As UAV swarm deployments become more prevalent in mission critical domains, collision avoidance remains a key challenge in ensuring safety, coordination, and autonomy at scale. This survey investigates the state of the art in learning based collision avoidance strategies enabled through collaborative intelligence in UAV swarms. We introduce a six dimensional taxonomy that classifies approaches across decision making paradigms, swarm coordination models, communication architectures, learning methodologies, execution strategies, and safety assurance mechanisms. The survey places particular emphasis on learning based methodologies, which we categorize into four prominent techniques: reinforcement learning, federated learning, neuro inspired models, and hybrid approaches. For each, we provide a detailed review of training architectures, scalability, robustness, and real-time feasibility. Drawing on peer-reviewed publications (2019 till early 2025), we synthesize comparative insights into their application contexts, including trajectory planning, vision-based navigation, decentralized coordination, and multi-agent conflict resolution, while assessing trade-offs in deployment complexity and operational safety. Beyond method specific analysis, the survey highlights key distinctions, practical challenges, and enabling technologies, concluding with open challenges and future directions for scalable and verifiable UAV swarm intelligence.

Time series datasets often exhibit complex and diverse characteristics, including seasonal fluctuations, long-term trends, and irregular variations, which makes it difficult for a single, fixed machine learning model to achieve consistently accurate forecasts across different domains. To address this challenge, we introduce a neuroevolutionary framework for time series forecasting that integrates Evolutionary Algorithms with a Transformer Encoder designed for time series forecasting. This hybrid approach leverages the Transformer’s ability to capture short- and long-term dependencies while using evolutionary search to adjust the model’s hyperparameters to the unique dynamics of each dataset. In this work, we adopt an encoder-only Transformer architecture as a design choice motivated by computational efficiency and the nature of time series forecasting tasks, where the prediction does not require a full encoder–decoder structure. Its optimization is guided by evolutionary algorithms, namely Genetic Algorithm and one of the Estimation of Distribution Algorithms; the Population-Based Incremental Learning. The effectiveness of the approach is validated by benchmarking against ten publicly available univariate time series datasets that cover various patterns and structures. The results demonstrate the notable performance of the proposed model in terms of forecast precision and robustness, highlighting its ability to generalize across various time series scenarios.

In this paper, we propose an approach to solve an “use-inspired” problem: the management of warehouses containing equipment to be installed at various sites throughout the territory by specialized personnel. The challenges to be addressed include the distribution of the warehouse, so as not to slow down work, but also the fact that the tasks to complete change constantly, requiring the opening, closing, and relocation of the corresponding warehouses.

In this article, we investigate a problem of invoice fraud detection in companies. We propose a method that can be applied to invoices received by a company and is used to check if their ID is consistent with the ID of previous documents emitted by the same company. This is one of several tests applied in the fraud detection pipeline and is used to mitigate the company’s daily fraud attempts. The proposed method is based on the automatic discovery of the ID structure, represented by a regular expression, which is used on newly received documents to validate them. The results show that the method can be successfully used to identify fraud invoices with good accuracy

Food security relies on factors like availability, access, and stability, often assisted by food imports when local production falters. Importantly, these imports stabilize supplies, mitigate shortages and price volatility, and enhance economic stability. Anticipating import requirements is vital for proactive food security planning. In this case study, we employ multiple forecasting models to predict food import for a large number of products from multiple countries. The results highlight varying algorithm performance across datasets. Traditional statistical models remain highly competitive compared to newer alternatives, especially for shorter time series. Our study introduces a multi-model forecasting approach to predict periodic food imports, a pivotal tool for food authorities.

A passive in-building distributed antenna system (IB-DAS) is often used to enhance indoor mobile data coverage by introducing indoor antennas inside buildings. Such systems are created to ensure that traffic generated indoors does not heavily depend on base stations installed outdoor, as penetration issues of wireless signals can affect the quality of connection. The focus of this paper is on the automation of IB-DAS design. Particularly, it provides an extensive analysis of the performance of four binary Evolutionary Algorithms (EAs) for this problem and shows that the two tested Estimation of Distribution Algorithms (EDAs) performs well on this problem. Furthermore, it investigates the effect of different genetic operators on the performance of the considered EAs. The practice outcome of this is to select the best algorithm among others to be implemented in a DAS network planning tool and to help our industrial partner reduce both the design time and deployment cost.

This paper proposes an AI solution to optimize the site selection process for 5G Radio Network(RAN) planning. We study various analytic and machine learning techniques to accurately identify the demand for 5G, and use that to plan for optimum 5G site selection, with an aim to have a highest possible return on investment (ROI). The proposed approach first detects the relationship between various network attributes, such as cell performance counters, customer behaviour, handsets’ penetration, and their effect on the expected 5G network load. Then it clusters the cells according to their priorities and required quality of services. It then uses the supervised model to predict and simulate the expected movement of the user to the5G layer, and at the same time, predict the expected change in 4G network performance. Finally, it incorporates the result into a ranking metric with a scoring schema, and provides a list of 5G cell candidates for upgrade considerations. Experimental results are presented to show the validity of the approach.

The traveling thief problem (TTP) has emerged as a realistic multi-component problem that poses a number of challenges to traditional optimizers. In this paper we propose different ways to incorporate dynamic programming (DP) as a local optimization operator of population-based approaches to the biobjective TTP. The DP operators use different characterizations of the TTP instance to search for packing plans that improve the best current solutions. We evaluate the efficiency of the DP-based operators using TTP instances of up to 33810 cities and 338100 items, and compare the results of the DP operators with state-of-the-art algorithms for these instances. Our results show that DP-based approaches, applied individually and in combination with other types of operators, can produce good approximations of the Pareto sets for these problems.

Employing effective optimization strategies in organizations with large workforces can have a clear impact on costs, revenues, and customer satisfaction. This is particularly true for organizations that employ large field workforces, such as utility companies. Ensuring each member of the workforce is fully utilized is a challenging problem as there are many factors that can impact the overall performance of the organization. We have developed a system that optimizes to make sure we have the right engineers, in the right place, at the right time, with the right skills. This system is currently deployed to help solve real-world optimization problems, which means there are many objectives to consider when optimizing, and there is much uncertainty in the environment. The latest version of the system uses a multiobjective genetic algorithm as its core optimization logic, with modifications such as fuzzy dominance rules (FDRs), to help overcome the issues associated with many-objective optimization. The system also utilizes genetically optimized type-2 fuzzy logic systems to better handle the uncertainty in the data and modeling. This paper shows the genetically optimized type-2 fuzzy logic systems producing better results than the crisp value implementations in our application. We also show that we can help address the weaknesses in the standard NSGA-II dominance calculations by using FDRs. The impact of this work can be measured in a number of ways; productivity benefit of £1 million a year, the reduction of over 2500 t of CO2and a possible prevention of over 100 serious injuries and fatalities on the UK's roads.

The service chain planning process is a critical component in the operations of companies in the service industry, such as logistics, telecoms or utilities. This process involves looking ahead over various timescales to ensure that available capacity matches the required demand whilst maximizing revenues and minimizing costs. This problem is particularly complex for companies with large, multi-skilled workforces as matching these resources to the required demand can be done in a vast number of combinations. The vastness of the problem space combined with the criticality to the business is leading to an increasing move towards automation of the process in recent years. In this paper we focus on the tactical plan where planning is occurring daily for the coming weeks, matching the available capacity to demand, using capacity levers to flex capacity to keep backlogs within target levels whilst maintaining target levels for provision of new revenues. First we describe the tactical planning problem before defining a bi-level model to search for optimal solutions to it. We show, by comparing the model results to actual planners on real world examples, that the bi-level model produces good results that replicate the planners' process whilst keeping the backlogs closer to target levels, thus providing a strong case for its use in the automation of the tactical planning process.

With an explosive increase in data traffic over recent years, it has become increasingly difficult to rely on outdoor base stations to support the traffic generated indoors mainly due to the penetration issue of wireless signals. Mobile operators have investigated different options to provide adequate capacity and good in-building coverage such as by deploying femtocells, Wi-Fi off-load or in-building distributed antenna systems (IB-DAS). A passive IB-DAS extends indoor coverage by connecting antennas to a base station through coaxial cables and passive components. This paper focuses on automated design of IB-DAS based on the real world requirements of a telecom service provider. A Genetic Algorithm (GA) is derived for this purpose, giving consideration to different factors, such as minimizing cabling and passive splitter costs, reducing power spillage and power deviation between the required and supplied power for antennas. The solution representation of the problem and the customized genetic operators to assist the evolution are described. The experimental results showing the effectiveness of the GA model on a number of different scenarios are also presented. The built model is incorporated into a software tool, which is being trialled by our industrial partner, delivering encouraging results, saving cost and design time.

Large service organizations such as telecom or utility companies face numerous decision management problems, many of which are related to the management of various resources. The management of spare parts and inventory is one of the key resource management problems an organization faces on a daily basis. Especially their timely availability can have serious impacts on the service quality and the customer satisfaction. Lack of visibility and availability of spare parts in the right place and at the right time can lead to traveling longer distance to supply the parts or in worst case a disruption of the service. We propose to automate the management of spare parts for a legacy technology in a telecom network by leveraging an evolutionary algorithm for optimizing the distribution of spare parts. Our results show that a significant gain can be made in comparison to a assignment typically performed in a manual mechanism.

In real world optimization problems there are often multiple objectives to consider. However, with traditional multi-objective optimization algorithms, like the Non-Dominated Sorting Genetic Algorithm, NSGA-II, one solution is not produced at the end of the process but a set of non-dominated solutions. This set of solutions make up what is known as the Pareto front. The Pareto front relies on calculating the dominance of each solution the multi-objective algorithm produces. Traditional dominance calculations are reasonable for a small number of objectives. However, the more objectives there are in the problem, the more unsuitable these dominance calculations become. This leads to poor selection criteria and ultimately a weaker form of optimization when compared to a small number of objectives. In this paper, we present a fuzzy logic system for computing dominance between two solutions. We have evaluated this fuzzy logic system in optimizing a set of black box test problems. In addition, we have also applied it to a real world many-objective system that optimizes five conflicting objectives, in the telecommunications domain. The implementation of the fuzzy logic system has led to the NSGA-II algorithm with Fuzzy Dominance Rules (FDRs) being able to perform better in a number of black box tests and improving the results of our real-world many-objective optimization problem, with a statistically significant improvement to the hypervolume of 5.46%.

In real world applications it can often be difficult to determine which optimization algorithm to use. This is especially true if the problem has multiple objectives, which is a common occurrence in real world applications. Both Genetic Algorithms (GAs) and Particle Swarm Optimization (PSO) algorithms have been explored, often being compared to each other. As problems are scaled up to more objectives, the suitability of these algorithms can change and would need to be modified. The most common multi-objective algorithms in use are Multi-Objective Genetic Algorithms (MOGA) and Multi-Objective Particle Swarm Optimization (MOPSO), which we are choosing to evaluate, as they can be tested in both their single and multi-objective forms. Real world applications often come with many conditions and constraints. The one being examined in this paper is concerned with the optimal design of working areas, for a large scale mobile workforce in the telecommunications utilities domain. This paper presents the suitable underlying algorithm to use for this problem with the aim of maximizing the utilization of the workforce, whilst having balanced and manageable working areas. The results show that genetic algorithms, in both its single and multi-objective forms, may be the most suitable option for this problem, when compared to PSO and MOPSO algorithms. The results also show that organizing the problem geographically helps the particle swarm algorithms.

In real world optimization problems there are often multiple objectives to consider. However with regular multiobjective genetic algorithms the more objectives there are the more of a problem this becomes for the Pareto front. This is why solutions for Many Objective Problems should be explored. Many objective problems differ from multi-objective problems in that they have more than three objectives [1], [2], [3]. The problem faced by many objective systems is that the more objectives there are the more likely that more solutions will appear on the Pareto front, especially if the objectives are conflicting. This is a problem in two instances, the first is that the genetic algorithm finds it difficult to distinguish between solutions for parent selection, the second is that the output of the system will usually give a big portion of the entire population set. This means that it might be very difficult for users to choose a single solution to apply to the given real-world problem. This paper presents a novel many objective genetic type-2 fuzzy logic based system for mobile field workforce area optimization. This system was employed in real world scheduling problems to handle the high uncertainty levels associated with these domains. We will present a distance measure to avoid the problems associated with the selection of one solution from the Pareto front of Many Objective Problems. The system in this paper uses five objectives from a real world many-objective problem where the objectives are conflicting and as a result the Pareto front becomes saturated with solutions. The distance metric will help to evaluate if optimizing fuzzy systems using a genetic algorithm improves the performance of the system, comparing both optimized and un-optimized type-1 and type-2 fuzzy sets. The results show that optimizing the membership functions of fuzzy sets using a genetic algorithm improved the overall performance of the fuzzy systems and that the distance metric helps to distinguish between the better solutions on the Pareto front. Such optimization improvements of the working areas will result in better utilization of the mobile field workforce in utilities companies.

Critical for successful operations of service industries, such as telecoms, utility companies and logistic companies, is the service chain planning process. This involves optimizing resources against expected demand to maximize the utilization and minimize the wastage, which in turn maximizes revenue whilst minimizing the cost. This is increasingly involving the automation of the planning process. However, due to unforeseen factors, the calculated optimal allocation of resources to complete tasks often does not match up with what is actually occurring on the day. This factor highlights a requirement for a method of predicting accurately the number of tasks that will be completed given a known amount of resources and demand in order to produce a more accurate plan.

In industries which employ large numbers of mobile field engineers (resources), there is a need to optimize the task allocation process. This particularly applies to utility companies such as electricity, gas and water suppliers as well as telecommunications. The process of allocating tasks to engineers involves finding the optimum area for each engineer to operate within where the locations available to the engineers depends on the work area she/he is assigned to. This particular process is termed as work area optimization and it is a sub-domain of workforce optimization. The optimization of resource scheduling, specifically the work area in this instance, in large businesses can have a noticeable impact on business costs, revenues and customer satisfaction. In previous attempts to tackle workforce optimization in real world scenarios, single objective optimization algorithms employing crisp logic were employed. The problem is that there are usually many objectives that need to be satisfied and hence multi-objective based optimization methods will be more suitable. Type-2 fuzzy logic systems could also be employed as they are able to handle the high level of uncertainties associated with the dynamic and changing real world workforce optimization and scheduling problems. This paper presents a novel multi-objective genetic type-2 fuzzy logic based system for mobile field workforce area optimization, which was employed in real world scheduling problems. This system had to overcome challenges, like how working areas were constructed, how teams were generated for each new area and how to realistically evaluate the newly suggested working areas. These problems were overcome by a novel neighborhood based clustering algorithm , sorting team members by skill, location and effect, and by creating an evaluation simulation that could accurately assess working areas by simulating one day's worth of work, for each engineer in the working area, while taking into account uncertainties. The results show strong improvements when the proposed system was applied to the work area optimization problem , compared to the heuristic or type-1 single objective optimization of the work area. Such optimization improvements of the working areas will result in better utilization of the mobile field workforce in utilities and telecommunications companies.

In utility based service industries with a large mobile workforce, there is a need to optimize the process of allocating engineers to tasks (i.e. fixing faults, installing new services, such as internet connections, gas or electricity etc.). Part of the process of optimizing the resource allocation to tasks involves finding the optimum area for an engineer to operate within, which we term as work area optimization. Work area optimization in large businesses can have a noticeable impact on business costs, revenues and customer satisfaction. However when attempting to optimize the workforce in real world scenarios, mostly single objective optimization algorithms are used while employing crisp logic. Nevertheless, there are many objectives that need to be satisfied and hence multi-objective based optimization will be more suitable. Even where multi-objective optimization is employed, the involved systems fail to recognize that these real world problems are full of uncertainties. Type-2 fuzzy logic systems can handle the high level of uncertainties associated with the dynamic and changing environments, such as those presented with real world scheduling problems. This paper presents a novel multi-objective genetic type-2 Fuzzy Logic based System for the optimal allocation of mobile workforces to their working areas. The method has been applied in a real world service industry workforce environment. The results show strong improvements when the proposed multi-objective type-2 fuzzy genetic based optimization system was applied to the work area optimization problem as compared to the heuristic or type-1 single objective optimization of the work area. Such optimization improvements of the working areas will result in improving the utilization of the workforce.

Within service providing industries, one of the challenges facing resource planners is to match the demand for services by trying to utilize the available resources as best as possible. The problem faced by the operational resource planner is to build a refined plan of tasks to resources for each day in a manner that the plan can be directly dispatched to the distributed available engineering field force. In this paper, we will introduce a genetic hierarchical interval type-2 fuzzy logic based operational planner. We will present experiments which will show that the proposed system is able to produce more efficient plans when compared to the traditional crisp logic based algorithms which employ hill climbing heuristic based search techniques. We will show also that the proposed system outperforms the type-1 fuzzy logic based counterparts.

Workforce management is one of the most important factors in the success of any company that provides its customers with services. Hence, in order for the company to achieve objectives like customer satisfaction and maximum resource utilization, there is a need to have a reliable means of efficiently managing the company workforce and making sure that the produced plan always gives a good choice when it comes to assigning the available technicians to the given jobs. As the quantity of services and the workforce grow, the use of an automated workforce management system becomes inevitable. However the automated workforce management system should allow full transparency to allow the user to interact with the generated plans. In addition, the workforce management systems face high levels of uncertainties when dealing with real-world scenarios, which necessitates employing systems, which are able to handle the linguistic and numerical uncertainties available in the real-world scenarios. Fuzzy Logic Systems (FLSs) are credited with providing transparent methodologies that can deal with the imprecision and uncertainties. However the vast majority of the FLSs employ the type-1 FLSs, which cannot directly handle the high levels of uncertainties. Type-2 FLSs which employ type-2 fuzzy sets can handle such high levels of uncertainties to give very good performances. In this paper, we will present a type-2 FLS based workforce management system that is being developed for a delivery unit in British Telecom (BT). We will show how the presented system was able to handle the faced uncertainties to give very good performance that outperformed the automated non-intelligent system and the type-1 FLSs based system.

Structure learning is a crucial component of a multivariate Estimation of Distribution algorithm. It is the part which determines the interactions between variables in the probabilistic model, based on analysis of the fitness function or a population. In this paper we take three different approaches to structure learning in an EDA based on Markov networks and use measures from the information retrieval community (precision, recall and the F-measure) to assess the quality of the structures learned. We then observe the impact that structure has on the fitness modelling and optimisation capabilities of the resulting model, concluding that these results should be relevant to research in both structure learning and fitness modelling.

Distribution Estimation Using Markov network (DEUM) algorithm is a class of estimation of distribution algorithms that uses Markov networks to model and sample the distribution. Several different versions of this algorithm have been proposed and are shown to work well in a number of different optimisation problems. One of the key similarities between all of the DEUM algorithms proposed so far is that they all assume the interaction between variables in the problem to be pre given. In other words, they do not learn the structure of the problem and assume that it is known in advance. Therefore, they may not be classified as full estimation of distribution algorithms. This work presents a fully multivariate DEUM algorithm that can automatically learn the undirected structure of the problem, automatically find the cliques from the structure and automatically estimate a joint probability model of the Markov network. This model is then sampled using Monte Carlo samplers. The proposed DEUM algorithm can be applied to any general optimisation problem even when the structure is not known.

The key ideas behind most of the recently proposed Markov networks based EDAs were to factorise the joint probability distribution in terms of the cliques in the undirected graph. As such, they made use of the global Markov property of the Markov network. Here we presents a Markov Network based EDA that exploits Gibbs sampling to sample from the Local Markov property, the Markovianity, and does not directly model the joint distribution. We call it Markovianity based Optimisation Algorithm. Some initial results on the performance of the proposed algorithm shows that it compares well with other Bayesian network based EDAs.

E-commerce has transformed the way firms develop their pricing strategies, producing shift away from fixed pricing to dynamic pricing. In this paper, we use two different Estimation of distribution algorithms (EDAs), a Genetic Algorithm (GA) and a Simulated Annealing (SA) algorithm for solving two different dynamic pricing models. Promising results were obtained for an EDA confirming its suitability for resource management in the proposed model. Our analysis gives interesting insights into the application of population based optimization techniques for dynamic pricing.

Markov random field (MRF) modelling techniques have been recently proposed as a novel approach to probabilistic modelling for estimation of distribution algorithms (EDAs). An EDA using this technique was called distribution estimation using Markov random fields (DEUM). DEUM was later extended to DEUMd. DEUM and DEUMd use a univariate model of probability distribution, and have been shown to perform better than other univariate EDAs for a range of optimization problems. This paper extends DEUM to use a bivariate model and applies it to the Ising spin glass problems. We propose two variants of DEUM that use different sampling techniques. Our experimental result show a noticeable gain in performance.

This paper presents a methodology for using heuristic search methods to optimise cancer chemotherapy. Specifically, two evolutionary algorithms- Population Based Incremental Learning (PBIL), which is an Estimation of Distribution Algorithm (EDA), and Genetic Algorithms (GAs) have been applied to the problem of finding effective chemotherapeutic treatments. To our knowledge, EDAs have been applied to fewer real world problems compared to GAs, and the aim of the present paper is to expand the application domain of this technique. We compare and analyse the performance of both algorithms and draw a conclusion as to which approach to cancer chemotherapy optimisation is more efficient and helpful in the decision-making activity led by the oncologists. Categories and Subject Descriptors

Markov random field (MRF) modelling techniques have been recently proposed as a novel approach to probabilistic modelling for estimation of distribution algorithms (EDAs) (S. K. Shakya et al., 2004). An EDA using this technique was called distribution estimation using Markov random fields (DEUM). DEUM was later extended to DEUM/sub d/ (S. Shakya et al., 2005). DEUM and DEUM/sub d/ use a univariate model of probability distribution, and have been shown to perform better than other univariate EDAs for a range of optimization problems. This paper extends DEUM/sub d/ to incorporate a simple Metropolis method and empirically shows that for linear univariate problems the proposed univariate MRF models are very effective. In particular, the proposed DEUM/sub d/ algorithm can find the solution in O(n) fitness evaluations. Furthermore, we suggest that the Metropolis method can also be used to extend the DEUM approach to multivariate problems.

This paper presents an empirical cost-benefit analysis of an algorithm called Distribution Estimation Using MRF with direct sampling (DEUMd). DEUMd belongs to the family of Estimation of Distribution Algorithm (EDA). Particularly it is a univariate EDA. DEUMd uses a computationally more expensive model to estimate the probability distribution than other univariate EDAs. We investigate the performance of DEUMd in a range of optimization problem. Our experiments shows a better performance (in terms of the number of fitness evaluation needed by the algorithm to find a solution and the quality of the solution) of DEUMd on most of the problems analysed in this paper in comparison to that of other univariate EDAs. We conclude that use of a Markov Network in a univariate EDA can be of net benefit in defined set of circumstances.