Autonomous Robotic Assembly through AI-based Computer Vision

As manufacturing tasks become increasingly complex and dynamic, traditional automation solutions struggle to keep up with rapidly changing markets and the demand for highly customized products. Smarter, self-governing manufacturing systems are needed.​ This research aims to develop an autonomous robotic assembly system driven solely by Artificial Intelligence (AI) algorithms and computer vision (CV), reducing human effort, increasing efficiency, and optimizing overall productivity in robotic automated production systems. Utilizing a 6-DOF robotic arm and a vision module consisting of cameras at different angles, this project demonstrates a use case for autonomous robotic assembly. The project has proven the concept of autonomous assembly with the capability to identify multiple objects in basic shapes and create robotic assembly tasks and the corresponding control commands. After executing each task, a quality check is performed in which the vision module detects any mismatch between the expected position and actual position of the assembled part. If a mismatch is detected the vision module issues instructions to the robot to execute corrective action (rework) until the quality check is passed. Furthermore, all the executed tasks and the associated quality data are stored in quality database. This helps in performing root cause analysis for any error generated due to rework or placement of the parts. This work demonstrates the feasibility of autonomous robotic assembly, contributing to advancements in smart manufacturing and Industry 4.0.