Autonomous Weed Plucker

Project Title

Autonomous Weed Plucker

Project Area of Specialization RoboticsProject Summary

 An Autonomous Vehicle plucking weeds from your fields, giving you live updates for field temperature, moisture, weeds density in your crop, overcoming hurdles faced and informing if it can’t overcome. A robot doing above all while sending you it’s own status like Battery level, current location, total distance covered, all sensors and camera status, approximate working time on remaining battery, answering cause if stopped working. All mentioned things on your palm. Machine learning model has revolutionized the world of Artificial Intelligence. Using these models, Cameras trained to both identifying target plant and driving vehicle, will notify the system. Raspberry pi will use obtained information to operate robotic arm toward pointed area and will pluck weed. Different installed Sensors will provide rest of the information about environment. This project aims to reduce farmer expenses,  hard work to clean fields and resulting maximum profit. 

Project Objectives

Goal of project is an ‘Autonomous vehicle self driving in fields and plucking weeds, not damaging cotton crop and sending live updates to user in mobile application’. There are many phases to pass through and reach goal state. Some major are listed below: 
? Train model for object detection

? Using installed camera, implement this model for vehicle to self drive

? Understand field patterns and visit accordingly ? Overcome hurdles and find its path

? Reporting back to user if stuck somewhere or stopped

? Well integrated robotic arm

? Use camera’s provided frames to differentiate between cotton and weed

? Locate weed and instruct robotic arm to move

? Successful plucking of weed

? Not damaging cotton plant during traveling and plucking

? Use installed sensors to get data and measuring moves

? Sending live updates of movements and operations

? Mobile App for basic control and data summary 

Project Implementation Method

Idea is to build a vehicle which will’ve the ability to distinguish between your crop plant and other unnecessary plant to pluck them. Below are major components of this robot: 
? A microprocessor (we’ll use Raspberry pi)

? A controllable Vehicle for autonomous drive

? Cameras to classify between cotton and weed

? A mechanical robotic arm with claw to move and pluck weed

? Sensor, Battery and mobile application for real time data summary 

Plan is to use machine learning models like tensorflow2 (use for object detection) on camera provided frames, which will help system identifying route to travel. Surrounding plants row and ground surface will help system to drive autonomously. System will also use these frames to identify cotton crop and other weed using Python Language libraries like OpenCV, Pillow,  PIL etc. We’ll train our model for cotton plant only, it’ll reduce complexity. After that Our Algorithms will calculate required information for robotic arm to reach target plant. Another camera will be installed just before claws of arm, whose recording will help system to get more precise location to pluck weed from roots and will be providing updates about movement arm is making, which will determine progress in reaching destination and in case of wrong move, redirect it. Also report system if hurdle occur. Battery will provide energy to operate. Temperature sensor, speed sensor, Voltage sensor, GPS etc. will help sending environment and robot information to user. Some hardware and software components are mentioned above that we’ll use. 

Benefits of the Project

Cotton is one of most important crop of Pakistani Agriculture as Pakistan is listed in biggest importer of garments all around the world bringing a lot of money to country. But it is mostly observed that farmers do a lot of hard work in order to meet the expected yield of crop. Farmers whole season utilized just to dig out unnecessary plants from crop and at end of day he barely have progress of 1 acre. After some period of time, when plant grow he manually can’t go inside and clean weed one by one. He will damage plant and reduce growth. Ultimately he has to buy expensive herbicides. 
Despite the fact that Pakistan’s economy is majorly dependent on Agriculture. Yet Agriculture is not being digitalized, while other countries are going miles ahead in this aspect. For sake of example we can see strawberry picker. A giant machine composed of multiple sections, identifying ripe fruit and calculating decision to make. Moving forward autonomously, storing them safely and then even packing them itself. Owner Careg Minkowski harvest strawberry on 600 acers. He stated in an interview 
“I need 600 worker to clean my fields after every 3-4 days for my 600 acer. Finding that huge human power, controlling them and getting max output from them is difficult. Labor cost increases much high level. Now, This single machine with few operator and some worker is speeding up work rate, replacing 30 workers at a time. It is saving me 20,000 US$ just in case  human labor.” 
Similarly many more machines like that have been developed. Like spraying machine, spraying on specific plant (weed only), automatically plowing field completely. All these effort scientist are making have strong reason. Top of them is population growth and rate of food required to meet all needs. Secondly, Expense to profit ratio. Most of farmer’s expenses are because of these irrelevant weeds  which destroy crop in two ways. First disturbing it in root and not letting it to grow properly. Secondly, chemicals farmer is using to power it’s crop, they also take their part in it.  

After all these problems and background, we decided to develop a machine that would allow farmer convenience in case of cotton crop, increasing growth rate, reducing expenses, reducing workload etc. This project is also a key of that chain. This is an initiative of modern and digital agriculture in Pakistan. 

Technical Details of Final Deliverable

Plan is to use machine learning models like tensorflow2 (use for object detection) on camera provided frames, which will help system identifying route to travel. Surrounding plants row and ground surface will help system to drive autonomously. System will also use these frames to identify cotton crop and other weed using Python Language libraries like OpenCV, Pillow,  PIL etc. We’ll train our model for cotton plant only, it’ll reduce complexity. After that Our Algorithms will calculate required information for robotic arm to reach target plant. Another camera will be installed just before claws of arm, whose recording will help system to get more precise location to pluck weed from roots and will be providing updates about movement arm is making, which will determine progress in reaching destination and in case of wrong move, redirect it. Also report system if hurdle occur. Battery will provide energy to operate. Temperature sensor, speed sensor, Voltage sensor, GPS etc. will help sending environment and robot information to user. Some hardware and software components are mentioned above that we’ll use. 

Anticipated results / Expected Outcomes

Expectation is to get a Raspberry pi controllable Vehicle having a robotic arm fixed with vehicle’s body and some camera and sensor installed on it. It’ll be powered by battery. Targeted robot will have ability to traverse all cotton field model. It’ll overcome hurdles if found and report to user if it can’t overcome. This Weed picker will understanding difference between cotton and weed plant to pluck weed consciously. This robot will be sending live real time information about field, progress, expected more time to complete, energy source, environment heat etc. to user mobile app. 

Final Deliverable of the Project HW/SW integrated systemType of Industry Agriculture Technologies Artificial Intelligence(AI), RoboticsSustainable Development Goals Decent Work and Economic Growth, Industry, Innovation and InfrastructureRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	78700
Camera	Equipment	4	3000	12000
robotic arm	Equipment	1	17000	17000
metalic chasis car	Equipment	1	29000	29000
battery 12v	Equipment	1	4000	4000
raspberry pi	Equipment	1	7500	7500
sensors + wires	Miscellaneous	12	600	7200
other	Miscellaneous	1	2000	2000