Vol. 2, Issue 1, Part A (2025)

The study explores the application of reinforcement learning (RL) to enhance human-robot collaboration (HRC) in industrial automation, focusing on developing a multi-agent reinforcement learning (MARL) framework that enables dynamic, safe, and adaptive cooperation between human operators and collaborative robots (cobots). Traditional rule-based control architectures often fail to accommodate unpredictable human actions and non-linear task environments, limiting efficiency and safety in shared workspaces. To address these challenges, the present research integrates human-intent prediction, safety-aware reward shaping, and multi-sensor perception into a unified RL-based control model. Experimental evaluations were conducted in simulated and physical assembly settings involving ten participants executing cooperative tasks such as peg-in-hole assembly and object handover. Quantitative performance metrics—task success rate, cycle time, throughput, and safety incidents—were analyzed using paired statistical tests, while qualitative assessments measured perceived collaboration fluency and operator trust. Results revealed a substantial improvement in productivity, with MARL achieving higher task success rates and up to 20% reduction in cycle time compared to baseline controllers. Safety compliance improved significantly, evidenced by fewer speed-and-separation monitoring breaches, reduced contact forces, and greater human-robot distance margins. Subjective ratings also indicated enhanced fluency and comfort during interaction. These outcomes confirm that reinforcement learning empowers robotic systems with the capacity for continuous adaptation and shared decision-making, thereby promoting safer and more efficient human-machine partnerships. The study concludes that MARL-based frameworks represent a major step toward realizing the goals of Industry 4.0, where intelligent, learning-driven robotic systems can seamlessly integrate human insight with machine precision. Practical recommendations include embedding RL algorithms into industrial robotic systems, implementing simulation-based safety validation, promoting human-centered design in robot interfaces, and establishing standardized training protocols to facilitate human-robot co-learning. Overall, the research highlights the critical role of reinforcement learning as a foundational technology for next-generation smart manufacturing environments.