|
Improving search for deep reinforcement learning with ant colony optimization
|
Mohammad Hassan Nataj Solhdar *   |
|
|
|
Abstract: (255 Views) |
| The emergence of the massive Internet of Things (IoT) ecosystem is changing human lifestyles. The Internet of Things still relies on human assistance and has unacceptable response times for analyzing big data, and still faces significant challenges. Therefore, it is very necessary to create a new framework and algorithm to solve the specific problems of fast Internet of Things. Reinforcement learning and deep reinforcement learning (DRL) approaches have the ability to make decisions, but traditional modeling and training methods are time-consuming and limit their applications. To overcome this problem, this article proposes a reinforcement learning method suitable for the Internet of Things. In this way, we propose a feature selection method based on ant colony optimization (ACO). Since the heuristic functions affect the decision-making process of ACO during the search process, the use of heuristic learning method can help the algorithm to search the search space better. Finally, as a case study of IoT, the proposed method is applied to traffic light control, with the aim of reducing traffic congestion in smart city intersections. Experimental results show that the proposed method can learn better actions in a shorter execution time compared to traditional approaches. |
|
| |
|
|
Full-Text [PDF 922 kb]
(62 Downloads)
|
Type of Study: Research |
Subject:
Computer Engineering
Received: 2023/05/14 | Accepted: 2023/08/12 | Published: 2025/04/12
|
|
|
|
|
|
|
| Send email to the article author |
|
|
|
| Add your comments about this article |
|
|
|
