Deep U-Shape Vision Transformer for Scheduling and Trajectory Planning in Wireless Rechargeable Sensor Networks

Abstract

Wireless Rechargeable Sensor Networks (WRSNs) have emerged as an effective solution for extending the lifetime of traditional Wireless Sensor Networks through mobile charging mechanisms. However, challenges such as sequence scheduling, trajectory planning, obstacle avoidance, and real-time decision-making remain significant due to dynamic environments and limited energy resources. Recent advancements in deep learning, particularly U-Net architectures and Vision Transformers (ViTs), have shown strong potential in solving these optimization problems. This systematic review examines the integration of Deep Convolutional U-Shape Networks with jump attention mechanisms and transformer-based models for intelligent scheduling and trajectory planning in WRSNs. U-Net architectures provide effective spatial feature extraction through encoder–decoder structures and skip connections, enabling accurate environmental representation and obstacle detection. To overcome limitations in capturing long-range dependencies, attention mechanisms and Vision Transformers have been integrated to improve global context modelling and sequence prediction. Hybrid architectures such as TransUNet and UNetFormer combine local convolutional feature extraction with transformer-based attention, achieving improved performance in path planning, obstacle avoidance, and scheduling optimization. Furthermore, deep reinforcement learning and attention-guided frameworks enable adaptive and energy-efficient charging strategies, enhancing network stability, reliability, and overall WRSN performance in dynamic environments.