MRL-SCSO: Multi-agent Reinforcement Learning-Based Self-Configuration and Self-Optimization Protocol for Unattended Wireless Sensor Networks

摘要

Resource-constrained nodes in unattended wireless sensor network (UWSN) operate in a hostile environment with less human intervention. Achieving the optimal quality of service (QoS) in terms of packet delivery ratio, delay, energy, and throughput is crucial. In this paper, we propose a topology control and data dissemination protocol that uses multi-agent reinforcement learning (MRL) and energy-aware convex-hull algorithm, for effective self-configuration and self-optimization (SCSO) in UWSN, called MRL-SCSO. MRL-SCSO maintains a reliable topology in which the effective active neighbor nodes are selected using MRL. The network boundary is determined using convex-hull algorithm to maintain the connectivity and coverage of the network. The boundary nodes transmit data under high traffic load conditions. The performance of MRL-SCSO is evaluated for various nodes count and under different load conditions by using the Contiki's Cooja simulator. The results showed that MRL-SCSO stabilizes the performance and improves QoS.