강화 학습과 다기준 의사 결정 방법을 활용한 무선 센서 네트워크 성능 향상 기법

초록

This paper proposes a method to simultaneously enhance energy consumption efficiency and data transmission in wireless sensornetworks using a multi-criteria decision making approach. The goal is to enable each sensor node in a power-constrained wireless sensornetwork to dynamically adjust its sampling interval, thereby improving both energy consumption efficiency and data collection volume. To address this challenge, we introduce the TOPSIS-AL (TOPSIS - Adaptive Learning), which combines TOPSIS (Technique for Order ofPreference by Similarity to Ideal Solution) with the Q-Learning. Unlike traditional multi-criteria decision-making methods thatdeterministically present a single solution, our approach probabilistically suggests multiple solutions simultaneously. Furthermore, weautomated the weight estimation process by enabling each sensor node to determine the weights for the two criteria (estimated energyconsumption and estimated data throughput) based on its current state using the entropy weight method. Experimental results comparingthe TOPSIS-AL technique with existing self-adaptive and Q-Learning methods demonstrate that TOPSIS-AL outperforms these methodsin terms of five evaluation metrics, including the average energy level of sensor nodes, average cumulative data transmission volume,and average energy consumption efficiency per learning iteration.

키워드