HOA-OBL: hybrid opposition-based hippopotamus optimization framework for efficient UAV task allocation Authors: G Keerthana, R Padmanaban Published: 2026 ABSTRACT AbstractUnmanned aerial vehicles (UAVs) are widely deployed in public and civic areas, especially in environments where human presence is restricted due to safety concerns. UAVs play a vital role in maximizing resource utilization and minimizing operational redundancy in applications such as disaster recovery, infrastructure inspection, and surveillance. However, the conventional task-allocation frameworks often suffer from suboptimal resource utilization and limited adaptability in dynamic operational environments. To resolve these problems, this paper presents a hybrid opposition based Hippopotamus Optimization (HOA-OBL) algorithm framework for efficient UAV task allocation optimization. Particularly, Hippopotamus Optimization Algorithm (HOA) is one of the widely known population-based algorithms, motivated by the inherent behavior of hippopotamuses, but it has a limitation of local optima in certain iterations. To enhance the exploration, the Opposition-Based Learning (OBL) is integrated with HOA to explore the search space to prevent it from converging too early to local optima. We performed simulation experiments in dynamic environments with 20 UAVs and 100 tasks. The experimental results imply that the proposed framework achieves reductions of about 17% to 33% in energy consumption and performance enhancements of about 33% to 50% in load balancing when compared with existing models such as HOA, GA, PSO, and GWO. The results demonstrate that the proposed framework provides a more balanced and scalable approach to task assignment to UAVs, thereby improving efficiency and consistency in real-world UAV missions.Keywords: Unmanned aerial vehicles, Hippopotamus optimization, Task allocation, Opposition-based learning, Load balancingSubject terms: Engineering, Mathematics and computing Introduction Unmanned aerial vehicles (UAVs) are gaining significant focus as a groundbreaking technology for defense and civilian applications due to their flexibility, fast deployment and reduced operational risks. Many UAVs are equipped with cutting-edge sensors, communication modules and autonomous control mechanisms, which enable real-time sensing, coordination and decision making1,2. These technological advancements allow multiple UAVs to operate collaboratively and accomplish complex missions with high efficiency and reliability. As a result, UAVs are widely adopted in diverse applications, such as precision farming, environmental monitoring, infrastructure inspection, emergency response, and surveillance3,4. Recently, next-generation smart aerial systems require multitasking capability to operate autonomously and handle complex tasks.Optimization algorithms are essential in practical UAV deployment supporting efficient scheduling, path planning, energy management and adaptive mission execution in real- world scenarios such as disaster response, surveillance, infrastructure inspection, logistics and agriculture. However, it is challenging for a single UAV to manage tasks simultaneously. Hence, UAV swarm systems have emerged as an effective solution to handle multitasking efficiently by offering several benefits such as improved fault tolerance, faster mission completion, expanded area coverage, and enhanced resilience against individual UAV failures5,6. For instance, in search-and-rescue operations, a UAV swarm can simultaneously explore multiple regions, significantly enriching the efficiency of locating victims. Consequently, in real-time tracking and surveillance scenarios, several UAVs can share the sensing workload with each other to continuously track targets without interruption. Regardless of the benefits of the UAV swarm, there remain some operational challenges, such as task allocation, energy constraints, communication limits, collision avoidance, and safety issues. Of the various challenges to achieve the better coordination of swarms, task allocation is the crucial one. In real multi-UAV missions, the task allocation involves more than just a task-to-agent matching. It has to take into account the heterogeneity of platform capabilities, a lack of on-board energy, the unreliable communication links, the dynamic arrival of tasks, the possible failure of UAVs, the changing priorities of tasks, and environmental constraints such as no-fly zones and complex terrain. The combinatorial complexity of the assignment problem is significantly increased by the interaction of these factors, and this requires optimization strategies that are efficient, adaptive, and feasibility-aware for reliable real-world deployment.The process involves allocating tasks to the right UAVs in a mission which are distributed in space and time sensitive. The task allocation strategy affects directly the mission duration, the energy consumption, and team coordination7–9. When tasks have different priorities and UAVs have different capabilities, efficiently allocating tasks becomes more challenging. In circumstances where there exist hundreds of tasks in a big mission, it quickly becomes hard to find the best allocation. The space for assigning tasks grows exponentially as the number of UAVs and tasks grows. Beyond the computational load, UAV task allocation must include mission constraints such as battery endurance, non-uniform travel times, no-fly zones, sensor range limits, communication delays, and UAV conflicts. Owing to these limitations, optimization-based task allocation approaches have become very popular. Many researchers from various domains utilized optimization algorithms to solve their domain problems. In particular, the ant colony optimization algorithm10, grey wolf optimization11, particle swarm optimization12 and so on have been used to solve the task allocation problem in heterogeneous aerial vehicles.In spite of these optimization approaches demonstrated promising performance, yet many of them rely on static environments and simplified assumptions. Therefore they often suffer from challenges such as premature convergence, scalability issues and lack of adequate feasibility handling in large scale NP hard UAV task assignment problems.The algorithm has problems like premature convergence, poor global exploration ability and lack of dynamic mission inputs despite acceptable results.Especially in UAV task allocation, as swarms grow in size, the discrete nature of the feasibility of assignments, the complex interactions among constraints, and the rapidly expanding search space become more intense. When assigning tasks to multiple UAVs, the search landscapes are commonly high-dimensional, nonlinear and multi-modal. Optimization algorithms need to strike a fine balance between exploration and exploitation if they are to avoid getting stuck in local optima. The necessity for greater diversity, strong global search behavior and faster convergence has led to the development of hybrid and advanced metaheuristic methods.Recent developments in metaheuristics have introduced new frameworks such as Draco Lizard Optimizer (DLO)13, which mimics gliding and adaptive environmental strategies to attain efficient global optimization, and the Cape Lynx Optimizer (CLO)14, which enhances biologically inspired mechanisms for balancing exploration and exploitation. DLO has demonstrated strong benchmark performance and competitive computational efficiency on contemporary test suites, reflecting the rapid evolution of swarm intelligence design.However, despite these advances, such optimizers remain predominantly validated on benchmark and continuous optimization problems, with limited investigation into constrained, discrete multi UAV task allocation that involves dynamic mission uncertainty, feasibility repair, and heterogeneous resource matching. To address this requirement, the present study introduces a hybrid Hippopotamus Optimization Algorithm combined with Opposition-Based Learning (HOA–OBL) for the allocation of tasks within UAV swarms. The Hippopotamus Optimization Algorithm (HOA)15 is a recently introduced bio-inspired metaheuristic that mimics the cooperative and competitive social behaviors of hippopotamuses. During its exploration phase, HOA performs a global search, while in the exploitation phase, it rapidly intensifies the search around promising regions. However, like many swarm-based algorithms, it is also prone to getting stuck in local optima due to the gradual loss of population diversity over time. This restriction limits its ability to explore the best global regions in very complicated task allocation landscapes.In contrast to many recent optimizers developed for continuous benchmarks, HOA provides structured behavioral phases that balance exploration and exploitation. This framework is more appropriate for constrained assignment problems, when combined with discrete repair mechanisms. To enhance the performance of HOA, this study integrates opposition-based learning into HOA. OBL is a well-established technique that improves the optimization efficiency by comparing a candidate solution to its opposite at the same time, thereby increasing the probability of identifying better search regions at an early stage. In addition, OBL increases the probability of initializing the population nearer to the global optimum and enriches the exploration in concurrent iterations.To enhance HOA’s search diversity and reduce stagnation at local optima, Opposition Based Learning is incorporated during both the initialization and iterative search phases. This integration enables broader coverage of the search space and sustains diversification beyond the one time improvement achieved at initialization. By integrating OBL with HOA, enriches the population diversity, accelerates convergence, and minimizes the premature stagnation in local optima. This makes the hybrid system perfect for assigning tasks to multiple UAVs. In addition, it is noteworthy to specify that various researchers have effectively utilized the OBL approach in numerous evolutionary computation frameworks to enrich their exploration phase. However, its integration with HOA remains unexamined in the realm of UAV swarm coordination. Also, many current UAV task allocation frameworks only look at optimizing one objective while overlooking other critical factors that significantly influence mission execution. These limitations motivate the development of a novel hybrid approach that integrates the most effective features of HOA with OBL that enhances search diversity and global exploration capability. The main motive of this hybrid approach is to develop a well-balanced task allocation approach that reduces energy consumption and mission completion time, thereby enriching suitability for real-world UAV swarm operations.Although the optimization algorithms have been developed remarkably, studies on UAV task allocation so far have not provided integrated approaches to address multidimensional capability assessment, dynamic mission uncertainty, constraint feasibility repair, and scalability toward swarm level deployment. To overcome these shortcomings, the HOA-OBL framework is introduced. The main contributions of the proposed work are specified as follows. A robust UAV task allocation framework is proposed based on the HOA-OBL approach, in which the Hippopotamus Optimization Algorithm is augmented with a dual stage Opposition Based Learning mechanism. This enhancement improves population diversity, accelerates convergence, and mitigates premature stagnation in large scale search spaces. A robust UAV task allocation framework is proposed based on the HOA-OBL approach, in which the Hippopotamus Optimization Algorithm is augmented with a dual stage Opposition Based Learning mechanism. This enhancement improves population diversity, accelerates convergence, and mitigates premature stagnation in large scale search spaces. A multi objective, multi criteria evaluation model is formulated that jointly incorporates UAV specific parameters together with task related parameters .Such integration provides a more realistic and balanced UAV-task matching. A multi objective, multi criteria evaluation model is formulated that jointly incorporates UAV specific parameters together with task related parameters .Such integration provides a more realistic and balanced UAV-task matching. A discrete encoding and constraint repair mechanism is proposed to convert the continuous optimisation output into feasible UAV-task assignments. This guarantees feasibility and stability of the solution by ensuring compliance with the assignment uniqueness, energy limits, operational range boundaries, and environmental constraints. A discrete encoding and constraint repair mechanism is proposed to convert the continuous optimisation output into feasible UAV-task assignments. This guarantees feasibility and stability of the solution by ensuring compliance with the assignment uniqueness, energy limits, operational range boundaries, and environmental constraints. A Practical uncertainty aware modeling approach implementation, including local wind estimation, communication limited information exchange and environmental variability.These considerations generalize the applicability of the proposed method beyond idealized assumptions. A Practical uncertainty aware modeling approach implementation, including local wind estimation, communication limited information exchange and environmental variability.These considerations generalize the applicability of the proposed method beyond idealized assumptions. Extensive performance evaluation was conducted against classical and recent optimization algorithms, ablation studies to isolate the impact of OBL and repair strategies, and analyses of convergence behavior, computational runtime and scalability with increasing swarm size. Extensive performance evaluation was conducted against classical and recent optimization algorithms, ablation studies to isolate the impact of OBL and repair strategies, and analyses of convergence behavior, computational runtime and scalability with increasing swarm size. The remainder of this paper is organized as follows. “Related work” provides a review of related studies. “Proposed algorithm” explains the proposed algorithm and its working principles. “Experimental setup and parameter configuration” details the experimental settings, working environment, and performance metrics. “Results” reports and analyzes the results, “Discussions” presents the discussion part, and “Conclusion and future work” concludes the paper and outlines the future directions of the work. Related work Traditional methods In UAV swarm systems, traditional ways of assigning tasks mostly used deterministic mathematical models such as the Assignment Problem (AP), the Traveling Salesman Problem (TSP), and the Multiple Traveling Salesman Problem (MTSP). In these models, tasks are assigned by minimizing total mission cost and travel time. Exact optimization methods like branch-and-bound, Hungarian assignment, and mixed integer linear programming16 give the best answers, but they quickly become too hard to compute for real-world UAV swarm situations as the model is derived from NP-hard problems. Nam and Shell17 proposed exact allocation strategies to reduce resource contention. Centralized UAV task assignment strategies have been applied using deterministic formulations such as the vehicle routing protocol and the traveling salesman problem to order reconnaissance and logistics tasks for small UAV teams; however these methods incur substantial computational delays as the number of UAVs and tasks grows, limiting their real-time applicability18. Moreover, classical centralized solvers usually assume static mission maps and tightly synchronized communication, assumptions that are not valid in real UAV swarms where platform states, battery levels and communication links change asynchronously19. Several traditional optimization studies have focused on UAV/ARS trajectory and communication reliability in complex environments. For instance, outage probability analysis has been conducted for downlink NOMA ARS-assisted cooperative systems with trajectory optimization20 performance enhancement has been explored in NOMA multi-user networks using aerial reconfigurable intelligent surfaces and aerial relay21 and outage performance has been studied for UAV-aided V2V-NOMA communication systems over double Rayleigh channels22. These works highlight the importance of trajectory and communication optimization in UAV networks. However, they remain centered on communication outage and trajectory planning rather than heterogeneous multi-UAV task allocation, which motivates our focus on heuristic and metaheuristic frameworks for adaptive task scheduling. In cell-free multi-ARS communication systems, several traditional optimization studies have addressed advanced resource allocation challenges. Pham Thanh Hiep et.al23 power optimization and ARS selection strategies for downlink multi-ARS systems .Hiep et al24,25 , developed data power optimization algorithms for small-cell environments, and introduced a spatial pilot reassignment algorithm to improve channel estimation in cell-free multi-ARS systems. These works highlight the importance of resource allocation in complex communication infrastructures. However, their focus remains on ARS and cell-free communication optimization rather than heterogeneous UAV task allocation under dynamic mission constraints.Consequently, these limitations have motivated research toward faster, more scalable allocation strategies based on heuristic optimization techniques proficient in managing extensive and intricate mission environments. Heuristic methods Heuristic-based task allocation methods were introduced to provide the solutions to the problems with traditional models that make them hard to scale. These approaches provide fast, approximate solutions derived from decentralized bidding and rule-based decision frameworks. One of the most important among these is the Consensus-Based Bundle Auction (CBBA)7 . In this auction UAVs bid for tasks and share bundle information to ensure global consistency. CBBA is powerful and converges under diminishing marginal gains, but it can be very communicative and does not always give the best global allocations. Zhao et al.9 further improved decentralized coordination even more with the Performance Impact and PI-Softmax algorithm. These algorithms use exploration-driven distributed decision-making and work better in time-critical rescue situations. Other heuristic scheduling methods, like willingness-based task allocation5 along with distributed time-constrained rescheduling2, show that they can work well in practice, but they are limited since they rely on local information, can’t handle multi-objective optimization, and don’t work as well in environments with a lot of tasks. Recent heuristic and hybrid optimization studies have successfully extended UAV system design toward deployment-oriented applications. For instance, Nguyen et al.26 introduced a hybrid deep learning–heuristic framework that integrates adaptive multi-start PSO with a hybrid U-Net to optimize 3D UAV placement in multi-user systems, achieving near-optimal throughput with significantly reduced inference time. Similarly, PSO-based joint power and position optimization strategies27 have been shown to enhance fairness and throughput in multi-UAV small-cell environments. These approaches show the practical value of swarm-inspired heuristics for UAV deployment and communication optimization. However, their scope is still largely limited in the area of deployment geometry and network level performance, leaving the problem of heterogeneous multi-UAV task allocation in the presence of dynamic mission constraints under addressed.Thus, heuristic methods are fast and reliable, but they are not able to perform the global search needed for complex UAV swarm coordination. Metaheuristic methods Metaheuristic optimization techniques have become very popular for allocating tasks to UAVs because they can quickly search large solution spaces and deal with multiple objectives, such as the total cost of the mission, the maximum workload for the UAV, and the efficiency of the path. Genetic algorithms28,29 and particle swarm optimization-based methods30,31 have been extensively utilized for multi-UAV routing and workload distribution; however, they frequently experience premature convergence and require meticulous parameter tuning. Ant Colony Optimization variants, such as memetic ACO32 and the Multi-Objective ACO System (MOACS)33, exhibit robust routing capabilities and deliver enhanced Pareto fronts compared to NSGA-II and MOPSO in various benchmark instances. Wolf Pack Search (WPS)34 and hybrid PSO–GA models are two more bio-inspired methods that improve global search ability, but they still lose diversity and get stuck in later iterations. Geng et al.35 proposed a task allocation method in UAV clusters using a quantized particle swarm optimization (QPSO) algorithm. However, complexity can make it more challenging to implement and tune correctly for specific problems, particularly for complex multivariable optimization tasks. Duan et al.36 presents a predator–prey particle swarm optimization (PP-PSO) method that is integrated with dynamic game theory for task assignment in multi-UAV combat. The PP-PSO is used to compute the mixed Nash equilibrium at every decision step, generating a flexible plan for engaged parties in combat. Although PP-PSO improved convergence, resolving Nash equilibrium remains computationally complex for a large scale. Lizhi Chen et al.10 proposed a multi-objective ACO (MOACO) algorithm to solve the complex task of assigning cooperative tasks to different types of UAVs. The problem is formulated as a constrained multi-objective optimization problem with three main goals: to minimize task benefit, UAV damage, and resource use. However, MOACO problem is that it tends to converge too quickly on suboptimal solutions. Gao et al.37 also suggested a grouping ACO algorithm to solve the hard combinatorial optimization problem. They also came up with a categorization model for different types of targets to improve representational fidelity. Wu et al.38 introduced a new task allocation approach combining a double-chromosome GA with simulated annealing to reduce computation time and improve efficiency; however, it struggles in handling dynamic situations during task execution. Wang et al.39 proposed a multi-UAV task allocation problem where Improved Grey Wolf Optimization (IGWO) can solve the local optima problem well and suit large-scale complex environments. However, the algorithm suffers from increased computational complexity compared to other algorithms. Wei et al.40 employed the firefly optimization algorithm with multi-neighbor search and simulated annealing to escape local optima, but the method remains sensitive to premature convergence and lacks responsiveness in highly dynamic environments. OBL enhanced optimization methods Several studies have integrated OBL with PSO, GA, GWO, Firefly, harmony search algorithm, DE, and other algorithms, demonstrating the performance by improving the convergence rate, global search capability, and solution quality. For instance, Opposition-Based Learning was integrated with Particle Swarm Optimization to create the Opposition-Based Particle Swarm Optimization (OPSO) algorithm for cloud task scheduling, which improved convergence speed, load balancing, and energy efficiency compared to standard PSO, mPSO, GA, and heuristic schedulers41. The genetic algorithm was hybridized with OBL to form an opposition-based genetic algorithm with double chromosome encoding, which improves the population variety for enhancing the global exploration capability42. The firefly algorithm integrated with OBL was also employed to accelerate the convergence and improve the FA’s exploration capability and also demonstrated superior performance in small- to medium-scale dimensions, while maintaining comparatively moderate efficacy in higher-dimensional scenarios43. Similarly, the Opposition-Based Learning Parameter-Adjusting Harmony Search (OLPDHS) algorithm44was proposed to address multi-UAV cooperative task allocation in complex battlefield environments. The authors have developed a heterogeneous task assignment framework that classifies missions into surveillance, attacks, and evaluation categories, integrating considerations such as mission utility, execution cost, payload limitations, attack range, and terrain constraints. At the same time, an Improved Adaptive Differential Evolution (IADE) algorithm that used Opposition-Based Learning was introduced45 to make global search more efficient and speed up convergence in difficult optimization situations.The framework combines OBL-based population initialization with adaptive modulation of control parameters to keep the population diversity and avoid premature convergence. Moreover, Ntakolia et al.46 proposed a new hybrid algorithm called Fick’s Law Algorithm with Opposition-Based Learning (FLA-OBL). The algorithm balances exploration and exploitation in searching for better solutions. These OBL-enhanced methods demonstrate the utility of adding opposite solutions at the start and at each iteration. Extract from literature A comprehensive list of all task allocation and optimization methods discussed in the literature is provided in Table 1. A comparison illustrates that traditional heuristic and auction based methods are easy to implement and converge quickly but do not always obtain the best global solution and are not very adaptable to changing environments. Nevertheless, metaheuristic and multi-objective optimization techniques are efficient only if parameters are tuned appropriately, and they often fail in large-scale and time-varying mission scenarios, despite the fact that they provide better solution quality and adaptability. Furthermore, the existing approaches do not adequately consider the real-time operational constraints related to energy availability, task urgency, environmental disturbances, and the heterogeneous capabilities of UAVs in a comprehensive optimization framework. Such limitations stress the need for a robust and flexible task allocation model, which can effectively balance exploration and exploitation, preserve solution diversity, and guarantee constraint-aware optimization for dynamic UAV-assisted missions. The proposed model is aimed to overcome these problems by combining the adaptive search algorithms with the oppositional based learning. This will improve convergence behavior and solution quality in realistic mission conditions. Despite its proven effectiveness in different application domains, opposition based learning has not been systematically integrated into the HOA for the UAV task allocation. This study fills this gap by adding OBL to the HOA framework to make solutions more diverse, speed up convergence and create stronger task-UAV mappings. While OBL and HOA have significant potentials, there is no research on combining OBL and HOA for UAV task allocation, which is a specific research gap that this paper aims to fill. Table 1Comparison of task allocation and optimization approaches.Author [ref]MethodologyBenefitsPerformance metricsAccuracy / improvementLimitationsTurner et al. (2018)2Distributed task reschedulingImproves online rescheduling; robust to time constraintsCompletion time, feasibility20% increase in task allocationsCommunication overhead and limited scalabilityZhong et al. (2021)5Willingness-based task allocationPriority-aware assignment; lightweight computationTask completion, agent utility75% higher total gain than IGPADepends on willingness model; heuristic-basedChoi et al. (2009)7CBBA decentralized auctionScalable with provable convergenceConvergence rounds, optimality boundApprox. 50% optimality under diminishing marginal gainsCommunication load; sometimes globally suboptimalChoi et al. (2010)8Decentralized auction with cooperation constraintsSupports heterogeneity and distributed decision makingTask assignment quality17% performance improvementRequires reliable communication; no global optimalityZhao et al. (2016)9PI-based distributed heuristicBetter workload balancing; suitable for time-critical tasksMission completion timeOutperforms CBBA variants in rescue scenariosScenario-specific; depends on local communicationWhitbrook et al. (2018)47Reliable distributed schedulingRobust to dynamic task arrivalsTimeliness, reliability9% reduction in average completion timeComplexity increases with swarm sizeLi et al. (2020)48ACO-based UAV mission assignmentEffective routing; aligns well with MTSPPath length, mission time40–60% improvement over GA-BPNN and BPNNParameter sensitivity; pheromone stagnation riskNam and Shell (2015)17Exact assignment algorithmOptimal solutions for small-scale problemsOptimality vs. runtimeOptimal in small casesNot scalable to large problemsQibo et al. (2013)28Modified GA for heterogeneous UAVsHandles multi-capability constraintsFeasibility, mission costProvides good feasible and optimal solutionsPremature convergence; tuning requiredDeb et al. (2002)29NSGA-II multi-objective GAStrong Pareto diversity with elitismIGD, HVExcellent benchmark performanceSlow convergence for large search spacesWei et al. (2020)30Multi-objective PSO (MOPSO)Fast convergence; simple formulationIGD, Pareto quality26% reduction in total costPremature convergence; parameter sensitivityChen et al. (2018)34Wolf Pack Search (WPS)Handles uncertainty effectivelyMission success, completion timeBest objective value of 4650.5High tuning complexityDorigo and Gambardella (1997)32Ant Colony System (ACS)Adaptive pheromone-based routingPath length, convergence speedStrong TSP performancePheromone stagnation without controlKennedy and Eberhart (1995)31Particle Swarm Optimization (PSO)Simple structure; fast convergenceObjective value, convergence speedNear-optimal benchmark performanceProne to local minima and premature convergenceChen et al. (2018)49Memetic ACO for bi-objective MTSPLocal search enhances refinementIGD, HV, path qualityImproved Pareto frontsHigh computational cost; tuning requiredShi et al. (2010)50Multi-objective optimizationEffectively manages conflicting objectivesPareto metrics, mission timeFeasible multi-objective solutionsLimited validation on large-scale systemsKe et al. (2013)51MOEA/D–ACO hybridStrong decomposition with ACO searchIGD, HVHigh optimization accuracyComplex hybrid structure; tuning overheadShuai et al. (2019)52NSGA-II-based MTSP variantEfficient multi-route optimizationMission cost, convergenceImproved MTSP solution qualityEncoding and parameter sensitivity Comparison of task allocation and optimization approaches. Proposed algorithm The proposed model for allocating tasks to UAVs relies on the necessity for missions that consider capabilities, uses energy efficiently, and adapt to the situation. Five important key factors are considered for the UAV’s weight: sensor capability, transmission quality, wind impact factor, remaining battery life, and available payload capacity. The UAV’s ability to do surveillance, monitoring, and inspect tasks depends on performance of the sensor being able to gather important data. The quality of transmission shows the efficency of the UAV to keep a stable connection with ground control. This is important for stable data transfer while it is in use. The wind-impact factor tells the reliability of the UAV, the capability of the system to maintain flight stability under various environment conditions, and the amount of energy it uses. The UAV’s ability to complete its mission without running out of power is also affected by the left over battery power. The remaining payload capacity is also important for keeping UAVs from being overloaded, which let them carry equipment that are made for certain tasks. Five metrics are considered to figure out the task weight: task completion time, distance to task from UAVs, estimated energy consumption, task priority score and resources required. The time it takes to do a task and the distance between the UAV and the task both show how much time and space it takes to finish the task. This lets the allocation model give tasks to UAVs that can get to them and finish them with the least amount of delay. The estimated energy use makes sure that tasks are only given to UAVs that have enough energy left, which reduces the chance of failures during the mission. Task urgency, which is based on deadlines and mission-critical needs, helps determine which tasks are more important for operations. Finally, the required resources metric takes into account the specific needs of a task, including its payload, to make sure that the capabilities are right for the task. When examined together, these UAV and task metrics give a overall depiction of both the mission’s limits and the UAV’s capabilities. This helps the Hippopotamus Optimization Algorithm work better by adding Opposition-Based Learning , which makes it possible to do UAV-task allocations that are energy-efficient yet adaptable to the situation. UAV metrics Table 3 presents the nomenclature of UAV metrics. Each UAV is characterized by a composite weightage score that consolidates its intrinsic and environmental capabilities. The following parameters are considered: Table 3Nomenclature of UAV metrics.SymbolDefinitionSensor capability score of transmitter quality score of Wind Impact score of Residual battery energy score of Remaining payload / resource-balance capability of Number of sensor types equipped in the UAV fleetRelative importance (weighting coefficient) of sensor type kAvailability score of sensor k on Effective communication quality index (range/RSSI/throughput) of Maximum achievable communication index across the UAV fleetCurrent environmental wind intensityMaximum safe wind operating tolerance of Remaining energy reserve of Full nominal energy capacity of UAVCurrently occupied payload of Maximum payload carrying capacity of Remaining payload left: Fitness weighting coefficients for optimizationComposite UAV weightage score Nomenclature of UAV metrics. Sensor capability score The sensor capability score of ,denoted as is calculated as the weighted sum of onboard sensor performance. Different sensor types, such as cameras, lidar’s, thermal, and GPS, provide complementary information. where represents the number of sensor types equipped in the UAV fleet, is the standard measuring quality index of sensor type k on , and is the relative importance coefficient representing the mission sensitivity toward sensor type k. These coefficients show how important each sensor is for surveillance missions. This makes sure that perceptually dominant sensors have a bigger impact on the score. Transmitter quality The transmitter quality is denoted by . The transmitter quality indicates its UAV communication efficiency. This efficiency is vital in maintaining communication links between the UAV and the ground control stations and other UAVs in a swarm network. Efficient communication ensures the proper functioning of task updates, the transmission of sensor information, and coordination in real-time. where denotes the effective communication quality index of , that is based on telemetry link measurements like RF communication range, packet success ratio, and achievable throughput. is the highest communication quality index seen among the UAV fleet. This ensures that tasks are ranked based on how well they communicate, which stops highly communicative and time-sensitive tasks from being given to UAVs with poor wireless capabilities. This form of quantification ensures is confined to the [0, 1] range, where values close to 1 denote better performance. Wind impact factor with limited visibility The proposed framework is not fixed at global scale and perfectly known wind speed in order to realistically model the effects of wind. Instead, each UAV makes a local guess about the wind with some uncertainty and, if communication allows, uses information from nearby UAVs to improve this guess. Local wind estimation The instantaneous wind speed affecting at time step (t) is estimated as: where denotes the true local wind speed affecting UAV i, is the corresponding estimated wind speed, and represents zero-mean Gaussian uncertainty caused by sensor noise, turbulence, or environmental estimation error. If , indicating no successful neighbor communication, the fused estimate defaults to self-estimation, i.e., = Communication-limited neighbor fusion To avoid requiring global communication facilities, UAVs only share wind estimates with people in the same neighborhood. The uses a combined wind estimate that is defined as: where is the improved wind estimate, is the group of nearby UAVs that can talk to at time (t), and controls how strong the fusion is between the UAV’s self-estimate and the information from its neighbors. The binary variable shows if received the wind message from UAV (k) at time (t). This means that it also shows packet drops and intermittent connectivity. Wind impact factor The wind impact factor is calculated as a limited stability score based on the fused wind estimate: where, is the highest wind speed that can handle. The term stands for the expected wind robustness of : values close to 1, meaning that the UAV works well in low wind, while values close to 0 mean that the UAV is at a high risk of damage and performance loss in high wind. Wind-adjusted effective capability To incorporate environmental realism into task allocation, the nominal UAV capability score is adjusted by a wind impact factor: Here denotes the nominal multidimensional UAV capability score, derived from payload capacity, sensing capability, and residual energy metrics. The effective score represents the task suitability of UAV under prevailing wind conditions, ensuring that environmental disturbances directly influence allocation decisions without assuming globally available wind information. Battery level denotes the battery power left. It shows how much energy a UAV is available, which in turn affects the endurance of the UAV, the distance covered by UAV,and the success rate of the task without stopping. When planning UAV missions, battery limits are a big problem especially those that need to be done quickly and last a long time. shows how much energy is left in , which is based on battery telemetry from the onboard system. shows the nominal full-charge energy capacity of the battery pack that comes with it. This normalized formulation turns different battery states into a single scale from 0 to 1. Values closer to 1 mean that the UAVs have more battery life left. The UAV weightage calculation makes sure that the allocation process gives drones with enough energy reserves the most weight. Payload capacity left The payload capacity left score of is denoted by . The payload describes about the weights that UAV carry the equipments and sensors. This is more significant during missions involving the UAV in sensor deployment, package shuttling, and taking data collection instruments for UAV-assisted tasks. where, is the maximum payload carrying capacity of , and = denotes the remaining payload after considering installed sensors. This standardization guarantees that it is constrained under the condition that lies between 0 and 1. where, 0 represents the UAV having no remaining payload and 1 represents the UAV having full payload availability. To maintain dimensional consistency across heterogeneous UAV and task attributes, all metrics are normalized to the range [0,1] prior to weighted aggregation. Task allocation algorithms consider the payload availability, ensuring that the UAV assigned to a given mission can carry all the required materials without placing the UAV under a minimum operational load. This avoids the adverse effects of an underweight condition, which may result in increased energy expenditure, more safety hazards, and stable flight. Weight coefficient determination using analytic hierarchy process (AHP) To minimize subjective bias in assigning weighting coefficients for multidimensional UAV evaluation, the Analytic Hierarchy Process was applied to systematically derive the relative importance of each criterion. Through structured pairwise comparisons, AHP establishes weighting coefficients based on the operational significance of each factor within the UAV task allocation process. For UAV capability assessment, five criteria were considered: sensor capability , transmission quality , wind resilience , residual battery energy , and remaining payload capacity . The pairwise comparison matrices were built on the Saaty 1-9 scale with relative importance values assigned based on mission-critical priorities.The normalized principal eigenvector of each matrix was used to determine the weighting coefficients for tasks and UAV metrics. To check the consistency of these judgments, the consistency index (CI) and consistency ratio (CR) are used which is defined as: Where is the maximum eigenvalue of the comparison matrix, n is the number of criteria and RI is the random consistency index. A consistency ratio of CR < 0.1 is generally accepted as an acceptable level of logical consistency.The consistency of these judgments was verified using the consistency index (CI) and consistency ratio (CR) is defined as: The resultant AHP based UAV weighting vector was obtained as: . The AHP-based capability evaluation weights of UAVs are listed in Table 2, which shows the relative importance of sensor capability, transmission quality, wind robustness, residual battery energy, and payload capacity in the multidimensional UAV assessment framework.These coefficients are applied in Eq. (11) to compute the normalized overall UAV suitability score. The cumulative UAV weightage score is computed through consolidating various performance metrics into a unified measure, enabling the optimal UAV to be chosen for task allocation. Where is the overall weightage score of . is the sensor capability score of , is the transmission quality score, is the wind impact factor, is the residual battery energy score, is the remaining payload score. are non-negative weighting coefficients that control the relative importance of each normalized UAV factor, subject to: . Table 2AHP-derived weighting coefficients for UAV.CriterionWeightSensor capability0.22Transmission quality0.18Wind robustness0.16Residual battery0.24Payload capacity0.20 AHP-derived weighting coefficients for UAV. Task metrics The nomenclature of task metrics are presented in Table 5. The list of task metrics is presented below. Table 5Nomenclature of task metrics.SymbolDefinitionEstimated total completion time for to finish Euclidean distance from to Estimated energy consumption of when executing Normalized urgency/priority score of Required resourcesNormalized resource (payload/sensor) demand required by Cruise velocity of On-site execution time required by Energy/power consumption rate of Resource/payload requirement for Maximum resource/payload capacity available in UAV fleetComposite weightage scoreFitness weighting coefficients Nomenclature of task metrics. Task Completion time Task completion time is denoted by . It indicates the total time for to do , which comprises travel and execution portions. In time-critical operations, this metric performance turns to impact mission efficiency and responsiveness in the operations, highlighting its importance in mission scheduling and task allocation. where is the distance from to the location of . is ’s cruising speed, representing its average operational travel velocity. is the estimated time to perform the task on-site. The first term and is the travel time which is a function of the UAV’s speed and the distance to be covered. The second term pertains to operation time relevant to the task. This helps in ensuring the UAVs are assigned to tasks that can be done within the deadlines, minimizing the idle time and the travel overhead. Distance to task The parameter , measures the Euclidean distance from the UAV to the location of the task. Distance in UAV task assignment is an important factor because it affects travel time, energy use, and the overall success of the mission. where, and are the coordinates of ’s position in a 2D operational space, and are the coordinates of task location j. In multi-criteria decision-making situations with different UAVs and tasks, distance may need to be normalized to make sure that everyone is treated fairly. Using normalization, the data changes to a desired range, usually [0,1].In this case, and are the shortest and longest distances between the UAVs and the tasks they are assigned to. Adding and its normalized form to the task allocation algorithm lets UAVs do tasks that are closest to them, which reduces the travel time and energy use. Estimated energy consumption The estimated energy consumption for a given performing is modeled as: where , represents the UAV’s power consumption rate, measured in watts (W), and denotes the total time required to complete the task. The total task time comprises both the travel time from the UAV’s current location to the task location and the execution time required to complete the task. Task urgency/priority score The urgency of is quantified using the following formulation: where, = denotes the remaining allowable time before the deadline of , and is the maximum remaining time-gap span among all tasks in the task list. A task with is considered highly urgent and need immediate attention, whereas corresponds to tasks with ample remaining time. Integrating this system with task allocation algorithms ensures urgent tasks are fairly prioritized regardless of system workload. This is evident in UAVs with deadlines, where tasks are assigned based on incoming deadlines. Required resources The resource requirement is expressed as: where, is the amount of surveillance resources needed,or , and is the highest amount of payload that the UAV fleet can handle. This term makes sure that UAVs aren’t given jobs that are too hard for them to handle and helps with capability-matched allocation during optimization. Task weight coefficient determination using AHP The same AHP methodology was applied to derive weighting coefficients for task demand evaluation. Five task-related criteria were considered: task completion time , travel distance , energy consumption , task priority , and payload requirement .Based on pairwise comparison analysis, the final task weighting vector was determined as: . Table 4 reports the AHP-derived weighting coefficients for task demand evaluation, reflecting the relative significance of task completion time, travel distance, energy consumption, task priority, and payload requirement within the generalized task assessment framework. These weights are then used in Eq. (18) to calculate the normalized overall task demand score. Table 4AHP-derived weighting coefficients for task.CriterionWeightCompletion time0.22Travel distance0.18Energy demand0.20Task priority0.24Payload requirement0.16 AHP-derived weighting coefficients for task. The task weightage is computed as: where is the estimated completion time to complete the , is the distance between and , is the energy consumption of to execute , is the priority score of , is the required payload for and are non-negative weighting coefficients that represent the relative significance of each normalized task factor, satisfying: . Problem formulation The multi-UAV surveillance environment was examined that has a group of M UAVs and a set of N surveillance tasks that are carried out in a limited area of operation. denote the set of UAVs and denotes the set of tasks. The goal of the task allocation process is to give UAVs tasks in a way that keeps the overall mission cost as low as possible while still meeting operational feasibility constraints like battery life, payload limits, communication reliability, and sensing reachability. Decision variable A binary assignment variable has been defined as follows to model the allocation: This binary representation allows the UAV–task pairing problem to be framed as a constrained combinatorial optimization problem, assigning each task to precisely one UAV. Objective function Let be the cost of giving to . The proposed framework allows us to find from the calculated task weightage and UAV weightage. The global mission goal is stated as: Constraints The following limits are put in place to make sure that allocations are valid and possible. First, there must be one UAV for each task: The assignment variables still have two values: Limiting the energy use of when doing to the remaining available energy reserve makes sure that the battery works: where is the estimated amount of energy that will use to do , and is the amount of battery energy that has left. In the same way, payload feasibility is enforced by making sure that the resource demand of doesn’t go over the remaining payload capacity of : where is the amount of resources that needs, and is the extra payload space that has.Also, enforcing a minimum link quality requirement can help keep communication reliable: where is the communication quality index for and is the lowest level of communication that is acceptable. Finally, if you think about the limits of the sensing range, the assignments must meet the sensing feasibility condition: where, Δ ij is the distance from the UAV to the task and is the range of the sensor’s detection. To include UAV capability-awareness in the allocation, the assignment cost can be changed even more by using the UAV suitability score in the following way: where, is the task-side composite weightage of giving to , is the UAV capability score, and \varepsilon >0 is a small constant that keeps the numbers stable. This goal encourages giving more capable UAVs expensive while lowering the overall cost of the mission. The above formulation describes a constrained discrete optimization problem with binary decision variables and several coupled feasibility constraints. So, the proposed work uses the Hippopotamus Optimization Algorithm with Opposition-Based Learning to quickly find the best possible allocation that meets all mission requirements and minimizes the objective function J. Solution representation and encoding framework for UAV task allocation Given that the initial Hippopotamus Optimization Algorithm works in a continuous search space, a framework is needed to decompose it into smaller parts in order to solve the NP-hard and combinatorial UAV task allocation problem. In this study, each candidate solution is represented as a real-valued vector , where N denotes the number of tasks, and each dimension reflects a UAV-selection preference for a specific task. A valid assignment vector is created by applying a deterministic integer mapping to each HOA position update. Specifically, rounding and boundary clamping are used to make each task variable discrete ensuring that each task is given to one UAV index that is within the range of what is possible. A constraint-repair mechanism is also used to move any task-UAV pair that can’t be done to the nearest feasible alternative, making sure that the payload and battery limits are met. This encoding successfully shifts HOA from ongoing optimization to distinct UAV–task allocation. For binary search spaces, sigmoid or V-shaped transfer functions are usually used. Due to the fact that the task allocation involves multi-class integer assignments , an integer discretization scheme that is better suited for and more commonly used for assigning multiple UAVs. Constraint repair mechanism Although,the encoding function in Eq. (26) guarantees that each task is mapped to a valid UAV index. The HOA essentially works in a continuous search space while the UAV task allocation is a discrete constrained combinatorial problem. Hence, candidate solutions produced during the optimization process can still violate assignment feasibility conditions that relate to sensing compatibility, residual battery energy, payload capacity, operational range, or workload balancing constraints. Thus, a dedicated constraint repair mechanism is employed after the solution decoding and before the fitness evaluation to enforce the solution feasibility.A constraint repair mechanism is used after the solution decoding and before the fitness evaluation to ensure a valid task allocation.The repair process requires the following feasibility conditions : Task uniqueness: a task should be assigned to exactly one UAV. When more than one UAV is assigned to the same task, the one with the highest suitability score is retained and the conflicting assignments are re-assigned to the next best feasible UAVs. Task completion: every unassigned task is assigned to the nearest feasible UAV that satisfies the capability, payload and energy requirements. Energy feasibility: the lower priority tasks are assigned iteratively until the energy constraint is met when the total assigned task energy of a UAV exceeds its residual battery capacity. Operational feasibility: if a mission violates payload, communication, or mission-range limitations, it is replaced with the next highest feasible candidate that is compatible with the UAV-task. This repair strategy guarantees that all candidate solutions meet mission-critical constraints and preserves as much optimization structure as possible. The proposed approach improves the feasible solution ratio, improves the convergence stability and reduces the computational waste by correcting the infeasible assignments rather than discarding the whole candidate solutions. Constraint repair procedure for feasible UAV task allocation. Fitness function Maximum possible score The maximum possible score is the best performance that the UAV-task allocation system could theoretically achieve. For each , the one with the highest capability coefficient is found and matched with the task importance coefficient that goes with it. The total of these best-case pairings for all tasks is an optimality benchmark, which is defined as: This formulation makes sure that the optimization process is guided by a clear performance boundary, which is then used to scale and normalize the fitness values of the candidate solutions. Normalized fitness For a given allocation S, fitness is evaluated as given in where, is the result for ’s weightage, and is the outcome for the weightage. To find the best way to divide the work, the UAV with the most ability, called max , is picked for each .Subsequently, the importance score of task j is used to weight the score of the chosen UAV. The impacts from all tasks are added together to get the theoretical highest score that could be achieved if the UAV best suited for each task did it. Where represents the capability score given to , and stands for the importance score assigned to task i. To determine an optimal allocation, for every task j, the UAV that has the greatest capability, denoted as max , is chosen. The chosen UAV’s score is then weighted by the importance score of task j. The resultant contributions across all tasks are aggregated, providing the theoretical peak score that could be secured under the assumption that each task is carried out by the UAV best suited for its execution. Integration of opposition-based learning To enhance HOA performance, OBL is incorporated during both initialization and iterative updates. For a candidate solution μ (i,j) in dimension j, its opposite is computed as where, and stands for the dimensions of lower and upper bounds, respectively. This formulation guarantees that the opposing solution is symmetrically located within the search space’s feasible range. Both the original and opposing populations are created and assessed during initialization, and the best solution is then kept. Furthermore, OBL is applied at regular intervals during the iterative process utilizing a predetermined jumping rate , which probabilistically ascertains if the opposite of a solution dimension is taken into account. OBL increases population variety and speeds up convergence toward the ideal UAV–task allocation by assessing both the existing and opposing solutions and keeping only the best candidates. Hippopotamus optimization algorithm HOA mimics the foraging and survival strategies of hippopotamuses, applied to UAV allocation as follows: Exploration: Hippos look for new water sources at random throughout this stage. To vary the search space in the UAV domain, this is similar to randomly reassigning specific duties to several UAVs. By preventing premature convergence, this method makes it possible to find potentially better task–UAV combinations. Exploration: Hippos look for new water sources at random throughout this stage. To vary the search space in the UAV domain, this is similar to randomly reassigning specific duties to several UAVs. By preventing premature convergence, this method makes it possible to find potentially better task–UAV combinations. Exploitation: A hippopotamus approaches a prospective water source once it has been located. This is equivalent to shifting current allocations in UAV task allocation in favor of the dominant solution, also known as the dominant hippo. The search agents improve their allocations and raise the swarm’s overall fitness by partially following the task assignments of the optimal solution. Exploitation: A hippopotamus approaches a prospective water source once it has been located. This is equivalent to shifting current allocations in UAV task allocation in favor of the dominant solution, also known as the dominant hippo. The search agents improve their allocations and raise the swarm’s overall fitness by partially following the task assignments of the optimal solution. Proposed HOA-OBL approach The overall working process of the proposed HOA-OBL architecture is presented in Fig. 1. The overall working process of the proposed HOA-OBL algorithm is illustrated in Algorithm 2. In addition, the simplified representation of the proposed algorithm is described below. Proposed HOA-OBL architecture diagram. Normalize UAV and task-related metrics to obtain dimensionless weighted coefficients. Normalize UAV and task-related metrics to obtain dimensionless weighted coefficients. Generate an initial population of UAV-task assignment vectors and compute their corresponding opposite solutions using oppositional-based learning. Generate an initial population of UAV-task assignment vectors and compute their corresponding opposite solutions using oppositional-based learning. Evaluate the fitness F(S) of the combined population and select the top solutions to form the initial Hippopotamus Optimization Algorithm population. Evaluate the fitness F(S) of the combined population and select the top solutions to form the initial Hippopotamus Optimization Algorithm population. Execute the iterative optimization process: Update solution positions using the HOA exploration and exploitation mechanisms.Periodically apply OBL to selected individuals to enhance population diversity.Repair infeasible UAV-task assignments to satisfy operational and resource constraints.Re-evaluate the fitness of the updated solutions.Update the global best solution and its fitness if an improvement is observed. Execute the iterative optimization process: Update solution positions using the HOA exploration and exploitation mechanisms.Periodically apply OBL to selected individuals to enhance population diversity.Repair infeasible UAV-task assignments to satisfy operational and resource constraints.Re-evaluate the fitness of the updated solutions.Update the global best solution and its fitness if an improvement is observed. Update solution positions using the HOA exploration and exploitation mechanisms. Update solution positions using the HOA exploration and exploitation mechanisms. Periodically apply OBL to selected individuals to enhance population diversity. Periodically apply OBL to selected individuals to enhance population diversity. Repair infeasible UAV-task assignments to satisfy operational and resource constraints. Repair infeasible UAV-task assignments to satisfy operational and resource constraints. Re-evaluate the fitness of the updated solutions. Re-evaluate the fitness of the updated solutions. Update the global best solution and its fitness if an improvement is observed. Update the global best solution and its fitness if an improvement is observed. Terminate the algorithm upon reaching the maximum number of iterations. Terminate the algorithm upon reaching the maximum number of iterations. Output the optimal UAV-task allocation solution along with the corresponding performance metrics. Output the optimal UAV-task allocation solution along with the corresponding performance metrics. Hybrid oppositional-based hippopotamus optimization algorithm. Computational complexity analysis To evaluate the additional effort that Opposition-Based Learning adds to the computation, the time complexity of the baseline HOA and the new HOA-OBL were compared. Let N stand for the number of candidate solutions, and T stand for the most iterations. Each possible solution must be decoded into a possible task assignment and checked against the objectives and constraints. is used to show the amount of combined decoding and evaluation process costs. Initialization of HOA: in a standard HOA, each iteration updates and checks all N solutions once. So, the total computational complexity of baseline HOA is 31 here, where (T·N) is the total number of solution evaluations and is the cost of evaluating each solution, which includes decoding the task and calculating the objective. Initialization of HOA: in a standard HOA, each iteration updates and checks all N solutions once. So, the total computational complexity of baseline HOA is 31 here, where (T·N) is the total number of solution evaluations and is the cost of evaluating each solution, which includes decoding the task and calculating the objective. Initialization based on OBL : For every first candidate , an opposite candidate is made and tested, and the better one is kept. This necessitates an additional assessment for each member of the population, resulting in an extra: 32 Initialization based on OBL : For every first candidate , an opposite candidate is made and tested, and the better one is kept. This necessitates an additional assessment for each member of the population, resulting in an extra: 32 OBL jumping during iterations: OBL is used in evolution with a chance of jumping . On average, candidates are evaluated in the opposite direction for each iteration. This adds across T iterations, The overall complexity of HOA-OBL, when combining these terms, is: 33 OBL jumping during iterations: OBL is used in evolution with a chance of jumping . On average, candidates are evaluated in the opposite direction for each iteration. This adds across T iterations, The overall complexity of HOA-OBL, when combining these terms, is: 33 Experimental setup and parameter configuration All simulations were implemented in MATLAB R2025a on a 64-bit Windows 11 workstation with the proposed HOA-OBL framework. Each experimental scenario contained a swarm of 20 UAVs and 200 spatially randomized tasks over a 500 × 500 m mission area, unless otherwise noted. To provide environmental diversity and eliminate initialization bias, UAVs and task locations were randomly initialized for each run. Five normalized capability metrics were taken into account to evaluate each UAV: sensor capability, transmission quality, wind robustness, residual battery energy, and payload capacity. Similarly, each task was defined by five normalized demand metrics: completion time, travel distance, energy consumption, urgency level, and resource requirement. By applying the proposed weighted evaluation framework, these multidimensional criteria were combined to guide the decisions of task allocation. For a fair comparative study, all optimization algorithms were implemented with the same population size of 25 and maximum number of iterations of 200. Each experiment was conducted independently for over 30 runs and the average performance metrics were reported to reduce stochastic variability. The comparison included convergence behavior, fitness value, computational runtime and quality of allocation. Figure 2 depicts the initial random spatial configuration of UAVs in red triangles and task locations in blue circles before optimization. UAV-TASK allocation simulation environment. Table 6 summarizes the simulation parameters for all experimental scenarios including computational platform, mission area dimensions, UAV swarm size, number of tasks, population size, iteration limits and the number of independent runs. These standardized settings were used for reproducibility and to provide a common evaluation framework for fair comparison of all optimization algorithms. Table 7 shows the algorithm specific parameter configurations for HOA-OBL and the benchmark optimization methods. These are population size, number of iterations and control parameters such as OBL jumping rate, PSO acceleration coefficients, ACO pheromone parameters and other variables specific to each optimizer. To ensure fairness in performance comparison among all the algorithms evaluated, uniform baseline conditions were preserved where possible. Table 6Simulation parameters.ParameterValueSimulation platformMATLAB R2025aOperating systemWindows 11Mission area500*500 mNumber of UAVs20Number of tasks200Population size25Max iterations200Independent runs30 Simulation parameters. Table 7Algorithm-specific parameter settings.AlgorithmParameterHOAPop = 25, iter = 200HOA-OBLPop = 25, iter = 200, jumping rate = 0.3PSOw=0.7, =1.5, =1.5ACOAnts = 25, evaporation = 0.5GWOPop = 25, iter = 200DLOPop=25, iter=200, exploration coefficient=2, adaptive hunting factor=0.5CLOPop=25, Iter=200, search coefficient=2, prey adaptation factor=0.5 Algorithm-specific parameter settings. Parameter sensitivity analysis The robustness of the HOA-OBL framework with respect to different hyperparameter settings was investigated by performing a sensitivity analysis on the population size , the maximum number of iterations , and the OBL jumping rate . The evaluation was based on the convergence speed, objective value and runtime efficiency. The population size was set to 25 for each optimization algorithm, and the algorithm was run for 200 iterations. Results were computed over 10 independent runs. Since all comparative convergence analyses have been reported using mean fitness values over independent runs, the parameter sensitivity results are also reported in terms of mean fitness, to keep consistency. Table 8 presents the sensitivity analysis of HOA-OBL on key hyperparameters. The impact of population size, maximum iteration number and OBL jumping rate on best fitness values and computational runtime are studied. The results confirm the robustness of the chosen parameter configuration, while highlighting the trade-off between the quality of convergence and computational efficiency. Table 8HOA-OBL parameter sensitivity under varying population size, iteration count, and OBL jumping rate.Population sizeIteration countOBL JUMPING RATE()Best fitness valueRuntime(s)201500.10.91814.8202000.30.93118.5252000.30.94221.2252500.50.94426.8302000.30.94624.9302500.50.94831.4402000.30.94938.6 HOA-OBL parameter sensitivity under varying population size, iteration count, and OBL jumping rate. Results Performance evaluation Table 9 provides the performance metrics obtained in the UAV task allocations for all algorithms. The objective fitness value shows the global optimization efficiency, combining factors such as task completion time and energy consumption. The proposed HOA-OBL algorithm reaches the highest fitness value of 0.942, followed by standard HOA with 0.865, whereas PSO and GWO exhibit comparatively lower optimization performance. Table 9Performance comparison of the proposed algorithm with existing approaches.MetricHOAHOA-OBLGAPSOGWODLOCLOObjective function value0.8650.94800.8240.7960.7110.8890.920Makespan (s)322265345368412298281Total energy consumption (J)Task completion rate (%)98.3100.096.895.594.298.999.4UAV load balance index (LBI)0.2140.1420.2680.2440.2900.1810.159 Performance comparison of the proposed algorithm with existing approaches. The makespan represents the total time required for all UAVs to complete their assigned tasks. The proposed HOA-OBL attains the minimum makespan of 265s, whereas GWO records the maximum value of 412s. The task completion rate indicates the percentage of tasks successfully executed within the simulation time window. The HOA-OBL reaches a 100% completion rate, outperforming HOA with 98.3% and GA with 96.8%. The UAV’s load balance index measures the workload distribution consistency between UAVs, with lower values indicating better balance. The HOA-OBL framework produces the most balanced task allocation among all evaluated methods. Finally, the computation time denotes the average operational time necessary for each algorithm to converge. Although HOA-OBL incurs a slightly higher computational cost than the other approaches, all algorithms converge within closely matched durations, adapting the efficiency of the proposed optimization framework. Convergence comparison The average convergence curves of the classical and recent metaheuristic algorithms are shown in Fig. 3 for 200 iterations. The algorithms are HOA, HOA-OBL, PSO, GA, GWO, DLO, and CLO. The proposed HOA-OBL is always faster in the early stage of optimization than other compared methods and finally converges to the highest mean fitness of all compared methods. This result is mainly attributed to the effect of opposition-based learning and HOA’s exploration-exploitation strategy, where the opposite candidate solution enhances the population diversity and reduces the premature convergence. Among recent optimizers, both CLO and DLO have shown better convergence than some of the traditional methods, indicating their competitiveness. However, both remain inferior to HOA-OBL in terms of final fitness quality and convergence stability. GA shows moderate convergence whereas PSO converges more gradually due to weaker global exploration in later iterations. GWO has the lowest convergence performance, which is mainly due to the reduced adaptive search capability and weaker exploitation efficiency under the evaluated UAV task allocation conditions.In brief, the proposed HOA-OBL can achieve both the accelerated convergence and the better final fitness because it is able to utilize the OBL mechanism which preserves the population diversity at the local level and enhances the global exploration. Convergence Comparison with other algorithms. Discussions Statistical robustness Figure 4 demonstrates the statistical robustness of HOA-OBL in comparison with conventional and recent benchmark metaheuristic algorithms over multiple independent simulation runs. Each boxplot summarizes the distribution of final fitness values. The central line indicates the median, the box boundaries show the 25th and 75th percentiles, and the whiskers display the full spread of observed values.While recent algorithms such as CLO and DLO outperform several classical baselines, HOA-OBL exhibits better robustness and allocation quality further proving its effectiveness for complex UAV task allocation under heterogeneous mission conditions. Statistical robustness of classical and recent metaheuristic algorithms. Computational time and convergence efficiency analysis Table 10 presents the comparison of computational time and convergence efficiency of classical, recent and proposed metaheuristic algorithms. Mean fitness, average runtime and iterations to near-convergence are reported. HOA-OBL has a small runtime overhead, but it consistently achieves better fitness and faster convergence, showing its balanced performance in terms of quality, speed and efficiency. Figure 5 Friedman ranking analysis was performed using fitness inversion to ensure the level of optimization performance was consistent with the statistical rank. The results show that HOA–OBL achieved the best overall mean rank over repeated simulations, confirming its superiority over benchmark algorithms. Table 10Computational time and convergence efficiency comparison.AlgorithmMean fitnessRuntime (s)Iterations to near-convergenceHOA-OBL0.948023.4141HOA0.81418.6162PSO0.754714.8178GA0.87220.9156GWO0.57513.5 > 200 DLO0.88921.8150CLO0.92022.6145 Computational time and convergence efficiency comparison. Friedman mean ranking of classical and recent metaheuristic algorithms. Scalability analysis From Fig.6 , it can be seen that the proposed algorithm is the most effective one since it exploits the additional UAVs while maintaining the high performance as the number of UAVs increases.Similarly, in Fig. 7 the proposed HOA-OBL sustains a high fitness level even as the number of tasks rises from 10 to 200, while the competing algorithms experience notable degradation. Comparison of fitness values with varying UAV counts. Comparison of fitness value with varying task counts. Convergence scalability under varying swarm sizes Figures 8 and 9 together look at how effectively HOA-OBL work in swarm-level settings. Figure 8 shows that the number of iterations needed to get close to convergence goes up slowly as the number of UAVs in the swarm grows past 50. This shows that controlled convergence degradation is happening. Figure 9 further confirms that the proposed method keeps stable fitness improvement all through the optimization phase and does not show early stagnation, even though it takes longer for higher-dimensional cases to converge. These findings confirm the efficacy of HOA-OBL for extensive UAV swarm task distribution. Iterations-to-convergence vs number of UAVs. Mean convergence curves for 20, 50, 100 UAVs. Ablation study of HOA-OBL components An ablation study was performed to isolate the contributions of opposition based learning, the multidimensional weighted framework and constraint repair by incrementally activating each component within the optimization process. Table 11 summarizes the ablation study of HOA-OBL, where OBL, the weighted evaluation framework and constraint repair was progressively enabled. Reported metrics include mean fitness, feasible solution ratio and running time. Results show that OBL promotes exploration, the weighted framework improves UAV-task matching, and constraint repair guarantees feasibility.The complete HOA-OBL configuration achieves the best overall performance, confirming the necessity of integrating all components for maximum efficiency and robustness. Table 11Ablation study of HOA-OBL components.VariantMean fitnessFeasible solution ratio (%)Runtime (s)HOA0.81481.218.6HOA+OBL0.88988.420.4HOA+OBL+WF0.92093.122.1Full HOA-OBL0.948096.423.4 Ablation study of HOA-OBL components. Figure 10 illustrates the ablation study of the proposed HOA-OBL framework, jointly reporting the mean final fitness and the feasible solution ratio in the stages of integrating progressive components. The results show that the convergence quality is improved by using the opposition-based learning, such as improving the population diversity and avoiding the premature convergence. The multidimensional weighted framework enhances allocation accuracy further by refining the evaluation of heterogeneous UAV-task compatibility. The constraint repair mechanism improves the feasibility of the assignments and the robustness of operations under practical mission constraints. The progressive improvement observed across all ablation variations implies that each component contributes successive improvements to the optimization performance. The full HOA-OBL configuration consistently achieves the best fitness and feasible solution ratio, validating that its strong performance is due to the complementary integration of exploration enhancement, multidimensional evaluation, and feasibility enforcement, rather than any single isolated mechanism. Ablation analysis of HOA-OBL components. Dynamic and complex scenario evaluation Additional experiments under dynamic and constrained conditions were conducted to further evaluate the adaptability and operational robustness of HOA-OBL over static mission scenarios. random task arrivals, unexpected UAV failures and optional no-fly-zone restrictions, to name a few. The results evaluate the framework’s ability to retain allocation efficiency, recover from mission disturbances and maintain near-convergence under varying environmental challenges. As shown in Fig. 11, the introduction of new tasks at iteration 75 during mission execution leads to a temporary decrease in mean fitness due to the increased allocation complexity. HOA-OBL, however, can dynamically reallocate resources and gradually recover the performance to near-convergence without restarting the whole optimization. This behavior demonstrates its strong real-time adaptability in dynamic task environments. The robustness of the HOA–OBL framework to the unexpected UAV failure is shown in Fig. 12. When the available UAV resources are reduced mid-mission, the framework experiences a temporary drop in fitness, but is able to quickly reallocate pending tasks among the remaining UAVs. This recovery behaviour shows the fault tolerance and robustness of reassignment of proposed approach. By the no-fly-zone constraints given in Fig. 13, the framework maintains the feasible task allocation by including the spatial restrictions to the reassignment process. This shows its applicability in operationally constrained and regulation sensitive environments.The overall dynamic and constrained experiments demonstrate that the HOA-OBL is not restricted to static optimization and is able to cope with re-assignment, faults and environmental changes reliably.This highlights its practical relevance for UAV missions under uncertain and changing conditions. Dynamic task arrival. UAV failure recovery. No-fly zone map. UAV task execution time analysis Figure 14, where the UAV and the task are mapped by a grey line which represents the respective task assigned to the UAVs. The X and Y axes are the coordinated points in the mission area and denote the position of the UAVs with respect to task locations. The legend is on the right, it provides the task identifiers and allows the traceability of each task allocation. The figure depicts a spiderweb-like structure that emphasizes a large distribution of both UAVs and tasks, which reflects the dynamic and complex behavior of the allocation process. The time for each UAV is shown in Figure 15 , where the horizontal axis represents the time in seconds and the vertical axis represents the identifiers of each UAV. The horizontal bars show time spent traveling iin blue and time spent on-site executing the task in green. The final bar of the finish is the total mission time. Idle gaps in the bars represent waiting UAVs and provide a visual indication of an unbalanced load. This is why HOA-OBL obtains a lower makespan and better load balancing than the baseline algorithms. The good performance of HOA-OBL is attributed to the combined effects of the opposition-based learning, multidimensional weighted evaluation and constraint repair. Opposition-based learning expands the search space exploration, thus increasing the diversity of population and avoiding premature convergence. The weighted task-UAV framework can promote the heterogeneous capability matching that can lead to a more accurate allocation decision. The constraint repair mechanism further enhances the solution feasibility and stabilizes the convergence under operational constraints. These components together explain the improved fitness, robustness and scalability of the framework over benchmark methods. Despite these advances, the study remains still simulation based and does not capture real-world complexities that includes decentralized communication delays, dense urban obstacles, irregular three-dimensional terrain and hardware-specific uncertainties. Moreover, the existing single objective weighted formulation may not adequately capture broader trade-offs between fairness, resilience, and long term energy sustainability. UAV-TASK mapping. Gantt chart of UAV-TASK execution. Conclusion and future work In this paper, an enhanced UAV swarm task allocation framework based on the Hippopotamus Optimization Algorithm with Opposition Based Learning was proposed. The framework was designed to address key challenges in multi UAV environments including heterogeneous capability matching, premature convergence and constrained task allocation. By combining HOA’s exploration-exploitation dynamics with opposition based candidate generation, a multidimensional weighted evaluation model, and feasibility aware constraint repair, the approach significantly improves allocation quality , convergence efficiency, and operational robustness compared with both classical and recent metaheuristics. Experimental results show that HOA-OBL delivers a 17.7% improvement in mission completion efficiency, reduces the overall mission span to 265 seconds, achieves a 100% task completion rate, and consistently produces the highest mean fitness value of 0.9480, while maintaining strong statistical robustness across repeated runs. Although opposition based learning and repair mechanisms introduce moderate computational overhead, the framework achieves a good balance between runtime and optimization quality, validating its scalability and appropriateness for complex UAV swarm coordination settings. The strong performance of HOA-OBL is primarily attributed to the complementary integration of its key mechanisms. Opposition based learning results in increased population diversity and reduced stagnation in local optima. The weighted evaluation framework upgrades the heterogeneity assessment of UAV-task compatibility . The constraint repair strategy enhances feasible assignment generation under realistic mission constraints. Moreover, the dynamic scenario evaluations reveal that the framework has a good reassignment capability under random task arrivals, UAV failures and constrained operational environments. These results extend its importance beyond static settings of task allocation. Despite these promising results, several limitations still exist. The present validation is simulation based with partial simplifications of communication and environmental assumptions including abstraction of sensing uncertainty, local wind estimation and limited dynamic models. It did not fully consider real-world deployment factors like decentralized communication delays, non-uniform 3D terrain, dense urban obstacles, and hardware-specific uncertainties. Furthermore, the current formulation considers weighted single-objective optimization, which may not capture broader Pareto trade-offs among energy, latency and resilience. Future work will extend HOA-OBL to fully decentralized real- time UAV swarm coordination, multi-objective Pareto optimization, and more realistic 3D mission environments including mountainous terrain, urban no-fly zones, and communication-impaired settings. Overall, the proposed HOA-OBL framework provides a strong and scalable foundation for next-generation UAV swarm task allocation and shows strong potential for wider autonomous mission planning applications. Author contributions Conceptualization, methodology, software, validation, formal analysis, resources, data curation, writing original draft preparation, visualization-Keerthana G; Investigation, Writing review and editing, Supervision-Padmanaban R; All authors have read and agreed to the published version of the manuscript. Funding Open access funding provided by Vellore Institute of Technology. Data availability The datasets generated and/or analyzed during the current study are available from the corresponding author upon reasonable request. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement The datasets generated and/or analyzed during the current study are available from the corresponding author upon reasonable request.