In recent years, microservices have gained increasing popularity for developing cloud and edge applications. One of the key enabling technologies is application containerization, which allows multiple containers to be deployed on a single physical node. This technology has become widely adopted in cloud computing. Although several approaches have been proposed for container-based microservice scheduling, they suffer from major drawbacks, including ineffective load balancing, high network transmission overhead, and low service reliability. To address these challenges, this study introduces an enhanced version of the particle swarm optimization (PSO) algorithm, named QuadI-MOPSO (Multi-Objective PSO with Four Indicators). This approach leverages the$\epsilon$-indicator to ensure controlled advancement, encourage diversity, and guide convergence toward the Pareto front. In addition, three complementary indicators-coverage (CA), hypervolume (HV), and a behavioral indicator called the Anger Score (AS)-are employed to provide a more accurate evaluation of trade-offs between solutions. Experimental results demonstrate that QuadI-MOPSO significantly improves performance by reducing load imbalance, minimizing network transmission, and enhancing service reliability, which makes it a compelling solution for enterprises seeking to optimize microservice placement. Consequently, the proposed methodology offers a robust and high-performance framework for optimizing container-based microservice scheduling in dynamic and heterogeneous cloud environments.

Human Action Recognition (HAR) is integral to various domains, particularly in intelligent transportation systems where real-time understanding of driver behavior is critical for road safety. Accurately recognizing driver actions is challenging due to inter-class similarity, viewpoint variation, and real-world occlusions. This paper proposes a multi-view HAR approach utilizing a Siamese neural network architecture, fine-tuned with DenseNet201, to learn discriminative feature embeddings across front and side perspectives. Evaluations on the challenging 3MDAD dataset demonstrate the robustness of our method against viewpoint and modality shifts, achieving 58.22

Traditional Ant Colony Optimization algorithms have demonstrated strong performance across various optimization problems. However, they are prone to premature convergence and their computational cost increases remarkably as the problem size does. These limitations are largely attributed to the construction-based approach, where solutions are rebuilt from scratch in each cycle, contrasting with local search methods that iteratively refine existing solutions. This paper introduces a novel adaptation of the Ant Colony Optimization metaheuristic to solve large-scale Traveling Salesman Problems. The proposed Ant-ILS modifies the Ant colony optimization metaheuristic from a construction-based to a local search-based approach. A pheromone-guided 2-Opt function is integrated to refine an initial population of solutions generated by the Nearest Neighbor heuristic. Additionally, a simple mechanism is proposed to mitigate stagnation during the local search process. Experimental evaluations on various TSPLIB instances highlight the algorithm’s superior performance, showcasing its competitiveness against state-of-the-art methods.

The multi-objective Time-Dependent Green Vehicle Routing Problem (MOTDGVRP) is examinated in this study. A mathematical model for the MOTDGVRP where distance, transportation time and fuel consumption are minimized is formulated. Calculation methods for the travel time and the fuel consumption across time periods with time-dependent speeds are presented. In the Time-dependent fuel consumption calculation function, distance, load, and time-dependent speed are considered simultaneously. We propose an approximate heuristic based on the Iterated Local Search (ILS) to generate the Pareto solutions. This heuristic is tested on a real-world case and the results show that the proposed heuristic can scientifically plan driving route for each vehicle, effectively reduce the total distribution costs, the vehicle fuel consumption and protect the environment.

In the pharmaceutical industry, the production of psychiatric drugs has been seriously disrupted since the appearance of COVID'19. For that, Demand Forecasting of psychiatric drugs is among the big challenges in this industry. The objective is to avoid an excess of stock and, at the same time, to ensure that a stock rupture does not occur. Based on analysis of psychiatric drugs data, we compare in this paper several forecasting techniques which are Exponential Smoothing, seasonal ARIMA (i.e. SARIMA), SARIMAX, enhanced with the integration of exogenous (explanatory) variables, and LSTM. Through all the done tests, we make a comparison study of the results to identify the most promising models.

Balancing diversity and convergence seems to be a difficult task when solving multi-objective optimization problems (MOPs). For addressing this issue, researchers’ interest has been drawn to hybrid approaches since this cooperation allows benefiting from the advantages of both approaches and gains better trade-offs overall. Considering such fact, this paper aims at introducing a hybrid approach as a synergy of a multi-directional Ant Colony Optimization algorithm with a Local Search method based on a weighted version of epsilon Indicator using a self-adaptive neighborhood operator coined as Indicator Weighted Based Local Search with Ant Colony Optimization (IWBLS/ACO) to handle the knapsack problem within the multi-objective framework. In IWBLS/ACO, initial solutions are created by the ant colony. Then, the enhancement phase is ensured by the local search procedure. The algorithm is evolving based on different configurations of the epsilon quality indicator through different weight vectors. Moreover, we propose in this work, a novel self-adaptive neighborhood operator which changes automatically and dynamically as the IWBLS algorithm runs. The proposed IWBLS/ACO was tested on widely used Multi-objective Multidimensional Knapsack Problem (MOMKP) instances and compared with powerful state-of-the-art algorithms. Experimental results highlight that the proposed approach can lead to finding a good compromise between exploration and exploitation.

In this paper, the fully cooperative multi-agent system is studied, in which all of the agents share the same common goal. The main difficulty in such systems is the coordination problem: how to ensure that the individual decisions of the agents lead to jointly optimal decisions for the group? Firstly, a multi-agent reinforcement learning algorithm combining traditional Q-learning with observation-based teammate modeling techniques, called TM_Qlearning, is presented and evaluated. Several new cooperative action selection strategies are then suggested to improve the multi-agent coordination and accelerate learning, especially in the case of unknown and temporary dynamic environments. The effectiveness of combining TM_Qlearning with the new proposals is demonstrated using the hunting game.

Reinforcement learning comprises an attractive solution to the multi-agent cooperation problem, due to its robustness for learning in unknown and uncertain environments. The objective of this paper is to provide learning capabilities to a group of autonomous agents in order to efficiently perform a cooperative foraging task in a distributed manner. Firstly, the D-DCM-MultiQ learning method, presented in [1], is evaluated. To overcome the shortcomings of this method, new cooperative action selection strategies are developed. A new exploration alternative, favoring least recently visited states, is also proposed. The conducted simulation tests indicate the efficiency of suggested improvements in the case of large, unknown and stationary environments.

To ensure complex systems reliability and to extent their life cycle, it is crucial to properly and timely prognose faults. In this context, this paper describes a new intelligent approach to estimate the remaining useful life in complex systems. This approach is based on the combination of several intelligent techniques. This approach is based on Echo State Network (ESN) and Particle Swarm Optimization (PSO) technique to set the ESN with optimal parameters. The input of this model are the measurements of signals correlated to the component degradation state, whereas the model output is the component RUL. To validate the feasibility of the proposed approach, real life fault historical data from turbofan engines system were analyzed and used to obtain the optimal prediction of RUL.

Large combinatorial optimization problems may be overly complex to be processed by a single type of algorithm. This explains the growing interest of researchers in the hybrid resolution. The hybridization of algorithms aims to take advantage of each one benefits, thereby achieving better results.In this paper, a hybrid metaheuristic is proposed to solve one of the most complex variants of the knapsack problem which is the Strongly Correlated Knapsack Problem (SCKP).The proposed approach combines a proposed Ant Colony Optimization algorithm (ACO) with a 2-opt algorithm. The proposed ACO scheme used combines two ant algorithms: the MAX-MIN Ant System and the Ant Colony System.At a first stage, our proposed ACO aims to solve the SCKP to optimality. In case an optimal solution is not found, a proposed 2-opt algorithm is used. Even if the 2-opt heuristic fails to find the optimal solution, it would hopefully improve the solution quality by reducing the gap between the found solution and the optimum.The proposed algorithm was tested on a set of instances and compared with classical and recent methods reported in the literature.

Causal Temporal Signatures (CTS) is an efficient formalism for behaviors description and recognition of fault diagnosis in Discrete Event Systems (DES). The main advantages of this formalism are the readability and the expressivity. Indeed, it is able to describe clearly all desired behaviors and it is understandable and readable by an expert in the field. However, it raises the problem of acquisition and updating of expert knowledge stored in a CTS base. In this paper, we suggest an incremental learning approach based on the simulation to acquire and update automatically a consistent CTS base. The proposed approach is illustrated with an example applied to the turntable helps to understand the different modules of the method.

We study within this paper the effect of agents' management on the global performance of a multiagent system (MAS). Cooperative systems working under coalition formation strategies have been considered in this work. Our evaluation approach uses representation tools of the coalition formation process in addition to mathematical modeling techniques using utility functions. An empirical study has been led to validate this approach and present a benchmarking of three different cooperative strategies: Binary Max-Sum, Distributed Stochastic Algorithm and Greedy algorithm. Each strategy aims at solving the problem of task allocation in search and rescue scenarios.

This paper investigates the effect of agents coalitional strategy on their effectiveness using an evaluation approach that causes no alteration of the global system’s performance. Cooperation process is thus considered during the execution of a Multiagent System (MAS) in order to collect all necessary data for the evaluation phase. A structural analysis is then led to set instant and global assessments of the agents management and output. The observation and evaluation tools are implemented using aspect-oriented programming. We prove through experiments that they do not alter the global performance of the system regarding execution time and reached scores. The approach could then be validated as an objective evaluator of the system by judging real organizational capabilities of its agents in achieving their goals.

Guided genetic algorithm and dynamic distributed double guided genetic algorithm are based on nature laws and by the Neo-Darwinism theory. These evolutionary approaches were very successful addressing Maximal Constraint Satisfaction Problems (Max-CSPs). Our work is inspired by a little mistake when dealing with in these two algorithms guidance. In fact these approaches are guided by the min-conflict heuristic and the template concept. The used template is distorted. So, we introduce a new template concept in order to allow a better guidance. We suggest considering the percentages of violated constraints in place of their number. This concept is, then, applied to guide the genetic algorithms. In this paper, we compare the latter guided genetic algorithm with our new template guided genetic algorithm. The experimentations show that our new template guidance improves the optimization process to find best solutions in better time.

The presence of faults in complex systems can cause serious damage and even generate fatal situations. Proper and timely diagnosis of behavior of such systems is then crucial to detect the presence of eventual faults and isolate their causes. In this context, this paper describes a new intelligent approach to diagnose multiple faults in complex systems. This approach is based on the combination of a fuzzy system optimized by cultural algorithm and causal reasoning. The ongoing experiments focus on a simulation of the three-tank hydraulic system, a benchmark in the diagnosis domain.

In this paper, a method for on-line fault detection and isolation of uncertain parameter systems is proposed. Residual expressions are generated from bond graph model in derivative causality. Detection is based on residual fixed thresholds around a working point and adaptive thresholds that vary according to the set point. Thresholds are generated from the analysis of residuals' sensitivity to different parameters. However, isolation is based on the fault signature matrix and the study of residual tendency. A real simulation example is provided to show the efficiency of the proposed method.