Plastic waste is an immense problem in India’s already shrinking water bodies and wetlands. Over time, it degrades and leaches harmful chemicals like phthalates, heavy metals, and chlorides , worsening water contamination. During my survey of rivers and lakes in Delhi/NCR other than sewage waste, I found plastic waste to be a major problem.
To address this, I have designed "Ecodrifter"—a low-cost, portable robot that removes floating plastic waste from water bodies. Built with readily available materials like PVC pipes, sunboard sheets, a rubber conveyor belt, and plastic mesh, it runs on two DC motors and a rechargeable battery. Controlled via a flysky transmitter, the device costs ₹10,000. After analyzing the first version, I developed an improved second prototype for better efficiency. While the version 1 was built on a conveyor systems mechanism , the version 2 is built on an outrigger boat mechanism. The version 2 is more efficient to collect a larger volume of plastic including the one accumulated at banks .
The current device is created on a trimaran /Outrigger concept- a boat made up of three connected hulls: one main body and two side bodies (left and right).The boats are made using sunboard and joined using UPVC pipes. A plastic net is attached at the back which drags the waste along with it.
For control, an Arduino microcontroller, a FlySky transmitter and receiver and a L298N motor driver is used .The boat is maneuvered by four 500 RPM Johnson gear motors using 3 * 12 V Lithium Polymer battery .The navigation is guided by Raspberry Pi 5 camera module trained by ML using YOLO algorithm.
Here’s a breakdown of the circuit design and connections
Components used are
Arduino board
FlySky Transmitter and Receiver
L298N Motor Driver Module
4 × 500 RPM Johnson Gear Motors
2X 30 RPM motors
3* 12 V Lithium battery
Raspberry Pi 5 with a camera module that scans the water for floating plastic trash. I trained a machine learning model using the YOLO algorithm to detect garbage in images. YOLO is a super fast neural network that can identify objects in real-time. I trained it with hundreds of images of water trash to recognize different types of plastic floating in water. When it spots trash, the Pi sends commands to an Arduino microcontroller that controls the boat's motors.
Right now, the device can start from a home position, identify where the most trash is concentrated, navigate to that spot, circle around to collect debris in its net, and then return to its starting point
The device was taken to a nearby canal for a real life demo and test, the local villagers were pleased, here is a link of their comments. https://youtube.com/shorts/8lw9ui0h4Bk?si=QawThId5lRWioBZZ
I had to educate and convince the villagers and their sarpanch . It was a rich experience for me. While they were curious, there was a lot of resistance amongst them as they felt it is the government's responsibility and what will one act of cleaning by a kid change .Once I cleaned the lake partially , for a few days they stopped dumping plastic in the lake . So yes , it impacted environmental justice as they got access to clean water.
I have created a website to share the progress of my project , inspire and learn from other people. https://firstprincipletechie.com/
At the same time, I have participated in tech fests and competitions where my project got selected to learn from senior academicians and practitioners in this industry.
I started from version 1, which was a conveyor belt model. Issues with first model
1. Slow collection speed
2. Limited storage capacity
3. The first version struggled with handling plastic accumulation near riverbanks
4. Lack of autonomous capabilities meant constant manual operation
5. 12 V batteries drain out quickly
I tried addressing problems 1-3 in version 2 and in future iterations of this version I will be addressing 4 and 5 also by adding solar panels.
Version 2
1. It could collect larger amount of plastic at a faster pace as you can see from the video https://youtu.be/7LLMLa9pscQ and design above
2. With brushes attached to it, it could scrape against the banks also to push the waste towards the net.
3. Trained a machine learning model using the YOLO algorithm to detect garbage in images. YOLO is a super fast neural network that can identify objects in real-time. I trained it with hundreds of images of water trash to recognize different types of plastic floating in water. When it spots trash, the Pi sends commands to an Arduino microcontroller that controls the boat's motors. Right now, the device can start from a home position, identify where the most trash is concentrated, navigate to that spot, circle around to collect debris in its net and then return to its starting point all by itself!
contact Daksh Gandhi gandhijidaksh@gmail.com
I can help share my prototype with you , I would like to collaborate with mentors and private companies and build a more robust model which can be used near points where rivers merge with sea i.e delta regions.
