History-Aware Deep Reinforcement Learning for real-time robot navigation in dynamic industrial environments

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

The advent of deep reinforcement learning (DRL) has significantly advanced the capabilities of autonomous mobile robots (AMRs) in navigating complex and dynamic environments. However, existing DRL models often struggle in industrial settings characterized by their unpredictable and changing nature. To address this challenge, we propose a novel History-Aware Deep Reinforcement Learning (HA-DRL) framework designed to enhance AMRs’ navigation capabilities by incorporating historical context into the learning process. The HA-DRL framework enriches the decision-making process by leveraging past experiences, enabling robots to make informed decisions in real-time, thus improving navigation efficiency and safety. Through extensive experimentation in both simulated and real-world industrial settings, we demonstrate that HA-DRL consistently outperforms traditional DRL approaches in terms of navigation performance. This research highlights the importance of historical context in developing more adaptable, efficient, and safe autonomous navigation systems for industrial applications. The experimental video is available at https://youtu.be/6LQtWLcwzTg .