| Autor: |
Jebrane, Walid, El Akchioui, Nabil |
| Předmět: |
|
| Zdroj: |
Journal of Control, Automation & Electrical Systems; Dec2024, Vol. 35 Issue 6, p1059-1077, 19p |
| Abstrakt: |
In the pursuit of advancing robotic navigation, a notable enhancement to the distributed distributional deterministic policy gradients (D4PG) algorithm is presented. Cutting-edge methodologies, including the implicit quantile network (IQN) critic, prioritized experience replay (PER), and N-step Bootstrapping, are integrated to achieve significantly improved learning efficiency and performance in autonomous systems. The enhanced algorithm is implemented using PyTorch and systematically evaluated within the robo-gym environment, demonstrating substantial advancements over traditional deep reinforcement learning methods. The results highlight the augmented D4PG framework's superior capability in navigating complex control challenges and suggest promising avenues for further exploration in advanced robotics and autonomous systems. [ABSTRACT FROM AUTHOR] |
| Databáze: |
Complementary Index |
| Externí odkaz: |
|
Full text from SpringerLink