|
Aggregation aware distributed channel access in internet of things networks using reinforcement learning under partial observability
|
| |
|
|
|
چکیده: (5 مشاهده) |
The rapid growth of Internet of Things (IoT) networks has transformed data aggregation and medium access control into a large-scale sequential decision-making problem under uncertainty, where distributed nodes operate with incomplete and noisy observations. In dense and heterogeneous IoT environments, classical medium access protocols fail to efficiently balance channel utilization, collision avoidance, and aggregation delay. In this paper, aggregation-aware dynamic channel access is formulated as a partially observable Markov decision process, explicitly capturing imperfect sensing and environmental uncertainty. To enable practical distributed learning, the original model is approximated by a surrogate Markov decision process defined over an information state constructed from local observations and past actions. Based on this formulation, a reinforcement learning–based cognitive medium access framework is developed, in which each node autonomously learns an access policy by optimizing a reward function that jointly accounts for successful transmission, collision probability, idle channel usage, and aggregation-related delay. Simulation results obtained in dynamic multi-channel environments demonstrate that the proposed framework significantly outperforms classical schemes. In dense network scenarios, it achieves a channel utilization of up to 85.3%, while reducing the collision probability to approximately 0.08, compared with ALOHA and SDSA. Moreover, the average response delay is reduced to 12.4 ms, confirming the effectiveness of the proposed approach for aggregation-aware and uncertainty-aware IoT networks.
|
|
| |
|
|
متن کامل [PDF 632 kb]
(2 دریافت)
|
نوع مطالعه: پژوهشي |
موضوع مقاله:
عمومى
دریافت: 1405/4/11 | پذیرش: 1405/7/1 | انتشار: 1405/7/6
|
|
|
|
|
|
|
| ارسال نظر درباره این مقاله |
|
|
|
