Computation

Background:

• Population-based search techniques are effective for complex optimization but computationally expensive.
• Unmanned Aerial Vehicle (UAV) path planning is a large-scale optimization problem with many constraints.
• Optimizing multiple UAV problems simultaneously is time-consuming with traditional methods.

Purpose of the Study:

• To develop a novel Evolutionary Multi-Task Optimization (EMTO) algorithm.
• To improve the efficiency of solving multiple optimization problems concurrently.
• To enhance existing Particle Swarm Optimization (PSO) algorithms through knowledge transfer.

Main Methods:

• Developed a novel EMTO algorithm based on a classical PSO algorithm.
• Implemented a knowledge transfer strategy that synthesizes information from component tasks during velocity updates.
• Created two versions of the algorithm, each with a distinct knowledge transfer strategy.
• Compared the proposed algorithm against multifactorial PSO, SREMTO, MFEA, and classical PSO.

Main Results:

• The proposed EMTO algorithm demonstrated superior performance compared to existing methods.
• The knowledge transfer strategy effectively improved the optimization process.
• The algorithm showed strong results on both single-objective and multi-task optimization problems.

Conclusions:

• The novel EMTO algorithm with knowledge transfer offers a more efficient approach to solving complex optimization problems.
• This method has significant potential for real-world applications, including UAV path planning.
• The developed knowledge transfer strategies are effective in improving multi-task optimization performance.