This paper presents a novel method for crystal structure prediction using particle swarm optimization (PSO) within an evolutionary scheme. The method, implemented in the CALYPSO code, efficiently searches for stable and metastable structures of compounds based only on their chemical composition. Unlike genetic algorithms, PSO avoids the use of evolution operators such as crossover and mutation, and instead uses a stochastic global optimization approach inspired by the behavior of bird flocks. The method incorporates symmetry constraints to reduce the search space and computational cost, and a geometrical structure factor technique to eliminate similar structures during the search process. The PSO algorithm is particularly effective in finding global minima on the free energy surface, and has been successfully applied to predict the structures of various known systems, including elemental, binary, and ternary compounds with different bonding environments. The method involves four main steps: (1) generating random structures with symmetry constraints, (2) local optimization of the structures, (3) post-processing to identify unique local minima using a geometrical structure factor, and (4) generating new structures through PSO for iteration. The PSO technique enables the intelligent selection of cell sizes, significantly reducing computational cost. The method has been tested on various systems, including elements like lithium, carbon, silicon, and magnesium, as well as binary compounds such as silica, and ternary compounds like MgSiO3 and CaCO3. The results show that the method can efficiently predict both stable and metastable structures with high accuracy and success rate. The method is particularly effective in finding complex structures with high pressure, and has the potential to be applied to larger systems, such as nanomaterials, surfaces, and bio-materials. The PSO-based approach is shown to be a powerful and efficient technique for crystal structure prediction.This paper presents a novel method for crystal structure prediction using particle swarm optimization (PSO) within an evolutionary scheme. The method, implemented in the CALYPSO code, efficiently searches for stable and metastable structures of compounds based only on their chemical composition. Unlike genetic algorithms, PSO avoids the use of evolution operators such as crossover and mutation, and instead uses a stochastic global optimization approach inspired by the behavior of bird flocks. The method incorporates symmetry constraints to reduce the search space and computational cost, and a geometrical structure factor technique to eliminate similar structures during the search process. The PSO algorithm is particularly effective in finding global minima on the free energy surface, and has been successfully applied to predict the structures of various known systems, including elemental, binary, and ternary compounds with different bonding environments. The method involves four main steps: (1) generating random structures with symmetry constraints, (2) local optimization of the structures, (3) post-processing to identify unique local minima using a geometrical structure factor, and (4) generating new structures through PSO for iteration. The PSO technique enables the intelligent selection of cell sizes, significantly reducing computational cost. The method has been tested on various systems, including elements like lithium, carbon, silicon, and magnesium, as well as binary compounds such as silica, and ternary compounds like MgSiO3 and CaCO3. The results show that the method can efficiently predict both stable and metastable structures with high accuracy and success rate. The method is particularly effective in finding complex structures with high pressure, and has the potential to be applied to larger systems, such as nanomaterials, surfaces, and bio-materials. The PSO-based approach is shown to be a powerful and efficient technique for crystal structure prediction.