Problem Statement
The goal of this project is to implement a robotic system which autonomously harvest green peppers from outdoor farms. The robotic system will be mobile to navigate around the fields and be equipped with robotic arms to grasp and separate the peppers from the plant. A brief description of our system is as follows :
- A robotic system implementing a combination of computer vision models to detect peppers and compute reachable targets within the workspace.
- This system will not be restrained to greenhouse environments. This system’s ability to withstand environmental changes in lighting and weather allows continuous field operation, making it ideal for agriculture
- This system has a bimanual arm setup, with a gripper mounted on one arm for precisely grasping the fruit and a pruner/cutter mounted on another to perform extraction.
- Each of the system’s arms has independent control which accounts for dexterity, to adjust for different orientation of the peppers and to avoid accidentally drop of peppers during harvesting.
Use Case
It’s harvesting season, and our farmer Tom, who has a mid sized pepper farm is looking to make a bountiful harvest.
He deploys VADER, an autonomous pepper harvesting robot into the field, and selects the mode of operation. When set to autonomous mode, VADER reaches the appropriate position near the pepper plant using the farmer’s guidance to start harvesting.
It scans the plant and identifies reachable peppers and their peduncles, and selects one of them for harvesting. VADER moves it’s gripper arm near the pepper and it’s cutter arm near the peduncle. It aligns its gripper and cutter to match the pepper’s orientation and then performs precise extractions, and places it in the designated storage area. The entire process repeats until all the peppers are harvested.
In the case of failure, VADER prompts Tom to select the teleoperation mode, where he monitors a real-time video feed, and remotely guides VADER to pick peppers with accuracy.