Particle swarm optimization (PSO) is a bio-inspired stochastic optimization algorithm that simulates the foraging behavior of birds. Despite its simplicity and efficiency, PSO often suffers from premature convergence and a poor balance between exploration and exploitation. These drawbacks mainly arise from its limited learning sources and rigid position update scheme. To address these issues, this paper proposes an enhanced PSO framework, termed Exemplar Learning and Memory Retrieval-Based Particle Swarm Optimization (EMPSO). The design of EMPSO is inspired by the learning, memory, and adaptation mechanisms observed in biological collectives. It integrates three complementary strategies to improve swarm intelligence. First, an elite exemplar learning mechanism aggregates the positional information of top-performing particles to construct a more reliable guidance vector. Second, a memory recall strategy retains exemplars that have recently contributed to global improvements and reuses them probabilistically with a recency bias, thus enabling effective knowledge inheritance. Third, an adaptive position update scheme assigns exploration- or exploitation-oriented behaviors to particles based on fitness ranking, promoting dynamic role differentiation within the swarm. Comprehensive experiments on the CEC2017 and CEC2022 benchmark suites demonstrate that EMPSO consistently outperforms six representative algorithms. Furthermore, applications to three engineering design problems and the optimal PMU placement task verify its robustness and practical effectiveness.

Particle swarm optimization (PSO) is a widely used bio-inspired optimization algorithm, yet maintaining an effective balance between exploration and exploitation remains challenging. Most existing PSO variants rely on static or predefined regulation strategies, which restrict their adaptability to evolving search states and may lead to premature convergence or search stagnation. Inspired by division of labor and competitive selection mechanisms in biological populations, this paper proposes a dual-subpopulation competitive particle swarm optimization (DCPSO). In DCPSO, the population is explicitly partitioned into exploration and exploitation subpopulations with distinct search roles. A dynamic competition mechanism is designed to evaluate recent search progress, based on which stagnated particles are adaptively migrated between subpopulations, enabling flexible reallocation of computational resources during the optimization process. Experimental results on the CEC2017 benchmark suite demonstrate that DCPSO consistently outperforms standard PSO and several representative state-of-the-art algorithms, achieving statistically significant improvements on the majority of benchmark functions, particularly on hybrid and composition problems. Additional experiments on engineering design problems further verify the robustness, convergence stability, and practical effectiveness of DCPSO.

Focusing on the autonomous location method of lower extremity exoskeleton robot in complex environment, this paper presents a gait type recognition method based on support vector machine (SVM)C and an autonomous location method based on inertial information mapping model and system reconstruction. In this paper, support vector machine is used to effectively recognize various kinds of conventional gait types of exoskeleton robots. The inertial information mapping models among different parts of the lower limbs of the robots under different gaits are established respectively. Aiming at the problem of the failure of the inertial measurement unit at the end of the limb in the impact or high overload motion, a robust autonomous location method based on system reconstruction is studied. The experiment results show that, using different neural network model parameters under different gaits can be used to reduce the complexity of the network model. While the inertial measurement unit at the end of the limb fails, the location performance of the exoskeleton robot navigation system based on this method, is equal to that of the inertial navigation system at the end of the limb with same sensor precision.

Particle swarm optimization (PSO) has been widely used in engineering optimization because of its simple structure and easy implementation. However, standard PSO and most of its variants mainly learn from the personal best position and the global best position. Thus, they often fail to preserve and reuse population-level knowledge generated during the search process. This problem becomes more evident when the search state changes or the swarm falls into stagnation, at which point useful search information may be ignored or forgotten. To address this issue, this paper proposes a state-adaptive knowledge recall PSO algorithm, termed SKRPSO. It includes three cooperative components. First, a state-aware adaptive aggregation mechanism adjusts the elite knowledge-pool size according to population dispersion and builds a rank-weighted knowledge vector for stable population-level guidance. Second, a stagnation-driven knowledge recall mechanism stores historical knowledge associated with global improvements in a bounded memory buffer and recalls recently successful knowledge with a time-decay preference when stagnation is detected. Third, a knowledge-fusion position update strategy uses current aggregated knowledge during normal search and recalled knowledge under stagnation, balancing local exploitation and stagnation escape. Experiments on the CEC2017 benchmark suite show that, based on 30 independent runs, SKRPSO achieves the best mean error on 22 of 29 functions and the best overall Friedman average rank of 1.431 among all compared algorithms. Engineering design results further indicate stable performance.

The non-linearity and uncertainty of air pollutant data add a certain degree of difficulty to the prediction of air quality, and neural networks have a certain ability to solve this problem. But it has the shortcomings of slow convergence and is easy to fall into local minimums. Therefore, this paper optimizes the OIF Elman neural network (IOIF Elman) by the method of adaptive learning rate and direct input to output connections. And an air quality prediction model is established based on the IOIF Elman neural network. The simulation results show that the absolute average error (AAE) and accuracy of the IOIF Elman air quality prediction model are better than those of the OIF Elman and Elman neural network. The establishment of this model can better meet the needs of practical applications and provide a more reliable basis for decision-making for environmental protection departments.

Existing heuristic algorithms are based on inspiration sources and have not yet done a good job of basing themselves on optimization principles to minimize and utilize historical information, which may lead to low efficiency, accuracy, and stability of the algorithm. To solve this problem, a personalized-template-guided intelligent evolutionary algorithm named PTG is proposed. The core idea of PTG is to generate personalized templates to guide particle optimization. We also find that high-quality templates can be generated to guide the exploration and exploitation of particles by using the information of the population particles when the optimal value remains unchanged, the knowledge of population distribution changes, and the dimensional distribution properties of particles themselves. By conducting an ablation study and comparative experiments on the challenging CEC2022 test and CEC2005 test functions, we have validated the effectiveness of our method and concluded that the stability and accuracy of the solutions obtained by PTG are superior to other algorithms. Finally, we further verified the effectiveness of PTG through four engineering problems.