Recursive Hierarchical Particle Swarm Optimization for Signal Processing

The patent implements a hierarchical structure where multiple PSO swarms are organized in nested levels. Each swarm operates at a different hierarchical level, with lower-level swarms exploring local regions and higher-level swarms coordinating global search. This nested arrangement allows the system to maintain simplicity at each individual level while achieving global optimization accuracy through the coordinated interaction of multiple levels.

The patent divides the population into multiple swarms segmented by hierarchical level rather than using a single homogeneous swarm. Each swarm is segmented to focus on specific regions or aspects of the solution space, with control parameters adjusted according to hierarchical position. This segmentation enables parallel exploration of different regions while maintaining overall coordination, resolving the contradiction between implementation simplicity and optimization accuracy.

The patent implements periodic hierarchical coordination where lower-level swarms perform local exploration for predetermined iterations, then periodically synchronize with higher-level swarms to share information and adjust search directions. This periodic action pattern allows extensive exploration to occur in controlled intervals rather than continuously, reducing overall computational time while maintaining the ability to find global optima through periodic global coordination.

The patent employs preliminary local exploration by lower-level swarms before higher-level swarms initiate global search. This preliminary action allows the system to quickly identify promising regions through simple local searches, then focus computational resources on global optimization only in those identified regions. This staged approach reduces total computational time while maintaining optimization accuracy by avoiding exhaustive search of entire solution space.

The patent applies different attraction force strengths and update rule parameters tailored to each hierarchical level and local region. Lower-level swarms operating in local regions use stronger attraction forces for quick local convergence, while higher-level swarms use weaker forces to maintain exploration capability. This localized customization of control parameters allows fast convergence in appropriate contexts without sacrificing global solution quality.

The patent implements dynamic adjustment of control parameters including attraction forces and update rules based on hierarchical level and iteration progress. Control parameters are not fixed but dynamically adapted during optimization, allowing the system to transition from exploration-dominated early stages to exploitation-dominated later stages. This dynamic behavior enables the system to achieve both fast convergence and high solution quality by adjusting forces appropriately at different stages.