Underwater Wireless Sensor Networks (UWSNs) face significant challenges due to noise, propagation loss, and delay, which affect network performance and reliability. This research introduces an adaptive routing protocol incorporating multi-agent reinforcement learning for efficient multi-hop transmission, a modified k-Means algorithm for optimized cluster head selection in Low-Energy Adaptive Clustering Hierarchy (MK-LEACH), and multi-layer trilateration to enhance deployment and sensor coverage. A quantitative approach was employed, utilizing numerical and statistical analysis based on Python-based simulations. The proposed methods were evaluated against the distance-and energy-constrained k-Means Clustering Scheme (DEKCS) for cluster formation, as well as the Q-Learning-Based Energy-Efficient and Lifetime-Aware Routing (QELAR) protocol and the Energy-Balancing Routing Protocol for WSNs based on Reinforcement Learning (EBR-RL) for multi-hop transmission from the cluster head to the base station. Key performance metrics included network lifetime, node failure rate, total packets sent, and packet data ratio. The results indicate that the modified k-Means algorithm reduces node failure by 84.58% compared to Low-Energy Adaptive Clustering Hierarchy (LEACH) and 18.08% compared to k-Means, while multi-layer trilateration decreases redundancy by 75.5% compared to random deployment. Additionally, the MK-LEACH protocol achieved a 9.99% improvement in packet data ratio over EBR-RL and a 187.14% improvement over QELAR, with a total data transfer of 294,325 bytes. These findings demonstrate the enhanced robustness and efficiency of the proposed approach for UWSNs in underwater monitoring applications. © 2025 by the authors.