Team Members: Mohamed Ali, Anjana Hevaganinge, Ravidu Hevaganinge, Sri Ram, Sri Kiran, and Joshua Ehizibolo
Our lab is partnering with the J.M. Clayton Seafood Company to address a pressing challenge in the crab-picking industry. Lump crab meat is highly valued, often fetching premium prices, yet its extraction remains labor-intensive, relying on skilled migrant workers. With rising socioeconomic pressures, challenges in retaining labor, and health risks like allergies from prolonged crab exposure, there is an urgent need to automate this process.
We are developing a robot capable of extracting lump crab meat precisely and efficiently. The process involves two key steps: first, capturing the intricate hand movements of skilled workers as they extract the meat, and second, programming the robot to replicate these movements. To achieve this, we’ve designed an advanced pose-tracking system that uses lasers and high-speed cameras to recreate an RGB depth map of human wrist movements. This setup captures fine details with unmatched accuracy, ensuring the robot can extract meat without damaging it.
Deep Imitation Learning (DIL) automates the labor-intensive task of crab meat picking, traditionally reliant on skilled human workers. By integrating machine vision, robotics, and advanced learning algorithms, the system replicates precise manual processes while maintaining product quality. Human demonstrations of crab handling are recorded using a lightweight replica robot, capturing visual and tactile data to train machine-learning models. Supervised learning replicates trajectories, while reinforcement learning optimizes for minimal meat loss and maximum yield. Robots equipped with specialized end-effectors execute tasks like scooping jumbo lump meat and brushing chamber meat, guided by DIL-trained policies that adapt to variations in crab morphology. This scalable approach addresses labor shortages and inefficiencies, redefining food processing with enhanced efficiency, safety, and sustainability.
Beyond crab picking, this technology holds promise for broader applications, such as surgical robotics, where precise replication of human motions is critical. By combining cutting-edge vision systems with robotics, we aim to redefine automation in food processing and beyond.