Adaptive TD3: Robust Mobile Robot Navigation in Dynamic Environments with Mixed Obstacles

Abstract

Mobile robot navigation in dynamic environments poses intricate challenges due to the coexistence of static and dynamic obstacles. This abstract presents a comprehensive study focused on leveraging the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm to enhance mobile robot navigation capabilities in such dynamic environments. The TD3 algorithm, renowned for its robustness in complex scenarios, incorporates twin critics and delayed policy updates to promote effective exploration and stability. In this study, TD3 is applied to enable mobile robots to navigate amidst a combination of static and dynamic obstacles, addressing key aspects such as obstacle avoidance, path planning, and collision prevention. Through rigorous experimental evaluations within a simulated environment, the performance of the TD3-based navigation system is assessed. The evaluations encompass the system's adaptability and responsiveness to environmental changes caused by both moving and stationary obstacles. Furthermore, the trade-offs between exploration and exploitation in dynamic environments are analyzed, accompanied by proposed techniques to optimize the TD3 algorithm for improved navigation performance. The findings obtained shed light on the efficacy of the TD3 algorithm in dynamic environments with mixed static and dynamic obstacles. They offer valuable insights for researchers and practitioners, empowering them to deploy TD3-based navigation systems capable of efficiently navigating complex scenarios. Additionally, the outcomes have practical implications for a wide range of applications, including autonomous vehicles, warehouse automation, and robotic surveillance, where reliable and adaptive navigation in dynamic environments is of paramount importance.

Cite as

Dinesh Kumar Sahoo. (2023). Recent Trends in Automation and Automobile Engineering. Recent Trends in Automation and Automobile Engineering, 6(2), 4–18. https://doi.org/10.5281/zenodo.8004578