DESIGN AND APPLICATION OF VISION-BASED ASSISTIVE ROBOTIC ARM FOR MATERIAL HANDLING IN INDUSTRIAL ENVIRONMENTS

Abstract

This thesis presents the design, implementation, and evaluation of a vision-guided assistive robotic arm system developed entirely in simulation using CoppeliaSim. Aimed at material-handling tasks in industrial environments, the system leverages classical image processing techniques, specifically blob detection and inverse kinematics, to perform real-time object localization and adaptive grasping. The robotic arm, modeled with five degrees of freedom and a dynamically actuated gripper, is guided by a ceiling-mounted vision sensor. Using Lua scripting and simIK-based control, the robot autonomously detects a cylindrical object, computes its pose, and executes a pick-and-place operation. Mechanical improvements to the gripper and tip alignment corrections were introduced to enhance grasp reliability. The system’s kinematic accuracy was validated using MATLAB, applying the Product of Exponentials formulation and comparing predicted and simulated end-effector orientations. Experimental results in simulation demonstrated a 100% grasp success rate post-modification, sub-centimeter placement error, and minimal deviation in predicted Euler angles. The system’s modular architecture, use of open-source tools, and demonstrated performance highlight its potential for low-cost prototyping and educational applications. Recommendations for future work include integrating closed-loop feedback, expanding object diversity, and bridging the simulation-to-reality gap through physical prototyping.