Deep Q-Learning-Based Optimization of Path Planning and Control in Robotic Arms for High-Precision Computational Efficiency

Abstract

Optimizing path planning and control in robotic arms is a critical challenge in achieving high-precision and efficient operations in various industrial and research applications. This study proposes a novel approach leveraging deep Q-learning (DQL) to enhance robotic arm movements’ computational efficiency and precision. The proposed framework effectively ad-dresses key challenges such as collision avoidance, path smooth-ness, and dynamic control by integrating reinforcement learning techniques with advanced kinematic modelling. To validate the effectiveness of the proposed method, a simulated environment was developed using a 6-degree-of-freedom robotic arm, where the DQL model was trained and tested. Results demonstrated a significant performance improvement, achieving an average path optimization accuracy of 98.76% and reducing computational overhead by 22.4% compared to traditional optimization methods. Additionally, the proposed approach achieved real-time response capabilities, with an average decision-making latency of 0.45 seconds, ensuring its applicability in time-critical scenarios. This research highlights the potential of deep Q-learning in revolutionizing robotic arm control by combining precision and computational efficiency. The findings bridge gaps in robotic path planning and pave the way for future advancements in autonomous robotics and industrial automation. Further studies can explore the scalability of this approach to more complex and real-world environments, solidifying its relevance in emerging technological domains.