Vision-Based Collaborative Robots for Precision Assembly: Comparative Evaluation of Rule-Based Control and Deep Learning-Enabled Adaptive Manipulation

Abstract

Collaborative robots increasingly support precision assembly in electronics, biomedical device manufacturing, and high-mix industrial production. However, conventional rule-based robotic control systems often perform poorly when parts vary in orientation, surface reflectivity, tolerance deviation, or human–robot interaction condtions. This article comparatively evaluates two collaborative robotic assembly architectures: deterministic rule-based visual servoing and deep learning-enabled adaptive manipulation. Using computational robotics simulation, vision-based pose estimation, force-feedback control analysis, and benchmark assembly-task evaluation, the study investigates how adaptive perception-control integration affects assembly accuracy, cycle time, error recovery, safety responsiveness, and industrial scalability. The analysis demonstrates that rule-based control provides transparency, stability, and predictable behavior in structured environments but becomes brittle under visual uncertainty and component variability. Deep learning-enabled manipulation improves object recognition, pose adaptation, and recovery from misalignment, although it requires stronger validation, explainability, and safety assurance. The article contributes to engineering scholarship by linking robotic perception, control theory, machine learning, human–robot interaction, and manufacturing systems analysis into a unified framework for adaptive collaborative automation.