Deep Reinforcement Learning Assisted Spectral Clustering and Hybrid RRT-ACO Optimization for Energy-Efficient Obstacle-Aware Mobile Sink Trajectory Planning in WSN
Year: 2026
Type: conference-paper
Abstract: WSN applications need energy-efficient and dynamic schemes for collecting sensory information. Current mobile sink trajectory algorithms have problems with long latencies, premature energy dissipation, and poor obstacle avoidance. The proposed work offers a hybrid approach that combines Spectral Clustering, Rapidly-Exploring Random Tree Star (RRT), Ant Colony Optimization (ACO), and Deep Reinforcement Learning (DRL) to develop optimal trajectories for mobile sinks. Spectral clustering algorithm is utilized to select an Rendezvous Point (RP) with optimal residual energy, density, and cost of communication. RRT is used to create collision-free trajectories and ACO optimizes sink movement with the lowest cost of travel. On the other hand, DRL helps to adapt sink trajectory depending on varying conditions in the network. Formulations of mathematical models concerning energy conservation, reducing latency, and packet delivery are considered to increase its analytical significance. Results obtained from simulations indicate improvements in terms of increased network lifetime by 18%, energy conservation of 21%, reduced latency by 15%, and better packet delivery ratio when compared to existing EDMSP and RRTMT approaches.
Language: English
Subfield: Computer Networks and Communications
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7202379550