Ecoferroquatics: Aquatic Plastic Clean Up With Ferrofluid and an Autonomous AI-Piloted Robot

View the attached video in second slide.

Video Transcript

“Every year, 11 million tonnes of plastic enter our oceans, and experts estimate that soon there will be more plastics than fish in the ocean. Current cleanup methods are basically blind: They catch everything, kill wildlife, clog quickly, and miss the smallest pollutants.”

“So instead of building bigger, I built smarter.”

“I designed, constructed and programmed an autonomous robot that uses AI to detect plastic and a custom ferrofluid - liquid filled with magnetic nanoparticles - to selectively attract and remove plastics.”

“Plastics are non-polar, so they interact with my cooking oil-based ferrofluid, allowing the system to target plastics without collecting everything in its path.”

“The AI was trained on my own dataset and guides the robot in real time.”

“In testing, it removed up to 50% of plastics from a small pool in just 3 minutes, with smaller particle sizes, it even outperformed a representative of current methods

“With more finetuning, this could see real-world deployment.

“This is just Mk-1 but it shows a new direction: smarter, selective, and scalable plastic cleanup.”

“Thank you for your time.”

Background Information

Plastic pollution contaminates water bodies worldwide. Without intervention, it will saturate our water sources and cause irreversible damage: killing our ecosystems, harming human health, and disrupting food chains.

11 million tonnes of plastic enter the ocean every year. (Fig.1)

Experts predict that, at our current pace, by 2060, there will be more plastic in the ocean than fish.

Ferrofluid, by definition, is magnetic nanoparticles suspended in a carrier fluid, making a fluid that can be moved around with magnets. It was invented by NASA in 1960 to operate as a zero-gravity fuel. Previous research has noted its capabilities as a possible tool in plastic pollution cleanup, but it has been limited to cleaning up small batches at a time.

Problem

Current methods of plastic recovery are nets, traps and filters. These have worked for years, but have glaring issues:

Clog quickly.

Requires lots of manhours.

Fail with smaller plastics.

Kill and catch everything in their path, including marine animals. (Fig. 2)

To sustainably address plastic pollution, we need a more effective solution.

My Solution

Plastic recovery needs to be easy, smart, efficient and selective. That's why I was inspired to build a new solution. From this project, I aimed for three goals.

Design a solution that picks up only plastic particles.

Make the solution automated and smart.

Test and evaluate performance, identify the strengths and limitations of this new method of plastic remediation.

I aim to improve efficiency, automate plastic recovery, and innovate a real-world solution

Building & Programming Mk-1

Ecoferroquatics Mk-1 is a robot which aims to accomplish three tasks:

Identify: Locate plastics in water using AI.(Fig3)

Mobilize: Move towards plastics in water.

Remediate: Use ferrofluid to pick up plastics in water.

This methodology was taken into account when designing the robot and ordering electronics. Ecoferroquatics Mk-1 was designed in Tinkercad and printed in carbon-fibre-infused PETG. A breadboard was used to wire the robot, which is equipped with a Raspberry Pi, a built-in power supply, an Arduino ESP32, and motors to accomplish its task. (Fig. 4)

How It Works

Identify: Ecoferroquatics Mk-1 is equipped with a camera positioned ahead of it, and an AI model analyzes the images captured by this camera within the robot to locate plastics and estimate their size and relative position to the boat.

Mobilize: The AI then targets the optimal plastic and works with an algorithm to make a path for the best way to reach that plastic. Custom computer algorithms convert these predictions and paths into 8-bit commands, which are then sent to the Arduino and subsequently to the motors. The robot then uses the paddles to steer and drive towards plastic pieces. (Fig. 5)

Remediate: Plastic pieces will come into contact with the front-facing magnetic drum. This drum is coated in ferrofluid, and as the plastic collides with it, it becomes wetted and soaked in magnetic ferrofluid. Once the plastic is sufficiently coated, it is magnetically attracted to the rotating drum and is rotated out of the water, then it is lifted into the scraper and is scraped off the drum; the plastic falls into a collection bin and is officially recovered. (Fig.6)

Evaluating Mk-1

The robot is evaluated in a controlled aquatic environment and tested on its ability to remediate different sizes and polymers. (Fig.7)

Key Results

⮞ Overall efficiency was 51.85% ± 18.34% (SD).

⮞ Collected 1.73 pieces every minute.

⮞ Best recovery: Macro-sized Foam PE = 96.67%.

⮞ Size was statistically relevant to efficiency (One-way ANOVA p = 0.005 < 0.05).

⮞ Hydrophobic, porous, simulated damage, and buoyant polymers(Like Foam PE) were best, while plastics like PVC and ABS did worse.

⮞ The comparison of the trendline and baseline test shows the robot performed 39.48% better with microplastics.

Observations

Ecoferroquatics Mk-1 could clear around half the plastics from a 0.785 metre diameter pool in 3 minutes. It collected an average of 1.73 pieces per minute.

The robot performed worse than the baseline for macro and meso sizes but managed to perform better with microplastics.

Decreasing the size harmed its cleanup capabilities; this is due to the AI model failing to detect and pilot towards these smaller plastics efficiently.

Lightweight Foam polymers were incredibly easy to pick up and were the most picked-up plastic polymers.

The plastics that did the worst were plastics that did not float. This is because the robot's method was only designed to pick up surface plastics(and 2 centimetres below the water surface). I still included these plastics that do not float because I aim to simulate real plastic pollution and not just pollution optimal for the robots' remediation methods.

The tests properly identify the limitations and strengths of Ecoferroquatics Mk-1.

Statistical analysis was conducted to evaluate performance consistency across size classes. Macroplastics reported a mean recovery efficiency of 71.67% ± 18.23 (SD), mesoplastics achieved 54.44% ± 21.87% (SD), and microplastics achieved 29.44% ± 15.84% (SD). The relatively low standard deviation in macro- and mesoplastic trials indicates stable and repeatable performance, while the substantially higher variability observed in microplastic trials represents inconsistent capture at smaller scales. (n=3)

The larger standard deviation reflects the strong size-dependent performance drop from macro to microplastics. These results quantitatively confirm that plastic size significantly influences remediation efficiency and reliability.

No ferrofluid leakage occurred, but it should still be a concern.

Ecoferroquatics Mk-1

The autonomous environmental remediation system's weight is 1.42 kilograms; it floats, moves itself, utilizes AI, does not leak ferrofluid and easily picks up plastics. The construction of the system can be considered a successful result. Comparing the system to current methods provides us with a comprehensive view of the limitations and strengths of this technology.

The total cost of this project was around $335, this is including the price for materials and prints that were never implemented. Proper cost-saving measures likely lead the cost of the robot to be $250 per Mk-1.

Limitations

The greatest limitation is that not all plastic pollutants are solid, surface-level, buoyant, hydrophobic, or visible to the naked eye. The test environment fails to account for the turbulence of open waters and debris densities. Plastics in open water may be covered in biofilm and substances that could reduce the effects of the hydrophobic interaction it would have with ferrofluids. This creation would face limited deployment, as ocean adventuring is currently out of the scope. Marine life would also be a consideration. While preliminary ferrofluid research indicates that this robot would be able to pick up nanoplastics incredibly efficiently, the reliance on AI for piloting makes piloting towards nanoplastics in water almost impossible.

Sources Of Error

A plastic tub was used, so plastics actually faced attraction to the edges of the pool, making it difficult for the robot to target them. The physics of magnetic fields causes the ferrofluid film to be weaker and thicker in some parts, reducing the carrying efficiency. The dataset for the AI was rather small, reducing the effectiveness of the targeting system. The AI model faced a small dataset. All images were hand-annotated; mistakes could have been made.

Conclusion

In conclusion, by evaluating an AI-assisted ferrofluidic plastic remediation system under controlled conditions, this work identifies both the capabilities and physical constraints that define its practical effectiveness. This project hopes to inspire more scientific research into automating the cleanup of aquatic environments.

Future Work

Real-world deployment would require adjustments, such as improved waterproofing and the inclusion of a black box to track failures and monitor progress, as well as strict ferrofluid containment protocols.

Deployment is likely for: Wastewater clarifiers, wetland restoration projects, and industrial cooling ponds.

The next steps are to demo and gain insight at university/research expos/events, and construct an Mk-2 that solves issues with the current method and utilize the insight:

Better power monitoring.

More economically friendly.

After that, reaching out to wastewater plants as a first testing environment will be the best next move.

Thanks to Mrs. Matthews, Mr. Green and Mr. Hayward for all the support and advice

Thanks to my family for listening to all the tiring practice presentations.

I couldn’t have done this without their support.

Ahmad, S., & El-Kadi, S. (2023). Review on the application of ferrofluids in water treatment and

environmental remediation. MSA Engineering Journal.

https://msaeng.journals.ekb.eg/article_291921_6243c0e60054f2220e61704c327b78a1.pdf

Alshraiedeh, H., Al-Jarrah, R., & Al-Omary, A. (2024). Autonomous aquatic robot for plastic waste

collection using deep learning and computer vision. Results in Engineering, 21, 101734.

https://doi.org/10.1016/j.rineng.2024.101734

Haldorai, A. (2024). An improved single short detection method for smart vision-based water garbage

cleaning robot. Cognitive Robotics, 4, 100039. https://doi.org/10.1016/j.cor.2023.100039

Kowalczyk, M., & Nowak, P. (2023). Experimental investigation of magnetic separation for water

pollutants. Warsaw University of Technology Repository.

https://repo.pw.edu.pl/info/article/WUT532c64965a314696bdc391c3e92aa71e/

Li, X., Wang, Y., & Zhang, Q. (2023). Ferrofluid-assisted removal of microplastics from aquatic

environments: Efficiency and mechanism. Chemical Engineering Journal, 475, 146123.

https://doi.org/10.1016/j.cej.2023.146123

Random Nerd Tutorials. (n.d.). ESP32, ESP8266, and Raspberry Pi tutorials and projects.

https://randomnerdtutorials.com/

The Ocean Cleanup. (n.d.). What types of plastic do you find in the middle of the ocean?

https://theoceancleanup.com/faq/what-types-of-plastic-do-you-find-in-the-middle-of-the-ocean/

United Nations Environment Programme. (2022, February 16). World leaders set sights on plastic

Pollution. https://www.unep.org/news-and-stories/story/world-leaders-set-sights-plastic-pollution

Wang, L., & Liu, J. (2023). Innovations in autonomous aquatic debris recovery systems. Marine Pollution

Bulletin, 192, 115042. https://doi.org/10.1016/j.marpolbul.2023.115042

Cairns, Stuart & Meza-Rojas, Diana & Holliman, Peter & Robertson, Iain. (2024). Interactions Between

Biochar and Nano(Micro)Plastics in the Remediation of Aqueous Media. International Journal of

Environmental Research. 18. 1-22. 10.1007/s41742-024-00635-0.

Florence N.F. Parker-Jurd, Natalie S. Smith, Liam Gibson, Sohvi Nuojua, Richard C. Thompson,

Evaluating the performance of the ‘Seabin’ – A fixed point mechanical litter removal device for sheltered waters, Marine Pollution Bulletin, Volume 184, 2022, 114199, ISSN 0025-326X, https://doi.org/10.1016/j.marpolbul.2022.114199.(https://www.sciencedirect.com/science/article/pii/S0025326X22008815)