key: cord-0976745-8wvdc0ov authors: Bella, Salima; Belalem, Ghalem; Belbachir, Assia; Benfriha, Hichem title: HMDCS-UV: A concept study of Hybrid Monitoring, Detection and Cleaning System for Unmanned Vehicles date: 2021-05-25 journal: J Intell Robot Syst DOI: 10.1007/s10846-021-01372-8 sha: 3fcfaba1d311f7bdd874a3e50b1ba74083be24dc doc_id: 976745 cord_uid: 8wvdc0ov Incidents of hydraulic or oil spills in the oceans/seas or ports occur with some regularity during the exploitation, production and transportation of petroleum products. Immediate, safe, effective and environmentally friendly measures must be adopted to reduce the impact of the oil spill on marine life. Due to the difficulty to detect and clean these areas, semi-autonomous vehicles can make a significant contribution by implementing a cooperative and coordinated response. The paper proposes a concept study of Hybrid Monitoring Detection and Cleaning System (HMDCS-UV) for a maritime region using semi-autonomous unmanned vehicles. This system is based on a cooperative decision architecture for an unmanned aerial vehicle to monitor and detect dirty zones (i.e., hydraulic spills), and clean them up using a swarm of unmanned surface vehicles. The proposed solutions were implemented in a real cloud and were evaluated using different simulation scenarios. Experimental results show that the proposed HMDCS-UV can detect and reduce the level of hydraulic pollution in maritime regions with a significant gain in terms of energy consumption. Nowadays, research in swarm robotics is growing due to their robustness, parallelism and flexibility. Unlike distributed robotic systems, swarm robotics focuses on a large number of robots and promotes scalability [15, 24] . Several applications of these systems concern environmental exploration, resource detection and monitoring, cleaning, rescue, food search in agriculture, etc. [15] . Many applications, such as surveillance, perimeter / surface detection and cleaning of polluted zones, have several ecological uses. Maritime activity has gradually increased in recent years. Many ships carry products such as hydraulics that can harm the environment. These products can produce high levels cal actions) such as controlled combustion, solidifiers (absorbents), skimming, oil oxidation, manual recovery, dispersion, bioremediation / biodegradation, etc. which are largely limited by marine conditions [19, 20] . While these treatments are important to rapidly control the spread and drift of oil, they are not suitable for ecological restoration. All the methods and techniques mentioned are only relevant to the elimination of oil spills in water, but this is not an easy task, as the spill usually spreads more widely over time. The only technology used to collect the spilled oil is the use of dams, which are large floating barriers that supplement the oil spill and then lift the oil from the water between two ships [11] , as shown in Fig. 1 1, 2 . This process is long and costly, as ships carrying large containers have to leave the zone several times to dispose of the recovered oil. In the case of turbulent water, the spill spreads more widely, making it difficult to complete the cleaning process. Instead of using many large barges in the working zone, a swarm of robots equipped with skimmers and dams was proposed in [11] to collect the spilled oil in one place and limit its spread. Only vessels equipped with containers can be present to collect and store the oil. Such a swarm can also be sent to prevent the oil spill from moving to the shoreline, port or other such zone and save lives more effectively. Based on this technique, a hierarchical decision architecture is illustrated in the proposed hybrid system (HMDCS-UV) for a better management and easier control between the different unmanned vehicles. In addition, the proposed system is efficient, robust and scalable compared to the self-organized approach which hardly allows scalability with increased complexity for efficient management of its entities. Centralized management has a central node with deterministic decision-making capability and easy to implement coordination. This central node has a global view of the unmanned monitoring and cleaning vehicles activities. Distributed management begins when the monitoring vehicle is assigned to a region. It moves to its region according to a proposed trajectory planning method, detects dirty zones based on an unsupervised classification method specific to image processing, monitors and supervises its cleaning swarms in its region. This article proposes a concept study of Hybrid Monitoring, Detection and Cleaning System (HMDCS-UV) for a maritime region using heterogeneous semiautonomous unmanned vehicles. The HMDCS-UV is based on a cooperative hybrid architecture of an Unmanned Aerial Vehicle (UAV) to monitor and detect dirty zones (hydraulic spills) and clean them from a swarm of Unmanned Surface Vehicles (USV). A general coordinator is proposed to 1 https://www.itopf.org/uploads/translated/ TIP3FRUseofBoomsinOilPollutionResponse.pdf 2 https://www.arzewports.com/?pages=page&rub=13 manage all the tasks and coordinate these vehicles. The proposed UAV is supposed to be equipped with on-board sensors that allow it to move, detect and locate dirty zones and update the nautical chart using the specific planning and detection methods. After receiving and analyzing the collected data, the general coordinator assigns a swarm of USVs with the processed information to clean each dirty zone. This swarm navigates to the assigned dirty zone and cleans it according to the proposed solutions for trajectory planning and cleaning. The paper is organized as follows: Section 2 presents some related work from the literature; Section 3 describes the proposed system for different unmanned vehicles with methods for detection, trajectory planning and cleaning; Section 4 illustrates an example to simulate the operation of the HMDCS-UV; in Section 5, the proposed HMDCS-UV is compared with related work; finally, Section 6 concludes this paper and provides some guidance for future research. The technical literature abounds with various solutions that address the problem of pollution of land, air and / or sea environments, using methods and algorithms for monitoring, detection, reduction / mitigation and cleaning. For example, the design of autonomous units (autonomous vessels / drones) was developed in the framework of an EU-MOP [17] research project for the elimination of marine oil pollution, which are capable of mitigating and eliminating the threat of small and medium sized spills. These vessels are released in the zone of the spill, automatically monitor (using appropriate sensors) the specificities of the oil concentration of the spill and locally apply mechanical or chemical countermeasures. An alternative design of a multirobot autonomous aquatic vehicle system was proposed in [16] for lake cleaning and fisheries maintenance. The robots use tactile sensors and wireless communications to independently navigate and collectively perform cleaning operations such as removing surface impurities, pumping oxygen into the water, spraying chemicals, distributing food to appropriate locations while measuring water quality. Then, a chemical leakage localization and cleanup method was implemented in [25] using a robotic swarm and based on a bio-inspired exploration method. This method is based on a combination of two bio-inspired behaviors: aggregation and pheromone tracking. The main objective of the robots is to follow pheromone trails to find the source of a chemical leakage and then carry out a decontamination task by aggregation at the level of the critical zone. A new swarm robotic system was proposed in [11] to locate and collect an oil spill on the water surface (ocean, river, lake, etc.). A coordinator determines the position and Fig. 1 Examples of dams collecting a hydraulic spill the center of the oil spill using a GPS receiver, then a barge carrying the robots goes to the work site. Depending on the state of the oil spill, the coordinator may send a swarm of robots to surround and collect the spilled oil, then place the barge with oil suction equipment and move it to another location to safely remove the oil. The distributed system presented in [15] allows monitoring, recovery and containment of a resource using a swarm of homogeneous drones at low cost. A microscopic model of the swarm is presented, which defines individual behavior and is capable of locating and marking the perimeter of an oil spill. The functioning of a macroscopic model is analyzed and shows the trend of the swarm for a large number of agents. The authors of [15] suggest to use the signal power (at a given frequency) for obstacle avoidance tasks. In [18] , an experimental system was proposed for the autonomous control and coordination of the automatic spill response dam towing operation when using unmanned vessels. This system comprises two ASVs (Autonomous Surface Vehicles) and a ground station. Once the operation has begun, the ASVs tow the dam towards the target, minimizing the towing effort, close to the target that the ASVs deploy and advance, and finally approach, confined to the spill; then the drive moves to the destination. Another marine robotics system was designed in [19] to locate and quantify surface water (oil-based) pollution in lakes and ponds using aerial and marine robots, taking into account the influences of nearby buildings and trees. An aerial mobile robot equipped with two cameras such as the FLIR (Forward Looking Infra-Red) thermal imaging camera to locate and detect oil spills based on water temperature day and night, and a digital camera for trajectory planning. Quantification consists of a marine robot (boat) with a spectrometer on board. This marine robot controls its movement by fuzzy logic, distinguishes the received signal from the oil and water spectrum thanks to its integrated ultrasonic sensor, and collects samples of oil spilled by the suction pump. The neural network technique analyses the images received to differentiate between the different types of pollution on the water surface. E-drones (Environmental Drones) is a new approach, was proposed in [14] for the large-scale elimination of air pollution. The environmental drones used make it possible to independently monitor air quality, detect the presence of pollutants and measure their concentrations (in carbon dioxide (CO2), carbon monoxide (CO), etc.), implement an option for appropriate reduction at a specific altitude (E-altitude) to guarantee the elimination of these pollutants, then fly to their ground locations. When multiple environmental drones are used in different locations, custom software generates an Air Quality Health Index (AQHI) map of the region for current and long-term environmental analysis. Thus, a method of controlling air pollution based on an air purifying drone system was presented in [26] to clean or reduce the amount of pollutants present in the areas near the industries or highly populated cities. The air purifier drone will disunite pollutants by spraying water and chemicals into the atmosphere. Another hierarchical hybrid approach for the heterogeneous cooperation of unmanned vehicles (HA-UVC) was proposed in [1] . The proposed HA-UVC allows the cooperation of an unmanned aerial vehicle (UAV) to monitor an ocean region and a swarm of UAVs to clean up dirty zones. The UAV is supposed to be equipped with on-board sensors that allow it to locate the dirty zones using the proposed methods of travel, discretize its environmental map and update it with the information collected about the dirty zones. After receiving and analyzing the data collected according to the color of the water, the general coordinator (represented by a laptop computer and guided by a human operator) assigns the explored map to the swarm SVU to clean each dirty zone. This swarm moves to the assigned dirty zone according to the proposed solutions for trajectory planning, namely "Modified-GA" and "Proposed-CCA". In addition, these solutions incorporate a method to avoid static obstacles. When this swarm arrives in the dirty zone, it starts moving and cleaning the dirty cells according to a proposed algorithm. The proposed HA-UVC is complemented by a method of managing SUV failures during the execution of its cleaning task by measuring its energy quantity according to an energy threshold. Thus, a system for monitoring, detecting and cleaning up hydraulic spills was proposed in [22] using a UAV developed with integrated software. The UAV can be used as a "flying officer" that can monitor each part of the oil platform using a light Canon XS260 camera to capture images, a light detection and ranging (LiDAR) sensor for navigation and collision avoidance, and a GPS to map the spill zone where a geo-referenced method is used. The UAV can also clean up the spill by spraying chewing microbes with oil on the ocean surface. In addition, a fluorosensor system carried by a commercial drone was built in [28] for monitoring laser-induced fluorescence from the aquatic environment. The present version of the fluorosensor system could only be used at night time, which is a clear drawback compared to pulsed lidar systems with range gating, allowing daytime use, while again still functioning better in low ambient light level conditions. Also, a new system architecture derived from the integration of a low-cost laser-based network of detectors for pollutants interfaced with a more sophisticated layout mounted on an UAV was proposed in [29] to identify the nature and the amount of a release. Once this system is in place, the drone will be activated by the alarm triggered by the laser-based network when anomalies are detected. The area will be explored by the drone with a more accurate set of sensors for identification to validate the detection of the network of Lidar systems and to sample the substance in the focus zone for subsequent analysis. A new elite group-based evolutionary algorithm (EGEA) was presented in [13] for maximum ocean sampling by several Unmanned Marine Vehicles (UMVs). The EGEA integrates a group-based framework and two proposed elitist selection methods, GIES (Group Individual Elitist Selection) and WPES (Whole Population Elitist Selection) to facilitate the selection of preferred solutions to be passed on to the next generation. The EGEA trajectory planners are based on the Simulated Annealing (SA) algorithm and Particle Swarm Optimization (PSO) to find the trajectories of the UMVs and to collect the maximum information on the studied regions. Mixed integer linear integer programming (MILP) was used by EGEA to solve the problem of adaptive sampling. Other algorithms were introduced in [21] for the detection of oil spills using MIMO (Multipleinput multiple-output) radar remote sensing integrated on a UAV. The first Single-Frequency Single-Observation (SFSO) sensor with power reflection coefficients is used to detect the presence of oil. The results show that the performance of this type of detector is related to oil thickness values, where they operate for one range but fail for another. An improvement of this detector is the Dual-Frequency Single-Observation (DFSO) detector where two frequencies of electromagnetic waves are used. Analysis of the performance of the second detector allows accurate detection. An improvement of both detectors is the use of multiple observations. A metaheuristic algorithm based on the simulated annealing methodology to generate the UAV movement directions, integrated in a platform was proposed in [27] to measure atmospheric pollutants and to monitor contamination sources and treat them in real time. Another pollution-driven UAV control (PdUC) algorithm was presented in [30] to guide drones equipped with offthe-shelf sensors to perform air pollution monitoring tasks. This algorithm makes it possible to autonomously monitor a specific area by prioritizing the most polluted zone. In particular, it is able to find the most polluted areas more accuracy and cover the surrounding area, thus obtaining a complete and detailed pollution map of the target region within the time bounds defined by the UAV flight time. In addition, the authors of [31] were developed a realtime detection and monitoring system for the coronavirus (COVID-19) from the thermal image integrated into a UAV based on the Internet Of Things (IoT). The proposed system can detect ground surface temperatures from a height above the ground. Furthermore, the proposed design has capability for using Virtual Reality (VR), so the live video scanning process will be monitored through the VR screen to make it realistic and less human interaction. Thus, the diagnosis of the screening process will be less time consuming and less human interactions that might cause the spreading of the coronavirus faster. A trajectory planning control method was developed in [20] for an Autonomous Surface Vehicle (ASV) that is capable of mitigating and bypassing the propagation of an oil spill while deploying micro-organisms and nutrients (bioremediation) with the collaboration of a UAV. The UAV is responsible for detecting the zone of the oil spill with the thermographic camera and controlling internal leakage zones by spreading freeze-dried indigenous microbial consortium dust on the oil spill, while the ASV releases the product mixed with water on the boundary zones of the line. The potential field method is used so that the ASV can plan its route, cover the entire zone and at the same time robustly avoid the oil spill. In addition, a multi-resolution navigation algorithm was presented in [12] to clean up oil spills in dynamic and uncertain environments using autonomous vehicles as the sole agent. A proposed algorithm for adaptive decision making based on sensory information, provides a complete coverage of the search zone for cleaning that does not suffer from the problem of local minima using potential field methods. It is in this sense that this present work aims to provide a new solution to the challenges of maritime pollution by proposing a concept study of Hybrid Monitoring, Detection and Cleaning System for Unmanned Vehicles (HMDCS-UV). This proposed system enables cooperation and coordination between heterogeneous semi-autonomous air-sea unmanned vehicles to monitor maritime regions and clean their dirty zones. Thus, the solutions are proposed for the trajectory planning towards the dirty region / zone, the monitoring of a region, the detection and cleaning of the dirty zones and the supervision of unmanned vehicles. This paper aims to propose a hybrid system for the surveillance, detection and cleaning of dirty / polluted maritime zones using the cooperation and coordination of semi-autonomous unmanned vehicles (HMDCS-UV). Thus, the HMDCS-UV is seen as an improvement and an extension of the HA-UVC solution [1] . In this section, a process for preventing and combating oil pollution in the port of Arzew (Algeria) is presented. Based on this process, the HMDCS system is described with its architecture, its constituent entities, the solutions proposed for supervision and detection, trajectory planning and cleaning, and their operation. Marine oil pollution from ships is a real problem that threatens the port of Arzew (oil port in western Algeria). Given the impact of these harmful products on the port and industrial activity of the site, prevention has become a constant concern for the authorities. The Arzew port has considerable material (tugs, oil waste recovery barge, self-floating containment dam, etc.) and human resources (people working in the anti-pollution cell, mooring staff, etc.) to combat oil pollution, and works in collaboration with other services at the port to intervene in the event of spills. -Presentation of the port of Arzew: the port of Arzew is a rather important port complex, it is the largest oil port in Algeria. It is composed of 2 ports, the first one is the port of Arzew which is the old port whose construction dates back to the Roman period; it underwent several modifications and extension works. Today, it receives two types of goods, general goods and hydrocarbons. The second is the recently built (1975) (1976) (1977) (1978) port of Bethioua (Arzew El-Djedid), which mainly supplies the liquefaction of natural gas and oil, crude oil and condensates. Both ports are managed by the Arzew Port Company (APC), which is attached to the Directorate of Captaincy (DC) 3,4 . -Pollution prevention: pollution prevention is an integral part of the daily missions of the Arzew port company. This mission is reflected in particular by traffic monitoring measures enabling permanent monitoring on the VHF marine radio (Very High Frequency), monitoring the port area, identifying incoming and outgoing ships, etc. As part of the fight against marine oil pollution, the directorate of captaincy has implemented an intervention plan, this plan consists of action planning aimed at organizing the response to a spill taking into account the collaboration with various services at the port level: civil protection, towing and mooring, environment, the Management and Operating company for marine oil Terminals (MOT), and the Territorial Grouping of Coast Guards (TGCG). To this end, the appearance of an oil spill in the port (due to an accidental or deliberate spill) is directly reported to the directorate of captaincy and to the territorial grouping of coast guards 5 . A hierarchical organization chart is proposed for a new Central Unit (CU) for the Arzew port company. The hybrid system is intended as an improvement and extension of the proposed organization chart for the directorate of captaincy. Figure 2 shows the proposed reporting structure, which is composed of: -Master station: it is composed of a general coordinator and several central unit officers. The general coordinator is represented by a laptop, guided by a human operator and / or an officer. This coordinator is responsible for several tasks such as launching and control of vehicles with / without driver used, use (processing) and storage of data in the database, regulation of the movement of the navigation of vehicles and coordination with other services (police and security, armament and maintenance). -Security, Armament and Maintenance (SAM) department: it comprises the three departments mentioned concerning the three services, namely security, police and ship movements, armament and naval maintenance. i) System architecture A hierarchical hybrid architecture is proposed in the HMDCS-UV (Fig. 3) ; it comprises a maritime force base, a central unit, a surveillance vehicle to monitor the maritime region, and a swarm of cleaning vehicles to clean a dirty zone (each swarm is guided by a leader). The maritime force represents the base of the TGCG, it includes the coastguards who work in collaboration with the port for surveillance and intervention in pollution response operations in the port. The central unit consists of a command room, a SAM department, a living base and a database. The main room includes a general coordinator who stores and consults the data in the database. The SAM department interacts with its agents (SAM agent ) in the living base by VHF marine radio, and with the control room by VHF and other means of communication such as WiFi. The master room interacts with the base of life and the surveillance vehicle via WiFi, and with the TGCG base via VHF and WiFi. This WiFi network also allows the exchange of messages between the surveillance vehicle and the leader of each swarm of cleaning vehicles. The hierarchical decision on the proposed architecture, illustrated in Fig. 3 , is made at the first level of the Maritime Force. This force includes the TGCG base, which represents the core memory of the port. It works in collaboration with the port to monitor it, intervene in pollution response operations and initiate requests to the central unit (i.e. the master's room). The latter is located on the second level, which includes a general coordinator. This coordinator represents the central memory of the central unit and has the highest decision for the execution of various tasks such as the launching and control of the used manned / unmanned vehicles that are located in the living base, the use and storage of data in the database and the coordination with SAM agent . The surveillance UAV is responsible for a lower decision located at the third level. Its role is to monitor a maritime zone and supervise its swarms of clearance vehicles. These swarms are located on the fourth level and are composed of a lead vehicle and follower vehicles. Their objective is to carry out the cleaning operation of the dirty zone according to the energy availability of each member. Each leader has two necessary roles; it is responsible for the tasks of its followers, and also shares and cooperates with them in the cleaning action. Finally, each vehicle of the swarm has a local memory which constitutes the fifth level, and it can communicate with its neighbors. The working environment is described by the elements listed below [1] . Before defining these elements, Table 1 presents a description of some acronyms of the entities used and Table 2 shows the applied parameters in the proposed system. (t mr ) for a region r; Cleaning task (t cz ) of a dirty zone z; Cleaning supervision task (t sz ) of a zone z; Launch task (t l ) of the various previous tasks (allocation, preparation, monitoring, cleaning and cleaning supervision task). -Set of vehicles: the different vehicles used in HMDCS-UV are the same vehicles used (UAV mr , USV cz , Leader cz and V ehicle rec ) in HA-UVC [1] except the V ehicle nt which is a floating object (nautical boat). It is a means of maritime transport used to embark port officers, pilots, customers, anti-pollution equipment, water supply and blasting. -Set of agents: the different agents of HMDCS-UV are: General crd : a general coordinator is represented by a computer which contains coordination software, guided by a human operator. It is responsible for: the base of life, the data of the regions T hreshold A fixed threshold which makes it possible to classify the coordinates of the zone according to the degrees of the cells "List Degree cell " A list of identifiers of USV cz List zone (List Degree cell , List P osition cell , P osition zone ) A list of coordinates of the dirty zone A list of the coordinates of the dirty zone compared to the dirt threshold compared to the degrees of dirt (List Degree cell ) P os Start (x, y) A starting position of USV cz / UAV mr from the base of life. This position is a Cartesian coordinate (x, y) P os End (x, y) A final position represents the arrival of USV cz / UAV mr at the region / dirty zone. This position is a Cartesian coordinate (x, y) P os Gol (x, y) A goal position means the next position of USV cz / UAV mr in its movement. This position is a Cartesian coordinate (x, y) List P osE (x, y) A list of Cartesian coordinates (x, y) representing the final positions (to reach the dirty zone) List P osB (x, y) A list of Cartesian coordinates (x, y) representing the border positions of the dirty zone List tabooP os(I d USV , P os Gol (x, y)) A list to save the positions already crossed by the USV cz . It contains the USV identifier (I d USV ) which has already visited the position P os Gol where its Cartesian coordinate is the pair (x, y) in the grid A list of dirty cells to memorize the cells already cleaned by USV cz . It consists of the USV identifier (I d USV ) and its cleaned cell Cell(x, y). This cell represented by the Cartesian couple (x, y) in the grid List distCost(link ij , Cost ij ) A list of distances between positions marked as an energy cost. It is composed of a link link ij between the position i and the position j with its energy cost Cost ij P arameters start−up (M) (I d region , P osition region , P os Start (x, y), P os End (x, y), Map atmosphere space , Map nautical space ) A list of parameters to start monitoring for each UAV mr which represents the region's identifier and position, the start and end position, the map of the atmosphere space and the nautical space map before arriving at the region P arameters start−up(C) (I d region , I d zone , I d Supmr , I d LeaderCZ , P osition zone , P os Start (x, y), P os End (x, y), List P osE (x, y), List P osB (x, y)) A list of cleaning startup parameters for each USV cz which represents the identifier of its region, its zone, its supervisor, its leader and the position of the zone with the start and end positions, the list of end positions and the list of borders List of UAV mr characteristics: the identifier of a region, the identifier of a UAV (I d UAV mr ), the deplacement energy consumption from the base of life to the region (Cons ED1 ), deplacement energy consumption in the region (Cons ED2 ) and monitoring energy consumption List of characteristics of USV cz : the identifier of a zone (I d zone ), the identifier of a USV (I d USV cz ), the identifier of the supervisor (I d Supmr ), the deplacement energy consumption from the base of life to the dirty zone (Cons ED1 ), the deplacement energy consumption in the dirty zone (Cons ED2 ) and cleaning energy consumption (Cons EC ) and dirty zones, the use (processing) and storage of data in the database, the launching of tasks / missions and the allocation / diffusion of tasks to vehicles and SAM agent . Sup mr : a supervisor vehicle is the unmanned aerial vehicle (UAV mr ). It allows to: monitor regions, dirty zones and swarms of cleaning vehicles; supervise unmanned vehicles, request / inform the cleaning vehicle by tasks; return data and results to the general coordinator. Leader cz : it is an unmanned surface vehicle which has two roles: it is an intermediary between the supervisor Sup mr and the swarm USV cz , and cooperates with its followers in the cleaning operation. SAM agent : a SAM agent. This agent represents the Security, Armament and Maintenance (SAM) department for the execution of the various tasks requested by the General crd . -Set of regions: the monitored maritime space is divided into a maritime force base, a central unit and maritime regions. The region is made up of two subspaces; an atmospheric subspace where there are UAV mr and a nautical subspace with swarms of USV cz , V ehicle rec , V ehicle nt , dirty zones and can be the base of life. -Set of base of life: it is an zone (which can be a boat, a ship, an island or a coast) to store a fixed number of UAV mr , USV cz , V ehicle rec and V ehicle nt . -Set of database: these are databases (servers) to store and save all the data and characteristics of the maritime space as well as the different vehicles with / without pilot used. -Set of dirty zones: represents the dirty part where water pollution is found, for example hydraulic sheets (oil, gas, etc.) or plastic waste. This work focusses on oil pollution where the proposed measure is the degree of dirt for each zone: strong dirt, medium-strong dirt, medium dirt and weak dirt. Each zone is characterized by a list "List zone " which delimits its borders by the coordinates; they are composed by the list of the degrees of dirt "List Degreecell ", the list of cell positions of this degree "List P ositioncell " and they are attached to a zone by "P osition zone ". The proposed HMDCS-UV system is mainly based on two main steps: "monitoring" and "cleaning". Figure 4 illustrates the steps of the HMDCS-UV. Step 1: Monitoring. This step presents the monitoring actions performed by each UAV mr . These actions take place in two phases: path from its P os Start (x, y) to a final position P os End (x, y) using the explored atmosphere map Map atmosphere space (a discrete space of dimension 2, represented by a square grid G) and an algorithm proposed based on Cartesian coordinates "Planning towards the region" (Algorithm 1) with a rectilinear movement to reach its region (Fig. 6 ). Grid G is made up of identical square cells, containing free / occupied positions by UAV mr . These positions construct a dynamic graph where the arcs are presented by connection links (edges) between the neighboring positions. The link represents the distance between the positions. This distance is represented by an energy cost that the UAV mr has to consume in the displacement between the positions. The position has a maximum of eight links with neighboring positions j th . After each movement between these positions, the UAV mr saves the value of energy consumed in its List characteristics (M). -Phase 2.2-Navigation in the region: once the UAV mr has arrived in its region, it can plan its movement. It therefore uses its sensory sensors (a camera and an ultrasonic sensor) and the Map atmosphere space card so that it can fly or move on the grid based on the Algorithm 2 "Modified Boustrophedon". This algorithm is inspired by the "Boustrophedon path" algorithms defined in [2] where four cases are proposed: The first case (A) is executed when the number of maritime space columns (nc-1) is odd and the number of the column of the goal or start position (Y g ) is odd; The second case (B) is executed when the number of maritime space column (nc-1) is even and the number of the goal position column (Y g ) is even; The third case (C) is executed when the number of maritime space columns (nc-1) is even and the number of the goal position column (Y g ) is odd; And finally, the last case (D) is realized when the number of columns of the maritime space (nc-1) is odd and the number of the column of the goal position is even. For example, the UAV mr in the first case (A) crosses a new position (a goal position Y g ) which is the intersection between the second row and the first column until get to the final position which is the intersection between the last row and the last column, but the return path is to start by crossing a goal position which is the intersection between the last row and before the last column until arriving at end position which is the start position at the beginning. For each case, the UAV mr scans repeatedly until the cleaning is finalized and at the same time saves the energy consumed in List characteristics (M), as illustrated in Figs. 5 and 6. A supervision and detection solution is proposed so that the UAV mr can update its Map nautical space (lower level, in second square grid G 2D), where the UAV mr uses an unsupervised classification method specific to image processing to process its captured data using the "swipe" movement. This proposed method is inspired by k-means clustering [3, 4] . The operation of this proposed solution is illustrated by the following phases: a) Remote sensing; it designates the techniques allowing the acquisition of images to obtain remote information on an object, a surface or a phenomenon found on the surface of the earth, by means of a measuring instrument (for example, an airplane, a boat, a spacecraft, etc.) having no direct contact with the object studied [5] . When the UAV mr detects a sudden change in the light intensity of the color of the water with its ultrasonic sensor in spatial resolution, it captures the complete image of the polluted zone and identifies its matrix points of landmark. There are different contour methods like gPb (globalized probability of boundary) [6] , GraphCut road detection [7] , etc. The contour method used in [8] is integrated into HMDCS-UV as a pretreatment phase, as shown in Fig. 7 . Then, the UAV mr sends positions A, B, C and D of this zone directly to its General crd so that swarms of USV cz from other regions avoid this zone in their movements (see ; remote sensing data is received as an image in the process of UAV mr . This image is made up of many squares called pixels, as shown in the example of an oil slick in Fig. 8b . Image segmentation can be performed by several color space methods [9] . However, the defined method is based on the use of the RGB (red, green, and blue) color space. In this method, the pixel (a bright spot) is calculated by averaging the RGB color encodings. Each pixel in an image has a radiometric value between 0 and 255. c) Classification of data using the k-means algorithm; clustering is a process by which discrete objects with similar characteristics can be assigned to groups. In unsupervised classification, clustering methods aim at partitioning a set X = {x 1 , x 2 , ..., x n } of n objects described by p attributes into k classes also called clusters. The basic idea is that each object must be closer in terms of similarity to objects in the group to which it belongs than any object in another group. One of the most commonly used unsupervised classification algorithms is the k-means algorithm [3, 10] . d) Cluster validity measures; many criteria have been developed to determine the validity of clusters [9, 10] such as Dunn's index, Davies-Bouldin, F-ratio (WB), etc., all with a common goal to find the cluster that gives well separated compact clusters. Since the k-means method aims at minimizing the sum of the squares of the distances of all points from the center of their cluster, this should result in compact centers and thus compact clusters. To this end, the "F-ratio index validity method" can be applied. e) K-means clustering; Algorithm 3 shows the K-means clustering applied based on the intra-cluster (1). Where, c i : is the center of the cluster i; n i is the number of data points (pixels) in the cluster c i ; x ij is the j th data point of the cluster c i ; k is the number of clusters and d is a Euclidean distance between x ij and c i . The initialization of the k-means algorithm is based on the specified number of k clusters. These clusters contain the pixels of the segmented image. Then, the algorithm starts with an initial set of centers of gravity (or centroids) of clusters {c 1 , c 2 , ..., c k }, chosen at random (an iterative process). In each iteration, each pixel of the image is assigned to the center of gravity of its nearest cluster. Then, the centroids of the cluster are recalculated. The center of gravity c of each cluster is calculated as the average of all pixels belonging to that cluster: The steps of the process are repeated until the centroids no longer move. The proposed algorithm allows to produce segmented images for 2 clusters up to Kmax, where Kmax is an upper limit of the number of clusters, and then to calculate the validity measure to determine which cluster is the best cluster and, consequently, what is the optimal value of K. This work aims to produce 2 clusters and then to determine the dirty zone by the best cluster. This best cluster is not detected by the validity measure in a nautical image but is found by a proposed dirt level. This dirt level is calculated by the sum of the dirt levels found in the cluster. Thus, the cluster that has a maximum rate is defined as a dirty zone. Then, the UAV mr compares the List Degree cell of this detected cluster with a predefined "Threshold" to eliminate the non-dirty degrees. After the comparison, it saves the result in the List threshold d egreeZ list and sends it to General crd along with the explored and modified nautical chart. Step 2: Cleaning. This step illustrates the cleaning actions in three phases: A) Phase 1-Cleaning configuration: in this phase, the cleaning process is carried out as follows: The General crd analyzes the data received from each UAV mr . Then, the human operator or central unit officer prepares a report (PV) from the List threshold degreeZ to determine the position and zone of the dirty zone. Then, the central unit officer sends this report to the TGCG base via the SAM agent , and requests the SAM agent to send a nautical vehicle (V ehicle nt ) with agents to determine the cause of the pollution, check the climate, take samples and limit the pollution flow. Thus, the TGCG base sends the report to the local authorities (in this study, it focused on the city of Oran in Algeria), who represents the President of the local committee, and to the Regional Operational Centre for Surveillance and Rescue (ROCSR) of Oran. Then, the TGCG base asked the regional operational centre to send a nautical vehicle (V ehicle nt ) with an evaluation team to collaborate with the SAM agent . The assessment team and the SAM agent found in the V ehicle nt record the information necessary for this pollution. Then the SAM agent and the assessment team contact the SAM agent from the SAM department and the TGCG base respectively, by VHF to send this recorded information in formation. This SAM agent prepares an initial report based on the information received, and sends it to the central unit officer. The latter sends this IR1 to the TGCG base to prepare a second report also based on the information received from the evaluation team. This second report defines the position, zone, nature, cause of the dirty zone and the climatic report. The TGCG base sends the second report to the local authorities, the regional operational centre and the central unit. If the climatic conditions are not favorable, it does nothing and sends this information to the local authorities. Otherwise, the central unit officer launches a cleaning plan. This plan begins when the central unit officer determines the number of USV cz via the General crd for each dirty zone according to the List threshold degreeZ . This action is described by Algorithm 4 for a single dirty zone in a region. Then, the General crd prepares a list of end positions (List P osE (x, y)), illustrated by the dark gray color in Fig. 9 , so that the swarm does not go beyond the zone borders, and a list of border positions (List P osB (x, y)) in light gray color based on the positions received by the UAV mr , so that the USV cz is located near the zone cells using the regular cell decomposition. In this phase, a proposed trajectory planning solution is applied by the USV cz swarms, namely the Proposed-Cartesian Coordinate Algorithm (P-CCA). The P-CCA algorithm guides the USV cz swarm to plan its trajectory to its dirty zone. This algorithm is already defined in [1] with adaptations and extensions for the HMDCS-UV, by adding the P os End (x, y) defined by (x e , y e ) which represents the first position found in the ListP osB(x, y), and by replacing the pair of abscissas (i 1 , i 2 ) in the dirty zone by the pair (x e , y e ) (Algorithm 5). When the General crd chooses the swarm of USV cz , it sends the list of its List I D (I d USV cz ) identifiers to its Sup mr . Then, the General crd asks the SAM agent to prepare the USV cz forming the swarm in the base of life to check their hardware components, to prepare a Leader cz for each swarm with a high energy capacity compared to the other USV cz and to drop them on the water by a crane. Then it transmits a set of parameters to the selected swarm, as shown in the sequence diagram in Fig. 10 ; and it sends only the start position P os Start (x, y) and end position P os End (x, y) to the Leader cz . These fields are empty for the other USV cz of the same swarm. Thus, the General crd launches via WiFi the USV cz swarm to move into the dirty zone and clean it. The nautical zone is already discretized in a square grid (G). Each cell of the G can be dirty or clean. The USV cz moves on G and perceives a detection zone R s nearby ( Fig. 8 [1] ), and can clean the dirty cells. For this purpose, two solutions are proposed and applied by the USV cz to move and clean the dirty zone. -Solution 1: the first solution is an improvement of the proposition (Algorithm 2) defined in [1] . The novelty of this solution is that the USV cz are located around the dirty zone with a step between each USV cz . This solution eliminates collisions between the USV cz and reduces the cleaning time. Then, a final position adjustment function is executed by the USV cz to reach one of the positions found in the List P osB (x, y) list for each solution. This function is executed on the list of USV cz (List USV ) in two parts with two leaders, primary and secondary. The primary leader of the first part who arrived first at P os End (x, y), it starts searching for two successive positions found in List P osB (x, y) with an upward move and modifies its last P os End (x, y) found in P arameters start−up(C) , then it sends its old position to its neighbor, and the latter also sends its third position before the last one to its neighbor, and so on (Fig. 11a) . Then the secondary leader who is the last USV cz of the second part performs the same instructions with its followers but with a downward movement using the step concept. In the end, the primary leader is considered a leader for the whole swarm because it has consumed less energy than the secondary leader. Then, the General crd launches Algorithm 2 "Cleaning operation" on the USV cz . Each USV cz follows the algorithm so that it can avoid clean cells and select dirty cells to clean them. -Solution 2: this solution is based on the method presented in [11] , where the authors proposed to place USV cz around the dirty zone in a circular pattern and each USV cleans a slice from its starting point of Then, the dirty zone is divided into two parts. The first part is cleaned by the USV cz at the top and the second by the USV cz at the bottom of the zone, as shown in Fig. 11b . Next, the General crd runs Algorithm 6 on the USV cz . Each USV cz follows Algorithm 6 to avoid clean cells and to select dirty cells for cleaning. C) Phase 3-Cleaning termination: this phase consists of identifying the termination of the cleaning process [1] . When the USV cz finishes its cleaning task, it informs its Leader cz of the end of its mission by sending a message with the List characteristics (C) via WiFi. Upon receipt of the message, the Leader cz informs the USV cz to return to the base of life. When the USV cz arrives at the life base, it informs its Leader cz by sending a message. Then, this Leader cz sends the List characteristics (C) of the USV cz to its Sup mr for registration. Upon receipt of the message, the Sup mr adds the information on the displacement energy consumption between the zone and the base of life to the List characteristics (C) of the USV cz , and sends it to the General crd for registration. When the cleaning operation is completed by all the USV cz , the General crd prepares a initial final report containing all the information received from the Sup mr and the SAM agent compared to that of the USV cz and other equipment used in this operation, and then sends it via WiFi to the TGCG base via the central unit. The TGCG prepares a second final report based on the second report and initial final report and the information received by the evaluation team, then sends it to the local authorities, the regional operational centre of the city of Oran and the central unit for registration. This section presents an example to simulate the operation of the HMDCS-UV. The evaluation of this simulation is based on the following measurements: displacement energy consumption plus monitoring energy consumption of UAV mr in the region, total UAV mr energy consumption, displacement energy consumption plus cleaning of the USV cz swarm in the dirty zone, total energy consumption of the USV cz swarm, total system energy consumption, speed of the USV cz swarm, cleaning rate and efficiency of the USV cz swarm. Furthermore, it is compared with a second modified proposal to study the behavior of the HMDCS-UV proposal and to analyze the results obtained from the simulation. The first proposal (P1) includes the first improved cleaning solution. However, the second proposal (P2) includes the second modified cleaning solution. A series of simulations were carried out using different parameters. Before starting the experiments, a real environment of the Gulf of Arzew (of the city of Oran, Algeria) with the positions of the most frequent hydraulic pollutions is presented, in addition to a virtual environment to apply the HMDCS-UV proposal. The real environment of the Gulf of Arzew includes the port of Arzew, Béthioua and the offshore installations of Mers El Hadjadj (Fig. 12) . The Gulf of Arzew is located on average on the Greenwich meridian and the latitude 36 • N and extends from Cape Ivi (36 • 37 N -000 • 54 W ) to Cape Carbon (35 • 54 N -000 • 20 W ). These two capes form the boundaries of the Gulf of Arzew 3,4 . The port complex extends along the maritime fringe of the western part of Arzew Bay for about 22 km in latitude and 22 km in longitude (Fig. 12) . The Arzew basin may have sources of marine pollution such as oil leaks from a ship loading station pipe or the underwater pipeline of a liquid hydrocarbon; uncontrolled or accidental spills after loading a ship with hydraulic products, for example. The most frequent marine pollution in this basin is (Fig. 12 ) 6 : "SMP 1, Single Point Mooring 1" Buoy Station, for loading crude oil located in the harbor; "Musoir" pier end, it consists of three secondary points S1, S2 and S3 for loading and unloading crude oil, fuel oil and bitumen; "Post P03", for loading crude oil and fuel oil; "Oued Tassmanit", the presence of a black oil slick and H / C (hydrocarbon) traces along this Oued and the main axis towards the sea. This pollution is the result of leaks from industrial water discharged from one of the petroleum gas liquefaction and petroleum refining complexes. An example of a virtual environment is proposed, based on the maritime space of the Gulf of Arzew and the most frequent marine pollution presented. The virtual environment (maritime space) is represented by a central unit and a region "Region 1" (Fig. 13) . The central unit is composed of a base of life, and the region constitutes two levels (maps) "atmosphere" and "nautical". The atmospheric level includes a UAV mr . The nautical level includes a dirty zone "Z11" treated by a classification of an aerial natural image proposed below. This dirty zone is illustrated by a real crude oil slick in the vicinity of the SMP1 station in an exercise of the Tel-Bahr plan, proposed on 14 May 2019 at 06:00 local time when an oil tanker "ALPHA" loaded with crude oil collided with a vessel of type "RO / RO" during its exist maneuver of the Gulf of Arzew 7 . Before defining the discretization step, the "K-means clustering (X, k)" algorithm is applied to a reduced natural image of an oil spill captured in 2010. This slick of nearly 800 million liters of oil is spilled in the Gulf of Mexico 8 (see Fig. 14) . After reading the image, an RGB color space method is applied. A matrix is constructed based on this method which includes the coding of three colors (red, green and blue) forming the pixel, the Cartesian coordinate (x, y) of each pixel and the number of its cluster. This cluster number is filled in after performing K-means clustering, as shown in the example shown in Fig. 15a . After the RGB matrix, the K-means algorithm is executed to classify the reduced image, whose number of clusters is set to 2, and then the dirt rate of each constructed cluster is calculated. The cluster with the maximum rate represents the polluted zone, as shown in Fig. 15b and c of the previous example. Next, the proposed environment is randomly defined and adds the data of this classification. Then, a grid is obtained where the grey cells represent the dirty zone and the white cells represent the borders of this zone. Figure 16 shows the discretization of the previous example. The application was produced with the open source Java language and the simulations were executed on a virtual machine (VM) in a heterogeneous Cloud, built with the middleware VMWare vCloud Suite 6, with a VM created Fig. 13 Virtual environment (presentation of 3D level maps) from 16G RAM and two Xeon (R) CPU E5-2620 v2 processors (@2.10GHz, @2.09GHz) running under the Windows 7 operating system. The proposed HMDCS-UV was implemented for a maritime region that includes a surveillance vehicle (UAV mr ), a swarm of cleaning vehicles (USV cz ) and a dirty zone. This zone does not present any solid obstacles that could prevent the vehicles from navigating and planning their trajectories. Based on a 78 × 48 cell metric map (Grid), the UAV mr plans its trajectory from the base of life to its region according to the proposed Algorithm 1 "Planning toward region". When the UAV mr arrives in its region, it begins to plan its trajectory and collect nautical level data using the "Modified Boustrophedon" algorithm and the unsupervised natural image classification method. When it detects a dirty zone with the contour detection method, it processes it with the k-means clustering method. The captured data is measured against a water color metric. These colors are distributed over four intervals: ]0, 25], ]25, 50], ]50, 75] and ]75, 100] in which they contain white (light dirt), light brown (medium dirt), dark brown (medium-strong dirt) and black (strong dirt) cells. The UAV mr classifies this data (degrees) of the cells by comparing the predefined threshold value (equal to 26% of the degree of dirt) with the Degree cel of each cell. The number of USV cz containing the swarm used in the simulations varies between 34, 47, 60, 73 and 86 USV cz according to the enhanced cleaning algorithm (first proposal (P1)), and is equal to 136 USV cz according to the modified cleaning algorithm (second proposal (P2)). The ranked images for each simulation are shown in Fig. 17 . Each swarm of USV cz plans its movement along the optimal trajectory (using the Proposed-Cartesian Coordinate Algorithm (P-CCA)) to reach its dirty zone and clean it. The P-CCA was executed by the Leader cz of each USV cz swarm where each Leader cz of the classified zone starts from a starting position equal to (0, 0), and a final position equal to the first position in List P osB (x, y) of each dirty zone. When the Leader cz finds a goal position, it shares it with it followers. It is important to note that the trajectory of the UAV mr or of USV cz swarm obtained from the starting position (base of life) to the final position (the region or dirty zone) in the second proposition is the same as the one generated in the first proposition. The cost value between grid cells is randomly generated between 5 and 10. The energy required by the USV cz and UAV mr to turn left / right from its current position to another position is 0.2%, 0.01%, and 0.05%, 0.001% respectively to continue directly. When the USV cz swarm arrives in its zone, it starts cleaning the dirty cells. The energy required to monitor and capture data from a cell is 0.0001% for the second proposal using P-CCA (P2 PCCA). The proposed energy required to clean a black cell is 0.9%, for a medium-strength dirty cell is 0.5%, for a medium-strength dirty cell is 0.2%, and for a clear cell is 0.07%. And finally, the time required to clean a black, dark brown, light brown, light cell is 25, 15, 10, and 3 min respectively. A formula was proposed to calculate the total system energy consumption (TEC) for each proposal using P-CCA. The formulas for the swarm speed metric of USV cz (Speed swarm), the cleaning rate (Cleaning rate) and the efficiency of USV cz swarm (Swarm eff iciency) were calculated as follows: T otal time = T raveltime toZ + T raveltime inZ + Cleaningtime inZ T raveltime toZ = (T otalN b traversedcells toZ × Swarm size) T raveltime inZ = (NbCells traversed/cleaned × T raversetime cell) This section presents the results of monitoring simulations in the region based on the unsupervised classification method for monitoring and detection, planning the swarm's trajectory to its dirty zone using "Proposed-Cartesian Coordinate Algorithm (P-CCA)" and cleaning the dirty zone (classified image). The results of these simulations give the curves of UAV mr displacement and monitoring energy consumption in the region (DMEC UAV ), total UAV mr energy consumption (T EC UAV ), displacement energy consumption plus cleaning of the USV cz swarm (DCEC SU SV ) in the dirty zone, total energy consumption of the USV cz swarm (T EC SU SV ), total system energy consumption (T EC), speed of the USV cz swarm, cleaning rate (Cleaning rate) and the efficiency of the USV cz swarm (Swarm eff iciency) for both proposals. -Result of DMEC UAV : Fig. 18a shows the results of the DMEC of UAV mr in its region. The first proposal was compared to the second in terms of displacement and monitoring. The DMEC represents the energy consumption of the UAV mr when it plans its trajectory in its region, detects the classified dirty zone and monitors the swarms in this zone that have already planned their trajectories on the basis of a P-CCA. It can be noted that the UAV mr consumes less DMEC in the first proposal (P1 PCCA) by moving and monitoring the 34, 47 USV cz in each dirty zone compared to the second P2 PCCA by moving and monitoring the 136 USV cz swarms in each zone. In addition, the DMEC of P1 PCCA increases by monitoring one swarm of 60, 73 and 86 USV cz relative to P2 PCCA. As a result, the UAV mr consumes less energy to plan its trajectory in its region, detect the different proposed zones and monitor the swarm in P1 PCCA with a gain of + 35% compared to P2 PCCA. -Result of T EC UAV : Fig. 18b shows the result of the total energy consumption of UAV mr in the two proposals (P1 PCCA and P2 PCCA). The T EC combines the displacement energy consumption of the UAV mr from the base of life to its region and the DMEC. As a result, the UAV mr in the first proposal gained +0.9% of T EC compared to the second proposal for displacement, detection and monitoring. -Result of DCEC SU SV : Fig. 19a presents a simulation that evaluates the behavior of the USV cz swarm in the DCEC in the classified dirty zone for both proposals. It can be concluded that the displacement energy consumption increases much more than the cleaning energy consumption when the number of USV cz containing the swarm increases. In addition, the DCEC increases in P2 PCCA with the triple of 136 USV cz compared to P1 PCCA with 34, 47, 60, 73 and 86 USV cz . The HMDCS-UV proposal is generally better with 300 USV cz than the P2 PCCA of 680 USV cz in the DCEC with a gain of +52%. -Result of T EC SU SV : the two previous results were used to calculate the T EC SU SV based on the P-CCA for the two proposals. Figure 19b shows that the dark gray columns of the proposed dirty zone are greater in P2 PCCA than the white columns of T EC in P1 PCCA. As a result, the proposal P1 PCCA with 300 USV cz consumes less energy with a gain of +53% compared to P2 PCCA with 680 USV cz . -Result of T EC of the system: this consumption combines the total energy consumption of UAV mr (T EC UAV ) and the total energy consumption of USV cz swarm (T EC SU SV ). Figure 19c shows that P1 PCCA consumes less T EC with a UAV mr in the proposed region and a swarm of 34, 47 USV cz in each zone compared to P2 PCCA with a UAV mr in the region and 136 USV cz in each swarm. In contrast, the white columns of P1 PCCA with a UAV mr and a swarm of 60, 73 and 86 USV cz are greater than the black columns of P2 PCCA. It can be noted that the increase in T EC with a swarm of 60, 73 and 86 USV cz is due to the increased energy consumption of UAV mr in displacement and monotoring these swarms to find and clean unoccupied and dirty cells. Therefore, the proposal with a UAV mr and a swarm of 34, 47, 60, 73 and 86 USV cz is better than the second proposal with a UAV mr and a triple swarm of 136 USV cz for a gain of + 1.23%. -Result of Speed swarm: Fig. 20a shows the results of USV cz swarm speed in the cleaning step for both proposals. It can be seen that the swarm speed increases in P1 PCCA when the number of USV cz forming the swarm is equal to 34 and 47, then decreases when the swarm size is equal to 60 and 73 and remains almost stable at 86 USV cz and in P2 PCCA. It can be noted that the speed of the swarm decreases when searching for a dirty and unoccupied cell, and when cleaning. As a result, swarms of USV cz used in P1 PCCA clean dirty zones with an average increase in speed of +5.44% compared to P2 PCCA with 680 USV cz . -Result of Cleaning rate: Fig. 20b presents the results of the cleaning rate in three situations for the two proposals. The three situations are different depending on the swarm size used in each proposal. It is important to note that the black curve of P1 PCCA is above the gray curve of P2 PCCA in different situations, and the rate decreases as the swarm size increases. Moreover, the swarm lose a lot more energy in moving instead of in cleaning. Therefore, the proposal with 300 USV cz used has a high cleaning rate with an average percentage of +67% compared to P2 PCCA with 680 USV cz . -Result of Swarm eff iciency: Fig. 20c shows the results of the efficiency of USV cz swarm in cleaning for the two propositions. It can be noticed that the efficiency of swarm increases in P1 PCCA when the number of USV cz forming the swarm increases and remains almost stable in P2 PCCA. The curve of P1 PCCA is above the gray curve of P2 PCCA in all three situations, and the cleaning efficiency increases as the swarm size increases. Consequently, the USV cz swarms used in P1 PCCA cleans dirty zones with an average efficiency of +52% compared to P2 PCCA with 680 USV cz . This section positions our HMDCS-UV solution in relation to the related work cited above (section 2). Each mentioned study has its own characteristics / parameters which differentiate it from the others. Tables 3 and 4 present a comparative study of these studies and the strengths of the proposed HA-UVC. For example, the UAV mr of the HMDCS-UV system moves to its region using a proposed Cartesian Coordinate Planning Algorithm (P-CCA), locates and detects the positions of the dirty zone in its region on the basis of its sensors (a camera and an ultrasonic sensor) and with a proposed solution. This solution applies an unsupervised classification method of natural / satellite image processing (K-means clustering) such as the UAV developed in [22] which makes it possible to monitor an oil platform using a camera and a LiDAR sensor for navigation and collision avoidance, and a GPS to map the spill area. The aerial mobile robot of [19] , for example, is equipped with two cameras, a FLIR thermal imaging camera to locate and detect oil spills, and a digital camera to plan the route. In addition, the authors of [21] have introduced algorithms for the detection of oil spills using MIMO radar remote sensing integrated on a drone. Another sensor integrated into a commercial drone, called fluorosensor, was built in [28] for monitoring laser-induced fluorescence from the aquatic environment and also at night which is an obvious drawback compared to pulsed LiDAR systems with beach porting. Another sensor integrated into a commercial drone, called fluorosensor, was built in [28] for monitoring laserinduced fluorescence from the aquatic environment and also at night which is an obvious drawback compared to pulsed LiDAR systems with beach porting. Thus, the authors of (c) [29] proposed a new system architecture derived from the integration of a low cost laser array of pollutant detectors mounted on a UAV to identify the nature and amount of a release. The authors of [12] proposed an adaptive decision-making algorithm based on sensory information from autonomous vehicles that provides complete coverage of the search area for oil spill cleanup. On the other hand, the homogeneous drone swarm of the distributed system proposed in [15] makes it possible to monitor, locate and mark the perimeter of an oil spill and surround it, thus, avoid obstacles using the intensity of the signal (at a given frequency). In addition, the drone integrated in a system proposed in [31] makes it possible to monitor and detect in real time the temperature of the coronavirus (COVID- 19) from the thermal image based on the IoT. The system of [15] allows a large-scale evolution like the HMDCS-UV, [1] and [25] and not in other works. Subsequently, the UAV mr sends the nautical chart to its General crd after discretization. Based on the data received, the General crd assigns the explored nautical chart to the swarm USV cz to clean each dirty zone. For this purpose, two cleaning solutions are proposed in HMDCS-UV which were applied to the swarm USV cz where each swarm can simultaneously move and clean dirty cells from its dirty zone without specifying how to remove the dirt (hydraulic spill). Thus, each USV cz measures its quantity of energy by an energy threshold, then sends its information to the supervisor via its leader. On the other hand, the swarms of robots in the work [11] can place the barge with oil suction equipment and move it to another location to safely remove the oil. This barge is also used by the APC in the event of the presence of hydraulic pollution at the level of the port of Arzew, such as the [18] work which proposes an experimental system for the automatic towing operation of a dam in spill case using two ASVs and a ground station. Thus, a design of autonomous units (autonomous ships / drones) was developed in a research project EU-MOP [17] for the elimination of pollution of marine hydrocarbons, which are able to mitigate and eliminate the threat resulting from small and medium spills. Another design of a multi-robot system of autonomous aquatic vehicles was proposed in [16] for removing surface impurities, pumping oxygen into water, spraying chemicals, distributing food to appropriate places while measuring water quality. Thus, a method of controlling air pollution based on an air purifying drone system is presented in [26] to clean or reduce the amount of pollutants by spraying water and chemicals into the atmosphere. The works [17, 18] and [26] allow a feasibility in their systems and not in the HMDCS-UV. This paper proposed a concept study of HMDCS-UV, allowing the monitoring, detection and cleaning of polluted marine zones, based on the cooperation of several semiautonomous unmanned vehicles (a UAV mr and a swarm of USV cz ) and their coordination by a general coordinator. Thus, this cooperation allows the swarms to clean the polluted zones from the map explored by the drone using a detection method based on a K-means clustering algorithm. In addition, an effective cleaning method for USV swarms is proposed so that they can move around and clean polluted zones in maritime regions (oceans / sea). This method is better in terms of energy by comparing it to another modified method which is inspired by the method proposed in Zahugi, E. M. H. and al. (2013) [11] . The proposed HMDCS-UV uses a UAV mr for each maritime region and a swarm of USV cz to clean up dirty zones. HA-UVC: Hybrid approach for unmanned vehicles cooperation Optimal control for constrained coverage path planning Generalized K-Means Based Clustering for Temporal Data under Time Warp (Doctoral Dissertation) Satellite imagery land cover classification using kmeans clustering algorithm: computer vision for environmental information extraction The fusion of satellite and unmanned aerial vehicle (UAV) imagery for improving classification performance Contour detection for UAV-based cadastral mapping Efficient road detection and tracking for unmanned aerial vehicle Détection et identification des nappes d'hydrocarbures a partir des images satellitaires radar SAR Determination of number of clusters in kmeans clustering and application in colour segmentation Contributionsà la segmentation non supervisée d'images hyperspectrales: trois approches algébriques et géométriques (Doctoral dissertation) Oil spill cleaning up using swarm of robots Navigation of autonomous vehicles for oil spill cleaning in dynamic and uncertain environments Path planning of multiple unmanned marine vehicles for adaptive ocean sampling using elite group-based evolutionary algorithms Autonomous monitoring, analysis, and countering of air pollution using environmental drones Modelling oil-spill detection with swarm drones Aquatic multi-robot system for lake cleaning Design and operation of small to medium scale oil-spill cleaning units Preparing the automatic spill recovery by two unmanned boats towing a boom: Development with scale experiments Detection, localization and analysis of oil spills in water through wireless thermal imaging and spectrometer based intelligent system Controllaw for oil spill mitigation with an autonomous surface vehicle Performance analysis of detector algorithms using drone-based radar systems for oil spill detection Resheniye Drone, the Answer of Digital Oil Spill Recovery The river runs red -catastrophic oil accident in the Russian Arctic A comprehensive review of path planning algorithms for autonomous underwater vehicles Cooperative pollution source localization and cleanup with a bio-inspired swarm robot aggregation Application of Air Purifier Drone to Control Air Pollutants in Domestic and Industrial Areas Design and implementation of an uav-based platform for air pollution monitoring and source identification* Aquatic environment monitoring using a drone-based fluorosensor Application of miniaturized sensors to unmanned aerial systems, a new pathway for the survey of polluted areas: Preliminary results Using UAV-based systems to monitor air pollution in areas with poor accessibility Toward a novel design for coronavirus detection and diagnosis system using iot based drone technology Sea cleaning drones. Yankd Design Patroller tactical drone: a modular design to address the multiple needs of maritime surveillance This new solution is seen as an improvement and extension of the HA-UVC solution [1] which defines a solution for fault tolerance and scaling up. The results of the simulations carried out are very encouraging, allowing a very significant reduction in energy consumption in monitoring, movement and cleaning. Additionally, the results indicate the effectiveness of deploying heterogeneous unmanned vehicles in a surveillance, detection and cleaning task in a partially known maritime environment.On the basis of the good results obtained, in future work, the authors intend to study the influence of speed variations of unmanned vehicles in the control and cleaning phase, and also to implement an intelligent planning approach for swarms using other cooperative techniques avoiding obstacles such as line of sight, GPS intelligent buoy, fuzzy logic, etc. specific to predictable environments [24] , and the methods such as RRT, virtual bodies and artificial potential, etc. for unpredictable environments [24] . They thus plan to improve the proposed process by using other physical properties of hydrocarbons such as density, viscosity or pour point [8] . Finally, they will think of collaborating with the port of Arzew (of the city of Oran, Algeria) in the future to develop a real drone for maritime surveillance such as the "Patroller [33] " system, and a drone for cleaning hydraulic layers such as the design of the "bio-Cleaner [32] ". Funding The research has not been funded.Code Availability All code generated or used during the study are available from the corresponding author by request. Ethical Approval The authors declare that the rules of ethics (International, National) are respected in the conception and the realization of this work. Informed consent was obtained from all individual participants included in the study. Competing Interests No conflict of interest exists in the submission of this manuscript, and the manuscript is approved by all authors for publication. Hichem Benfriha is a computer science teacher in the Department of Technical Sciences, University of Mascara Mustapha Stambouli, Algeria. He is currently a Research Member of Laboratory of Computer Science of Oran. He is currently a PhD candidate in the Computer Science Department of Oran 1 University (Algeria). He received his Master of Science degree in 2012 from the same university. His research interests focus on CBR, data Mining, text mining, information extraction, information retrieval, natural language processing, machine learning and Multi-label classification areas.