Emergency evacuation in densely populated facilities remains a pressing concern where human behavior under stress directly influences evacuation outcomes. A behaviorally-informed adaptive quantum ant colony optimization (BIAQACO) framework is proposed that integrates psychological stress coefficients and behavioral adaptation mechanisms into multi-objective evacuation planning. The model minimizes total evacuation duration while managing psychological congestion through anxiety-weighted path selection. An adaptive quantum rotation strategy adjusts exploration-exploitation balance based on real-time behavioral feedback, while mutation operators prevent premature convergence in complex psychological state spaces. Experiments across benchmark traveling salesman instances and campus evacuation scenarios validate framework effectiveness. Results demonstrate BIAQACO achieves 36.58% faster evacuation times and 43.23% lower psychological stress accumulation compared to the best-performing baseline algorithm under high-density conditions with one thousand evacuees.
Modern power systems face new challenges in maintaining stability and real-time balance with the large-scale integration of intermittent renewable energy. To address the persistent issue of insufficient participation in demand-side management, this study proposes a three-tier Stackelberg game coordination framework for Virtual Power Plants (VPPs) and Price-Based Demand Response (PBDR). To efficiently solve the high-dimensional nested multilevel optimization problem, a Hierarchical Nested Particle Swarm Optimization (HN-PSO) algorithm is developed. Theoretical analysis and simulation results demonstrate the existence and uniqueness of the global equilibrium in the proposed game model. A 24-hour simulation further shows that the VPP can mitigate price volatility while balancing internal dispatchable generation and external procurement. Therefore, the proposed approach can provide an efficient and practical computational method for decentralized market dispatch under renewable energy uncertainty.
The Honey Badger Algorithm (HBA) is a recent addition to the growing family of nature-inspired optimization techniques within the domain of Swarm Intelligence (SI). Inspired by the intelligent foraging behavior of honey badgers, HBA has demonstrated remarkable effectiveness in solving complex optimization problems. This review provides an analysis of HBA, exploring its core principles, applications (particularly within the engineering domain), and the evolution of the algorithm through various modified and hybrid versions. The authors further leverage bibliometric analysis to unveil research trends and offer valuable insights into the current state of HBA research. Finally, this study provides insights for future advancements and applications in diverse problem-solving tasks while harnessing the HBA.
The Squirrel Search Algorithm (SSA) has garnered considerable attention from the academic community and is increasingly applied across multiple disciplines to tackle a variety of optimization challenges since 2019. Inspired by the dynamic foraging behavior of southern flying squirrels, the SSA mimics their energy-efficient gliding locomotion—a special adaptive strategy that enables these small mammals to traverse large distances with minimal effort. This study offers a comprehensive analysis of the original SSA, along with its modifications, hybridizations, and variants. Furthermore, this study elucidated the practical applications of SSA across various engineering disciplines and incorporated analyses of studies demonstrating suboptimal SSA performance. Finally, the authors discuss future research directions to further enhance the SSA’s performance.
Predicting the short-term state of the Limit Order Book (LOB) is crucial in quantitative finance, yet challenging due to its high-dimensional and noisy nature. While deep learning models like Transformers excel at capturing temporal patterns, they often rely on suboptimal hyperparameters and lack robustness. This paper introduces PSO-Transformer, a hybrid framework for high-frequency LOB prediction. It combines a modified Particle Swarm Optimization (PSO) with adaptive inertia and a Temporal Attention Network. The framework operates through three dedicated modules: data preprocessing, PSO-based hyperparameter optimization, and temporal attention forecasting. Experiments on a large-scale LOB dataset show that PSO-Transformer outperforms state-of-the-art benchmarks, achieving an average Matthews Correlation Coefficient (MCC) of 0.612, an increase of 15.9% compared to the base line Transformer. Ablation and sensitivity studies validate each component's contribution and reveal emergent optimal model structures.
Constrained optimization problems are widely prevalent, with evolutionary algorithms serving as predominant solution approaches. However, these algorithms exhibit inadequate adaptability and lack causal interpretability due to excessive dependence on predefined heuristic rules. This paper proposes an evolutionary constrained optimization framework driven by doubly robust causal learning, which quantifies the causal interaction between objective optimization and constraint satisfaction through the use of causal random forests, and designs a dynamic adaptive strategy-switching mechanism to autonomously select constraint priority, objective priority, and their corresponding complete comparison strategies. Comprehensive experiments conducted on mainstream benchmark suites demonstrate that the proposed framework outperforms state-of-the-art algorithms in convergence speed, solution quality, and stability, while exhibiting enhanced robustness in high-dimensional and strongly constrained scenarios.
Differential Evolution (DE) often faces a critical challenge in striking an effective balance exploration and exploitation when tackling complex optimization problems. To address this issue, this paper proposes a novel Elite-Transition-Potential (ETP) model based adaptive multi-population multi-mutation DE algorithm, named ETPDE. This algorithm employs the ETP model to dynamically partition each generation's population into three complementary subpopulations, and tailors differentiated mutation strategies to them to further reinforce their respective roles, thereby enabling the algorithm to dynamically adapt to the search requirements of different evolutionary stages. Furthermore, ETPDE incorporates a population size reduction method, along with an adaptive size control strategy for these three subpopulations, dynamically adjusting their proportional distribution within the entire population during evolution. Experiments are conducted on the CEC 2017 benchmark suite and the Lennard-Jones potential problem, and the results indicate that ETPDE exhibits competitive performance.
The Honey Badger Algorithm (HBA) is a recent addition to the growing family of nature-inspired optimization techniques within the domain of Swarm Intelligence (SI). Inspired by the intelligent foraging behavior of honey badgers, HBA has demonstrated remarkable effectiveness in solving complex optimization problems. This review provides an analysis of HBA, exploring its core principles, applications (particularly within the engineering domain), and the evolution of the algorithm through various modified and hybrid versions. The authors further leverage bibliometric analysis to unveil research trends and offer valuable insights into the current state of HBA research. Finally, this study provides insights for future advancements and applications in diverse problem-solving tasks while harnessing the HBA.
Appropriate model hyperparameters, transfer learning strategies, and loss balancing are still challenging to discover, while deep learning has achieved remarkable progress in picture segmentation and classification. An integrated encoder-decoder network is improved in this transfer learning system with the application of Particle Swarm Optimization (PSO). To get hierarchical data, a pretrained encoder is used, and lightweight segmentation and classification heads learn two tasks at once. By default, PSO, a universal meta-optimizer, changes learning rates, layer-freezing masks, batch sizes, and loss-weighting coefficients. Manual searches are accelerated in this way. With this hybrid approach, we may improve the accuracy-efficiency trade-off by combining global exploration with gradient descent and Particle Swarm Optimization (PSO). Cityscapes, PASCAL VOC, and CIFAR simulations demonstrate that swarm-based meta-learning has the potential to increase the flexibility of transfer learning, leading to more efficient and less expensive models for picture processing.
Government subsidies for renewable energy consumption and Renewable Portfolio Standards (RPS) are critical issues in the power supply chain. To address these issues, this study considers three decision-making scenarios: a centralized supply chain, a decentralized supply chain, and a coordinated supply chain employing revenue-sharing contracts between generators and retailers. Moreover, an improved Particle Swarm Optimization (PSO) algorithm is utilized to solve the three proposed models. The results indicate that wholesale prices for both renewable and fossil-fueled electricity are higher under the decentralized decision-making scenario than under the revenue-sharing contract mechanism. Furthermore, the retailer achieves the highest procurement volumes for both types of electricity and obtains maximum profit under the revenue-sharing contract mechanism. From the perspective of the overall supply chain, the revenue-sharing contract mechanism effectively mitigates the double marginalization effect inherent in the decentralized power supply chain.
Predicting online shoppers’ purchasing intention is crucial for e-commerce management. This study proposes SHAP-IGWO-EM, a hybrid ensemble framework to enhance prediction accuracy and interpretability for high-dimensional, imbalanced behavioral data. The method integrates XGB, LGBM, and GBM within a stacking architecture, optimized via an enhanced Grey Wolf Optimizer featuring a nonlinear convergence factor, improved leader selection, and Lévy flight perturbations to escape local optima. SHAP analysis identifies key features, reducing redundancy and enabling transparent model interpretation. Experiments on a public e-commerce dataset demonstrate SHAP-IGWO-EM’s superior performance over baseline and ensemble models. Ablation studies validate the optimization strategy and SHAP-based feature selection. By uncovering critical drivers of purchase behavior, the model not only boosts predictive power but also provides actionable insights for personalized marketing and operations.
5G private power networks require high levels of real-time performance and reliability; however, traditional threshold-based scaling methods struggle with the unpredictable bursts and periodic fluctuations in traffic. This paper proposes a BiLSTM-Attention model that incorporates Swarm Intelligence-based Multi-Agent Collaborative Prediction (SI-MACP), named BiLSTM-Attention-SI-MACP, for resource forecasting in these networks. The proposed framework combines bidirectional LSTM networks with attention mechanisms to capture multivariate resource dependencies. Meanwhile, the SI-MACP mechanism utilizes principles of swarm intelligence through distributed task decomposition, privacy-preserving local modeling, and the collaborative aggregation of parameters and predictions. Experimental results demonstrate significant improvements, including 32% MAE reduction, 35% higher resource utilization, and 99.9% SLA compliance, particularly benefiting latency-sensitive services like differential protection.
Unmanned Aerial Vehicles (UAVs) have been widely used for power line inspection tasks. However, UAVs are typically constrained by the limited endurance and payload capacity, which restricts their ability to perform long-range inspections and deliver large quantities of materials in a single mission. To address the above issues, a fairness-aware multi-objective multi-UAV path planning (FAMO-MUPP) method is proposed to facilitate both power tower inspection and the delivery of essential materials. The multi-UAV power tower inspection problem is decomposed into two sub-problems: path planning and charging. Accordingly, two different improved NSGA-II algorithms are used to separately evolve the routing population and the charging population. Additionally, this study introduces an information interaction method to achieve information completion and mutual evaluation between the two populations. The experimental results demonstrate that the proposed FAMO-MUPP outperforms two competitors on seven test scenarios.
Predicting residual energy in communication nodes of private power networks is crucial for maintaining stable power grids, yet current methods often fail to capture the complex relationships between periodic features and sudden energy fluctuations in power time-series data. This article proposes a novel prediction model integrating TimesNet and a swarm intelligence-based adaptive attention reweighting (SI-AAR) mechanism. The model employs channel-independent slicing to encode heterogeneous data, extracts dynamic patterns through period matrix reconstruction and Inception convolution, and dynamically allocates attention weights using neighboring node information via the SI-AAR module to enhance spatial anomaly detection. Experimental results on real datasets demonstrate that the proposed method reduces Mean Squared Error (MSE) and Mean Absolute Error (MAE) by 8.18% and 11.19%, respectively, compared to TimesNet. Ablation studies highlight the SI-AAR module's significant contribution, achieving a 7.66% reduction in MSE.
The precise feeding system with horizontal screw feeder is currently experiencing feeding accuracy, uneven water-to-feed mass ratios, and dust pollution in pig farming problems. To address these issues, a precise feeding system with horizontal screw feeder based on an Active Disturbance Rejection Controller (ADRC) is designed. The system is constructed by a cloud-edge architecture, which consists of an edge precision feeding device and an intelligent gateway, integrating spiral feed control, synchronized water-to-feed ratio adjustment, and feed intake monitoring functions. The edge feeding device employs an adaptive ADRC to compensate for feed discharge disturbances in real time. The gateway uses the Differential Evolution-Grey Wolf Optimizer (DE-GWO), a hybrid swarm intelligence algorithm, to adaptively tune ADRC parameters. Simulation and field tests indicate that the system's average feed dispensing error is 0.25% (maximum error 1.5%), with a feed-to-water ratio error of approximately 2%, validating its effectiveness in improving farming efficiency.
To address the imbalance among riders, platforms, and merchants in food delivery within the context of Smart Cities, this study proposes an Intelligent Food Delivery model incorporating Dynamic Weather Conditions. A multi-objective optimization framework considering customer satisfaction, rider delivery pressure, and platform efficiency is established. An Improved Sparrow Search Algorithm (ISSA), enhanced by Sin-chaotic, Golden Sine, Adaptive t-distribution, and Dynamic Selection mechanisms, demonstrates superior convergence speed and accuracy. Simulations under Clear, Rainy, and Snowy scenarios validate ISSA’s effectiveness in reducing rider pressure and optimizing delivery routes, providing robust theoretical guidance for intelligent logistics management in Smart Cities.
Fruit picking in complex orchard environments is limited by the low detection and recognition accuracy due to clustered background, illumination variation, and partial occlusion. To improve the picking performance, a hierarchical visual grasping architecture based on You Look Only Once (YOLO) algorithm and adaptive error compensation is proposed. The upper layer uses YOLOv5 and inverse kinematics to recognize and localize the target. Ant colony optimization is specifically used for hyperparameter tuning of YOLOv5 to improve the detection accuracy. The middle layer dynamically compensates for the output torque of the joint actuator through the feedback linearization method. The lower layer finishes precise grasping through nonlinear mapping model between the pulse width modulation signal and the servo angle. Experimental results validate that the proposed architecture outperforms traditional methods by reaching 98.2% and 97.7% recognition accuracy in obstacle-free scenarios and complex environments, a higher grasping success rate, and a lower positioning deviation.
Insulator defect detection is essential for ensuring the safety and reliability of power transmission lines. Although deep learning models like YOLOv7 achieve high accuracy, their manually designed architectures tend to be computationally heavy and unsuitable for deployment on resource-limited edge devices. To tackle these issues, the authors propose EHANN-NAS, a neural architecture search framework optimized for real-time insulator defect detection on edge platforms. EHANN-NAS employs a Graph Convolutional Network (GCN)-based surrogate model to efficiently predict the performance of candidate architectures without full training, significantly reducing search time and computational cost. Furthermore, it integrates an enhanced Particle Swarm Optimization (PSO) approach featuring uncertainty-aware sampling and adaptive diversity mechanisms to better explore the architecture space and identify lightweight yet effective network designs. Experiments on a real-world dataset show that EHANN-NAS achieves a mean Average Precision (mAP0.5) of 95.1%, outperforming state-of-the-art methods.
This paper proposes a novel framework, evolutionary intelligent multiple neural networks (EIMNN), for college student mental health analysis and guidance. EIMNN integrates modality-specific neural sub-networks, each tailored to textual content, temporal behavior, and peer interaction signals, within an evolutionary neural architecture search paradigm. Inspired by principles of swarm intelligence, the framework employs a cooperative co-evolution strategy, where multiple neural networks evolve in parallel and exchange structural knowledge to optimize multimodal representations. A co-evolved attention mechanism adaptively fuses outputs from these networks based on individual context. Furthermore, the authors introduce a reinforcement-guided psychological state transition model, which learns to anticipate emotional trajectories and supports proactive mental health guidance. Extensive experiments on a large-scale, multi-modal social media dataset collected from 3,500 university students demonstrate that EIMNN outperforms state-of-the-art baselines in terms of many metrics.
Traditional deep learning methods have excelled in power equipment defect detection, but their high computational requirements hinder real deployment on edge devices. To address this challenge, this article introduced particle swarm optimization for retraining-free mixed-precision quantization (PSOQ), a novel post-training quantization framework enabling retraining-free mixed-precision quantization, through layer importance-guided particle swarm optimization (PSO) over a one-shot trained Supernet. Specifically, this framework constructed a Supernet using Monte Carlo sampling and interference-aware bit-width scheduling for fast, accurate evaluation of mixed-precision quantization configurations—without fine-tuning. Experiments on benchmark and insulator defect datasets demonstrated that PSOQ, when applied to various network architectures, significantly reduced computational and storage overhead, while maintaining detection accuracy.
