PRE2022 3 Group9

Implementation and Simulation of Debris Recognition for Autonomous Drum Oil Skimmers

---

Group members

Name	Student Number	Email
Eryk Gruszecki	1731483	e.s.gruszecki@student.tue.nl
Mathilda Fogato	1656376	m.fogato@student.tue.nl
Oyku Sanlibayrak	1654519	o.a.sanlibayrak@student.tue.nl
Maud van Bokhoven	1664387	m.m.v.bokhoven@student.tue.nl
Siiri Jokiranta	1614207	s.h.jokiranta@student.tue.nl
Nika Tersteeg	1750828	n.n.q.y.tersteeg@student.tue.nl

Introduction

Ocean and other major body of water oil spills have a terrible impact on the ecosystem and people's health. With consequences ranging from physical pain to changes in behavior and reproduction, the hazardous compounds present in crude oil can disrupt marine ecosystems both temporarily and permanently. Fish, marine mammals, and birds who come into touch with the oil spill or consume polluted prey may experience particularly severe effects. Oil spills can harm human health in addition to having a harmful effect on wildlife. For individuals working to clean up oil spills, exposure to the hazardous chemicals found in oil can lead to respiratory troubles, skin irritation, and other health problems.For individuals working to clean up oil spills, exposure to the hazardous chemicals found in oil can lead to respiratory troubles, skin irritation, and other health problems. Economic losses may also occur in coastal areas as a result of the fishing and tourism industries. As a result, it is crucial that action be done to stop and lessen the harm caused by oil spills. The most concerning oil spills are primarily maritime ones. Since oil is a major source of energy, it is utilized frequently and all over the world, which makes spills like these more frequent. The majority of oil that is spilt remains on the surface and spreads out quite rapidly. Although conventional methods of oil spill cleanup, such the use of booms and skimmers, have been successful in removing oil from the ocean, they are frequently expensive, time-consuming, and heavily labor-intensive. One of the biggest problems oil skimmers face is clogging by floating debris, which reduces their efficiency and efficacy in oil spill cleaning efforts. As a result, the goal of this study paper is to delve more into a potential remedy for the problem of debris clogging oil skimmers, which involves fitting the skimmers with cameras and utilizing neural networks to sort trash photos in the ocean.

Approach / methodology

The research will focus on the stimulation of the oil spills in oceans and the performance of oil skimmers in cleaning the oil spills. An attempt to simulate an implementation of the suitable camera system to oil skimmers which can lead to a more effective way of cleaning oil spills in the ocean. This research will be conducted mainly by literature research, simulating the oil skimmer and contacting oil skimming companies and finally creating a simulation.

First of all, literature review will be needed to obtain relevant information on oil spills, the impact on the environment, current methods and existing oil skimmers. This information is needed to provide a clear overview and possibly new insights to design the autonomous oil skimmer.

Second, by contacting oil skimming companies another view will be obtained about the effectiveness of their current methods and machines. This will help to understand the perspective of the people in the industry, which will give a deeper understanding on the problems that oil skimmers face.

Third, we have chosen to use an image recognition system for plastics in oceans. By creating a database of images the robot will learn and optimize to minimize false positives and false negatives by learning by recognition plastics in the ocean and the distinction between plastics covered in oil, big chunks of plastics etc.

Fourth, the performance metrics play a big role in designing the robot to evaluate the performance and the effectiveness of the autonomous oil skimmer and compare the measurements with a "regular" oil skimmer. Performance metrics such as speed of cleaning, price/ production costs, lifespan, sustainable battery and the percentage of oil recovery (not sure about this).

Fifth and last, taking all these four aspects together in creating a simulation of the autonomous oil skimmer to obtain an insight how to robot eventually will work in a realistic environment. The simulation will be conducted using Unity in order to vary different conditions and aspects to test and optimize the robot.

Limitations:

- it is hard to design a large database for machine learning

- it is difficult to take every floating object into account. Think about the distinction between a dead fish and a floating piece of wood.

- what are we going to do when we do find a piece of plastic? avoid the plastic or pick it up?

- how are for example vessels going to recognize the robot in the sea or see the robot on a map? Is everyone be able to track the robot?

- How are we going to prove that this is more effective and safe than a regular skimmer, since it is autonomous?

- only using a simulation to prove the effectiveness and quality is weak. further research is needed for confirmation

Reliability and validity:

- internal validity: no influences by other factors

- external validity: generalizable/ results can be applied to other situations

- internal reliability: consistency of results across items within a test

- external reliability; the extent to which a measure varies from one user to another

Analysis technique:

Analyzing in Unity. How are we going to analyze the simulation? What is important to look at? What is our goal?

Reason for investigation -> problem statement

Problem Statement

Oceanic oil spills can harm marine life and their ecosystems, among other serious environmental effects. The research conducted by R.Manivel and R. Sivakumar in the publication "Boat type oil recovery skimmer" can be a useful source to comprehend in order to respond to the question of why focus on the oil spills. According to the publication, oil leakage is becoming a bigger issue due to an increase in ship mishaps and oil transportation, which endangers people's health. This is why it's so important to act and protect the environment as soon as possible.

Modern methods for cleaning up oil spills, such as ordinary drum oil skimmers, are efficient in removing oil from the water's surface. Oil and water are separated by a spinning, floating drum in the water. When the oil sticks to the drums, wiper blades have to remove the oil from the drums. Since they are portable, dependable, and effective, drum oil skimmers are said to be an effective oil skimmer. The oil skimmer, however, has its limitations.One of the main issues that stood out when examining the difficulties that oil skimmers encounter when collecting oil is the buildup of debris inside the machine, which wastes time and money. According to the School of Maritime Studies and Transport's research, this issue with plastic debris entering the machine and clogging the drum oil skimmers occurs quite frequently. In that situation, maintaining the oil skimmer will take a lot of time. Since oil disperses swiftly in water, it is critical that the oil be cleaned up as soon as a spill occurs.Also, as time goes on, cleaning up the spilled will become more and more difficult. To reduce the impact of the oil leak, it is crucial to respond and clean as soon as you can.

Minimizing maintenance and oil extraction speed are necessary in order to achieve this. Using a machine vision camera with image recognition capabilities that can identify debris and disregard it is one potential solution to this issue. This leads to extended working hours for the skimmer without the assistance of a second worker. A novel technology that has the ability to avoid trash while efficiently cleaning up oil spills is an automated drum oil skimmer outfitted with plastic identification software. Unfortunately, it is unknown how well this technique performs in practical situations when compared to a typical drum oil skimmer. This project's goal is to compare an autonomous drum oil skimmer with plastic detection software against a standard drum oil skimmer in order to determine which one performs better in terms of oil recovery, efficiency, and environmental impact.

Final Goal

The goal of this robotics simulation is to demonstrate the effectiveness of integrating object recognition and navigational path planning in oil skimmers for the purpose of improving the amount of oil collected and decreasing maintenance times. By creating a proof of concept, we aim to showcase the benefits of this technology and provide a basis for future development and implementation in the oil spill response industry. Through simulation, we will test the system's ability to accurately recognize oil and navigate a path that maximizes oil collection while minimizing equipment wear and tear. Ultimately, the goal is to provide a more efficient and effective method for oil spill cleanup, reducing the impact on the environment and minimizing the costs associated with cleanup efforts.

Functionalities

What is an oil spill?

Oil spills can be defined as the leak or spillage of petroleum, which is a mixture of hydrocarbons in a gaseous, liquid or solid form, into the environment. They are common in the sea and have been occurring for a long time, causing life-threatening scenarios. Some oil spills occur naturally due to the activity of earth movements that cause liquid or gas to escape on the surface. However, the most oil spills occurs due to human involvement such as accidents in which tankers release oil, especially marine oil spills.

Nowadays, shipping and environmental regulations are made to prevent oil spills, but still many oil spills happens. According to ITOPF, which is a company that maintains a database of oil spills from tank vessels, the average number of oil spills from 1970's till 2022 is reduces by almost 90%

Figure 1 shows that the number of medium and large tanker spills is reduced from an average of 78.8 to 5.7 oil spills per year. Furthermore, ITOPF stated that the main causes of oil spills were caused by collisions and groundings.

The size and location of an oil spill are important factors in the overall impact. Larger spills require more oil to clean up and the process is expensive. However, it happened that small oil spills incurred greater overall costs than larger spills, due to the lack of equipment and the difficulty of detecting such a small leak. Moreover, the cost of the spill depends on the distance from the coast, as the impact of the spill is greater when the oil hits the coast . Moreover, the type of oil and the concentration of chemicals play a role in the severity of the leak. To sum up, the impact of an oil spill depends on a lot of factors. The main factors are the size, location and the type of oil, but also the cause the of accidents, the type of oil skimmer, the currents/winds and the effectiveness of the clean-up. These factors are all important for the maintenance time/cost and play a role in the seriousness of the leak. .  

Another important aspect is the planning and the decisions that are needed to be made concerning making predictions in the distribution of oil when an oil spill occurs. Computer and mathematical models could provide insights to minimize the risks and impacts and in effectively clean up the oil with a searching algorithm.

What are oil skimmers?

Oil skimmers are devices that are used to clean up oil spills by collecting oil and separating it from water. The advantage of oil skimmers compared with other methods of cleaning up oil spills is that the oil that is collected can be reused. Other methods of cleaning up oil spills include oil booms, sorbents and burning oil. As the focus of the course is on robotics oil skimmers were chosen as the method of cleaning up oil.

The type of skimmer that is the most effective at collecting oil depends on the type of water body (sea, lake), the type of oil (heavy, light) and environmental conditions such as wind speed, the strength of surface waves and the presence of debris. There are three main types of oil skimmers: weir skimmers, oleophilic skimmers and suction skimmers. Weir skimmers float on the surface of water and use a dam at the oil-water interface to separate oil from water whereas suction skimmers use a vacuum to pump the oil into a storage tank. Weir skimmers and suction skimmers are only effective in calm waters and get easily clogged by debris.

Oleophilic skimmers are the most effective type of skimmer for collecting oil in the ocean. This type of skimmer uses an oleophilic material that absorbs oil to separate oil from water. Drum skimmers are a specific type of oleophilic skimmer where a rotating drum is coated with an oil-absorbing material. Some oleophilic skimmers can deal with small amounts of debris but drum skimmers are sensitive to being clogged .

Drum skimmers were chosen as the type of oil skimmer to focus on because they are commonly used for cleaning up oil spills in an ocean environment and they face the problem of getting clogged by debris.

Debris

Most oil skimmers are not effective at collecting oil when there is a lot of debris present because debris can clog the intake of the skimmer and prevent it from pumping oil. This can result in the skimmer breaking down or someone having to manually remove the debris from the skimmer which increases the amount of time it takes to clean up an oil spill. In addition to this, more people are needed to operate the skimmer as it has to be constantly observed to prevent it from getting clogged. In areas with a high amount of debris such as harbors, the debris must first be removed before oil skimmers can be used to collect oil which can further increase costs and the amount of time it takes to clean up the oil spill The main ways to deal with debris are to either manually remove the debris or to use an oil skimmer that collects the debris together with the oil.

Specifications of Oil Skimmers

Types of oil skimmers

There are several different varieties of oil skimmers, each with unique advantages and limitations. Beginning with belt skimmers, these machines use a rotating belt to collect oil from the water's surface and transport it to a collection container. They perform best when removing sizable amounts of oil from bodies of open water, such as lakes or harbors. Drum skimmers revolve a drum in order to gather oil from the water's surface. They are good in removing both light and heavy oils, and they operate well in both calm and choppy waters. The best place to use disc skimmers is in shallow water since they use a rotating disc to remove oil from a surface. Weir skimmers are best employed in calm environments like ponds since they collect oil off the water's surface.The drum skimmers will be the subject of this study, though. The cause behind that can be expressed in two key points: The ability of the drum oil skimmers to function in open ocean and sea is one of them. Focusing on a drum-type oil skimmer is an excellent idea because this research is concerned with debris in the open sea, which is a much more complex environment for the oil skimmer. Moreover, the drum oil skimmer is mobile. It can travel independently to get oil. This is an important factor to consider when selecting the oil skimmer that will be the subject of this study because our exploration of the oil skimmer needs to be mobile in order to detect debris in the ocean using the image recognition technology that will be used in the camera system and to ignore the debris that might clog the skimmer.

How long does it take to clean oil with a drum skimmer?

(Nika) As it is shown that the amount of medium, 7-700 tons, and large spills, >700 ton has been decreased, but over 80% of spills since 1970 are small oil spills, <700 ton . Since small oil spills occurs often, it is important to look further into these and questioning the impact of such a small oil spill compared to medium or large spills.

The article "Calculation of oil droplet size distribution in ocean spills" by I. Nissanka and P. Yapa reviews the impact of an single droplet of oil in the ocean based on model predictions

As it is said before, using models and simulations could make predictions in the possible risks and according to the I. Nissanka and Y. Yapa, the calculation of an oil Droplet Size Distriubtion (DSD) are essential in model predictions of ocean oil spills. The calculation of a droplet size can be split up into equilibrium models and droplet population dynamic models with both promising promising, but the equilibrium model is preferred since it can be used for field situations. As believed by I.Nissanka and Y.Yapa, modeling this DSD is complex and contains several processes but it is impossible to include all these processes in the model.

J. Montewka attempts to design a probabilistic and systematic model to estimate the costs of clean-up operations for the Gulf of Finland in "A probabilistic model estimating oil spill clean-up costs - A case study for the gulf of Finland" .

The author presents a model based on Bayesian Belief Networks that analyses the cleanup costs for the Gulf of Finland in risk framework. A cost model includes utility-, decision-, independent- and conditional variables. In addition, the oil type, spill size, season, wave height, evaporation, effect of booms, time for vessel to arrive, time to collect the oil, oil-combating efficiency, number of vessels, manual and machine costs.

In the first scenario, an oil spill of 5000 tons of medium oil the total clean-up costs: 12.1 M

Second scenario: oil spill of 30000 tons of heave crude oil: 95 M.

Doesn't include socioeconomic, environmental and waste management costs.

The model allows the user to select the location of an oil spill, its size, type of oil, time to collect the oil and the model calculates the total costs of clean-up operations.

In our research we can use this source to prove that the total costs to clean up an oil spill will decrease when the recovery time will decrease since the time to collect the oil is a factor in the model.

"Performance of rotating drum skimmer in oil spill" recovery by A. Hammoud and M. Khali investigates the effectiveness of the oil skimmer and finally predict the oil recovery rate of the device.

"According to the authors, the rate of the maximum oil recovery is dependent on drum rpm and oil viscosity. Oil increases with oil thickness for both low and high viscosity oils. Furthermore, oil recovery rate decreases with wave height for both low and high viscosity oil and increases with current velocity."

An experiment was conducted in which a rotating drum skimmer was tested in a water reservoir spilled with oil. For each test, the oil thickness, which varied between 10 to 40 mm, and the distance between the drum and the surface were measured.

In another similar experiment, the oil recovery rate and the oil recovery efficiency were measured.

The results show the effects between the rotation speed and the recovery rate in crude oil. The main factors to indicate the performance of the drum skimmer are the speed of the rotating drums and the drum diameter. This can be explained as the speed of the rotation increases, the oil recovery rate increases since the drum contact area with the oil increases. In addition, the oil recovery rate also increases when the distance between the oil and the surface decreases.

The same results were seen for diesel oil and SAE 10 w oil

Operational parameters for recovery rate

Speed:

In the article “Effect of operational parameters on the recovery rate of an oleophilic drum skimmer” written by V.Broje and A. Keller shows the effects of the different parameters on the recovery rate. Looking at the speed of the oil drum skimmer, the drum rotational speed was tested for 30, 40 and 65 rpm (rotations per minute). Varying the speed controlled the encounter rate of the oleophilic surface with the oil front. 30 rpm represents the minimal rotation speed and 65 rpm the maximal speed that could be achieved for average drum skimmers. A higher speed emulsifies the oil to a greater extent. It shows that the amount of entrained water will increase, especially in the case of thin oil slicks and/ or viscous oils. The amount of recovered oil can be increased by 50-100% using higher rotational speeds. 40 rpm seemed to be near the optimal rotational speed above which the drum starts to entrain significant amounts of water which results in less emulsification. If storage capacity is not limited, drums should be operated at their maximum speed.

In another experiment by Walaa Sabbar et al (2021) investigated in their experiment in which the temperature, viscosity, oil slick thickness and drum rotational. In addition, these parameters were tested for different type of oils such as light diesel, crude oil and heavy diesel oil. The results of the optimal speed for the rotations in shows that the recovery rate is proportional to the increase in the drum rotation speed and predicted in their experiment that 56 rpm  was the optimum rotation speed. Light diesel shows that this type of oil had the best recovery rate which increases with decreasing viscosity.

USE Assessment

When designing a new type of drum skimmer it is necessary to look at the USE assessment and take these different aspects into account.

User

Different users are involved such as oil spills companies, government agencies, oil and gas companies, environmental organization and research institutions.

Government companies

Oil spills have negative impacts on public health, drinking water, natural resources, ecology and the economy. As the oil industry is huge and unavoidable, large quantities of oil are used and transported, sometimes resulting in an oil spill, putting public health and the environment at risk. The U.S. Environmental Protection Agency's Oil spill Program plays an important role in the protection of oil spills that occur in and around the U.S. The organisation tries to prevent these oil spills and published a booklet containing information about oil spill which include potential effects. EPA proposed different techniques that may be used after an oil spill to minimize the impacts. The main response technique is the "mechanical containment or recovery". This includes the recovery equipment such as barriers, skimmers, booms and sorbent materials.

When there is a new design of an oil skimmer which shows to be more effective and efficiency, environmental organizations such as EPA could implement this as a standard practice for oil spill response.  This could lead to a better environment as EPA aim to minimize the negative impacts of oil spills.

Society

The impact on the society after the Gulf of Mexico oil spill in 2010.

An huge disaster happened in 2010 in the Gulf of Mexico, also known as the “BP oil spill” which is one of the largest oil spill in history with a total oil spill of 129000 m^3 and even 11 workers have died. After the explosion, different cleaning strategies were implemented in order to clean up to spill and minimize the effects. However, oil spill caused widespread environmental damage and social impacts for the economies and  cultures in the area. Nearly half of all Gulf Coast residents perceived damage to the environment and. To talk further about the socioeconomic impacts, nearly 223000 km ^2 of fishable waters were closed due to the oil spill. Many fishing communities had to close their fishing area and caused big economic losses. To be more specific, Floridians felt the economic effect on the tourism industry and Floridians on the tourism. The spill caused damage to the reputation of the area since there could not be fished and the area was associated with the oil spill.  The total economic impact is approximately 23 billion .

The spill had also impacts on the physical and mental health of the residents. After the exposing crude oil that includes chemicals that could cause cancer and brain damage and other disease. Some residents reported health issues related to their lungs, kidneys and heart functions in the long term effects. Not to forget is the impact on the mental health such as an increase in anxiety and depression, and even PTSD issues. There are still studies investigating in the long term effects since we don’t know the effects of the exposure with crude oil.

201.6x201.6px

Looking from the oil company perspective, BP was the main responsible for the leak and they faced criticism from the media, environmental groups and the public.

BP responded to the leak and started a clean up including oil skimmers and booms to remove all the oil. The total estimate costs were around 61.6 billion of which BP paid most of it which caused financial consequences for BP. Furthermore, the reputation of BP dropped significantly such as a decline in stock price, lost partnerships and faced a negative public image.

Environment

According to A. Clifton (2014), Petroleum is one of the most consumed raw material as an energy source and therefore transported in significantly large oil tankers. The annual estimated amount of petroleum that is spilled is between 10 million and 20 million ton and the petroleum spreads on the surface of the sea. When an oil spill occurs, there are great concerns to the environment on short term and long term. An oil spill harm ocean life by fouling or oiling, which means that the oil physically harms plants and animals such as coating bird’s wings due to oil and by oil toxicity, which means that the toxic particles cause health problems such as to the immune system and even death.

The impact of an oil spill has large effects and even after a few years, it has proven that there were still high concentrations found in the sea.

To mitigate the effects, new techniques are developed surrounding an oil spill clean-up and one of the most promising technique is the remediation process for reversing environmental impacts. Another way to reduce long-term negative environmental impacts is to clean up oil quickly and efficiently as soon as possible after an oil spill. If there is a new oil skimmer design that is more efficient and takes less time to clean up oil, the long-term impacts will be reduced.

Enterprise

While oil spills continue to represent a serious threat to our ecosystem, governments and organizations throughout the world are investing more in cutting-edge technologies to lessen the dangers associated with them. The oil skimmers in the market faces a lot of challenges while trying to perform their task. Especially one of the biggest problems that the oil skimmers struggle with, is the obstacles they face in the open sea environment, like debris. The classic oil skimmers don't know how to ignore these kind of obstacles, which make them much harder to navigate and manage. This problem makes the company to loose a lot of money and time. That's why, our brand-new oil skimmer, which is equipped with Sony XCG-H280CR machine vision camera, is in the leading-edge of this effort. Our skimmer is designed to work in the most extreme environmental conditions, ensuring that oil spills may be contained and cleaned up fast and effectively. It boasts powerful picture processing capability and a durable, waterproof shell. The Sony XCG-H280CR machine vision camera, which enables real-time detection and identification of aquatic waste, is the key component of our skimmer. With a high resolution of 1920 x 1200 pixels and a frame rate of up to 60 frames per second, the camera can take photos quickly and accurately, ensuring that the skimmer operates with the best speed and accuracy possible. Due to its powerful image processing capabilities, the camera can also function well in low light conditions, which can be typical during oil spill cleaning tasks. Our oil skimmer, which we believe to be a significant advancement in technology for cleaning up oil spills, offers organizations and companies looking to lessen the risks associated with oil spills a convincing solution.Whether in response to a spill emergency or as part of routine maintenance tasks, our skimmer is designed to deliver rapid, effective, and environmentally responsible solutions.

Navigational Search Methods

In literature, there are two search methods that have been implemented in autonomous water/ocean surface operating robots. Deep Reinforcement Learning (CARL-bot) and Coverage Path Planning (SMURF):

Deep reinforcement learning (DRL) is a machine learning technique that has shown great promise in robotics and other complex tasks. One DRL algorithm that has been applied to autonomous oil spill cleaning robots is called V-RACER (Value-based Reinforcement learning for Autonomous Cleaning and Environmental Restoration).

V-RACER is a DRL algorithm that combines deep learning and reinforcement learning to enable the robot to learn how to navigate and clean up an oil spill efficiently. The algorithm uses a neural network to map sensor inputs to actions, allowing the robot to learn how to navigate and clean up the spill in a simulated environment.

In the V-RACER algorithm, the robot is trained in a simulation environment that mimics the real-world conditions of an oil spill. The algorithm uses a reward system to encourage the robot to take actions that lead to the successful cleaning of the spill, such as moving towards areas with higher concentrations of oil and avoiding obstacles.

"CARL-bot is the Caltech Autonomous Reinforcement Learning robot, which is a palm-sized, underwater robot for testing our RL-based navigation in the real world. It will show us how our recent results in a simulated fluid flow hold up for navigating in the real world with all of the associated challenges, such as imperfect sensors and motors, all while training on-board and in real-time"

The algorithm also uses a technique called experience replay, which involves storing and replaying past experiences to help the robot learn from its mistakes and improve its performance over time. The experience replay technique enables the robot to learn from a diverse set of experiences and develop a robust and efficient cleaning strategy.

One advantage of using DRL algorithms such as V-RACER in autonomous oil spill cleaning robots is that they can adapt to different environments and conditions. The algorithm can learn how to navigate and clean up spills in a variety of conditions, such as rough seas, changing currents, and different types of oil.

On the other hand, coverage path planning is a search strategy commonly used in autonomous systems, including oil spill cleaning robots. The objective of coverage path planning is to ensure that the robot systematically covers the entire area of interest while minimizing the distance traveled and maximizing the cleaning efficiency.

In the context of oil spill cleaning, coverage path planning can be used to guide the robot to cover the entire area of the spill while minimizing the amount of oil left behind. The robot can use various sensors and imaging technologies to detect the extent and location of the spill, and then use coverage path planning algorithms to plan the most efficient cleaning path.

There are several advantages to using coverage path planning in autonomous oil spill cleaning robots. First, it ensures that the entire area of the spill is covered, reducing the risk of leaving oil behind. Second, it minimizes the distance traveled by the robot, reducing the amount of energy consumed and maximizing the battery life. Finally, it ensures that the cleaning process is carried out in a systematic and efficient manner, reducing the time and cost required to clean up the spill.

An existing fully autonomous water surface cleaning robot has been developed using a variation of the coverage path planning method SMURF by a research team in China. It is a trash collection robot that aims to efficiently clean the water surface that it is operating in.

“The environment perception of SMURF relies on the fusion of mmWave radar point clouds and RGB images. SMURF is controlled through our improved NMPC controller to sail along the generated path. When there are obstacles on the path that are detected during the task, a local path will be generated to bypass the obstacles. In addition, the SMURF detects floating trash autonomously and collects the trash in time to avoid trash moving. ”

There are various algorithms and techniques that can be used for coverage path planning in autonomous systems. One common technique is the grid-based algorithm, which divides the area of interest into a grid and assigns the robot to cover each cell in a predetermined order. Another technique is the Voronoi diagram-based algorithm, which divides the area of interest into polygons that are covered by the robot in a predetermined order.

In conclusion, the V-RACER algorithm is a powerful DRL technique that has shown great promise in the application of autonomous oil spill cleaning robots. By using a combination of deep learning and reinforcement learning, the algorithm enables the robot to learn how to navigate and clean up spills efficiently, even in challenging and changing conditions. The ability to adapt to different conditions and environments makes DRL algorithms such as V-RACER a valuable tool in the fight against oil spills and other environmental disasters. Additionally coverage path planning is an effective search strategy for autonomous oil spill cleaning robots, as it ensures that the entire area of the spill is covered in a systematic and efficient manner. By using sensors and imaging technologies to detect the extent and location of the spill and applying coverage path planning algorithms, the robot can maximize cleaning efficiency while minimizing energy consumption and time required to clean up the spill.

Coverage Path Planning

Firstly, coverage path planning is defined as:

"... the task of determining a path that passes over all points of an area or volume of interest while avoiding obstacles.

https://www.sciencedirect.com/science/article/abs/pii/S092188901300167X#:~:text=Coverage%20Path%20Planning%20(CPP)%20is,of%20interest%20while%20avoiding%20obstacles.

This type of navigational method is very common within autonomous household devices such as vacuum cleaners, lawn mowers and many more.

Coverage path planning is an essential technique for modern-day robots, which involves developing algorithms and strategies to enable robots to navigate through an environment and cover a specified area. Here are some steps to consider when implementing coverage path planning in modern-day robots:

1.    Define the problem: The first step in implementing coverage path planning in modern-day robots is to clearly define the problem you want to solve. You need to determine the area you want the robot to cover, the obstacles in the environment, and the resources available to the robot.

2.    Choose a suitable algorithm: Once you have defined the problem, the next step is to choose a suitable coverage path planning algorithm. There are various algorithms available for this purpose, such as potential fields, cellular decomposition, and Voronoi diagrams. You need to consider the specific requirements of your problem and choose the algorithm that best meets those requirements.

3.    Map the environment: Before the robot can start moving, you need to map the environment. This involves creating a digital map of the area the robot will be operating in. The map should include details such as the location of obstacles, the boundaries of the area to be covered, and any other relevant information.

4.    Plan the path: With the algorithm and map in place, you can now plan the path for the robot. The path should cover the entire area that needs to be covered while avoiding obstacles and using the least amount of resources possible.

5.    Implement the algorithm: Once the path has been planned, it's time to implement the algorithm. This involves programming the robot to follow the path and use the coverage path planning algorithm to avoid obstacles and maximize coverage.

6.    Test and refine: Finally, you need to test the robot and refine the algorithm as necessary. This may involve tweaking the path or adjusting the algorithm to improve performance. With each iteration, the robot should become more efficient and effective at covering the specified area.

Overall, implementing coverage path planning in modern-day robots requires careful planning, algorithm selection, and testing. However, with the right approach, this technique can be a powerful tool for maximizing robot efficiency and productivity in a wide range of applications.

However, CPP works well in static environments for example the inside of a home or the front lawn of house. In a dynamic environment, such as the ocean, environmental factors like wind and currents are huge challenges for CPP based robots. Therefore, the autonomous drum oil skimmer could only be deployed within a bounded area created by floating barriers.

"Therefore, the challenge of the CPP path optimality is to minimize the total travel time along the coverage path and reduce the turning cost."

https://ieeexplore.ieee.org/document/9523743

Hardware

Which camera to use?

The ocean environment presents unique challenges for industrial image recognition systems, as cameras must be able to withstand the harsh conditions of saltwater, strong currents, and changing light conditions.

The hardware used for image processing systems, namely cameras, are distinguished by two groups. The first group being industrial/machine vision (MV) cameras and the second group being network/IP (Internet Protocol) cameras.

Industrial image recognition requires high-resolution cameras capable of capturing clear and detailed images even in low light conditions. High-resolution cameras can capture images up to 4K or 8K, allowing for detailed analysis of underwater images. Remote sensing cameras can capture images from a distance, allowing for long-range monitoring of underwater environments. They are useful for monitoring oceanic currents and weather patterns

The goal of this study is to develop a camera system that can interact with a drum oil skimmer and record photos of debris while simultaneously identifying and ignoring plastic particles that could clog the skimmer. In this situation, using a machine vision camera with picture recognition skills is conceivable. There are numerous machine vision cameras available on the market. There are various considerations that must be made while choosing a camera. Resolution is a crucial feature to consider. To get clear images of the floating waste in the water, the camera's resolution must be high enough. The frame rate is another significant feature. The number of video frames that the camera records each second is known as the frame rate. The camera's frame rate must be high enough for the real-time image processing to function effectively. A decent camera system also needs to be able to process images. The camera should either have built-in image processing hardware or be compatible with external image processing units to enable accurate and effective trash detection. The lighting conditions should also be considered. The camera need should be able to take crisp pictures in both bright and dim lighting conditions. Environmental resilience and strength features are crucial, too. The camera ought to be able to withstand exposure to severe weather, including exposure to moisture, salt, and extreme heat.

For our oil skimmer, we decided on the Sony XCG-H280CR machine vision camera due to its exceptional performance and cutting-edge capabilities. Even in difficult climate conditions, the camera's high resolution of 1920 x 1200 pixels and quick frame rate of up to 60 frames per second enables it to record clear and accurate photographs of the ocean surface. Our skimmer can rapidly identify and remove oil from the water while leaving behind other items thanks to its excellent image processing capabilities, which include real-time debris identification and recognition. In addition, the Sony XCG-H280CR is a resilient camera designed to survive challenging environmental conditions, like those frequently present during oil spill cleanup efforts. While its industrial-grade components guarantee that it can withstand the demands of long-term usage, its waterproof and dustproof housing ensures that it operates reliably in wet and dusty circumstances. Overall, we've found that this camera's vision provides the perfect balance of performance, dependability, and longevity for our oil skimmer. We are certain that this camera will assist our skimmer to perform at its best, offering quick and effective oil spill cleanup services that contribute to the protection of our environment from oil spills.

How to implement this camera system to the oil skimmer?

We chose the Sony XCG-H280CR machine vision camera for the oil skimmer, and therefore the implementation process of the camera system to the CAD design of oil skimmer needs to be explained. First of all, the position of where to implement the camera system is very important since the clear visual image of the debris in ocean should be detected easily. From the picture of our CAD Design, the final decision of our implementation can be seen. After that, once the camera is mounted, you must connect it to the oil skimmer's control system so that it can handle real-time image processing. Installing specialist hardware and software, such as OpenCV, will allow the camera to connect with the control system. OpenCV can be used to process images in real time. It supports a wide range of image processing operations and may be used with C++, Python, and Java among other computer languages. Secondly, the camera's settings must then be adjusted to maximize its efficiency during oil skimming activities. In order to ensure that the camera takes clear and accurate pictures of the trash on the water's surface, this process involves adjusting the resolution, frame rate, and other settings. Finally, in order for the skimmer to react to camera input and modify its operation as necessary, the camera system must be integrated with the skimmer's control system. This can entail installing an automated system that modifies the skimmer's speed and position in response to information from the camera.

Simulation

Next to the research we conducted, we made a simulation of how a drum skimmer works with and without our proposed improvement. The main goal of these simulations is to be better able to explain how our variation on the skimmer would work. It can also give people who are interested a better idea of why it would be a good improvement, compared to just explaining our idea.

Besides from helping explain to potential users how our proposal would work, the simulation could also help potential users identify problems in the design that we did not see due to our limited knowledge about oil skimmers. The simulation can therefore help us include potential users in the design.

Next to helping potential users better understand how our system would work, the simulations are based on data, in order to show in which environments our system would be an improvement to the effectiveness of the skimmer.

Weekly Work

Logbook
Who?	Eryk	Nika	Matilda	Oyku	Sirii	Maud
Week 1 + hours	Created structure of the report	Thinking about subject and research question (4)	Researching about autonomous robots and AI-based navigation solutions (4)	Researching about ocean garbage collectors, and how they work		Researching the current state of the art of autonomous underwater garbage collectors (8)
Week 2 + hours	Problem Statement research	Research to different type of oil skimmers (6)	Researching image recognition techniques and implementations for sea plastic collection (6)	Research about the types of oil skimmers, and their advantages		Researching how oil skimmers work (6)
Week 3 + hours	Comparison of autonomous drum oil skimmer and non-autonomous drum oil skimmer	Methodology + start problem statement (6)	Working on Unity simulation (6)	Getting in contact with oil skimmer companies and continue working on the report		Working on netlogo-model (2)
Week 4 + hours	Naviagtional method for ocean surface robots	Analyzing 3 papers that focuses on the oil recovery rate of drum skimmer / the number of oil spills/ the effects of a droplet oil spill (8)	Working on creation of a realistic environment for the Unity simulation (6 )	Research about the camera system that is going to be implemented on the drum oil skimmer		Writing about why we chose to make a simulation (1)
Week 5 + hours				Finishing the "enterprise" part in the USE part. Work on the camera system, which one to use,why, how are we going to implement it
Week 6 + hours				Research about How to implement this camera system to the oil skimmer. Proofread some part of the report to make it better
Week 7 + hours
Week 8 + hours

References