Abstract:Planning the movement of the sink to maximize the lifetime in wireless sensor networks is an essential problem of great research challenge and practical value. Many existing mobile sink techniques based on mathematical programming or heuristics have demonstrated the feasibility of the task. Nevertheless, the huge computation consumption or the over-reliance on human knowledge can result in relatively low performance. In order to balance the need for high-quality solutions with the goal of minimizing inference time, we propose a new framework combining heterogeneous graph neural network with deep reinforcement learning to automatically construct the movement path of the sink. Modeling the wireless sensor networks as heterogeneous graphs, we utilize the graph neural network to learn representations of sites and sensors by aggregating features of neighbor nodes and extracting hierarchical graph features. Meanwhile, the multi-head attention mechanism is leveraged to allow the sites to attend to information from sensor nodes, which highly improves the expressive capacity of the learning model. Based on the node representations, a greedy policy is learned to append the next best site in the solution incrementally. We design ten types of static and dynamic maps to simulate different wireless sensor networks in the real world, and extensive experiments are conducted to evaluate and analyze our approach. The empirical results show that our approach consistently outperforms the existing methods on all types of maps.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is to maximize the network lifetime in wireless sensor networks (WSN) by planning the path of mobile receiving points (sinks). Specifically, the paper focuses on how to efficiently schedule the path of mobile sinks to balance the energy consumption in the entire network, thereby prolonging the overall lifetime of the network. This problem is of great significance in practical applications because wireless sensor nodes are usually energy - limited and difficult to replace or recharge, so maximizing the network lifetime is one of the key challenges in improving the practicality of WSN.
### Background and Problem Description of the Paper
Wireless sensor networks (WSN) are widely used in fields such as traffic control, military target tracking, and biomedical health monitoring. A typical WSN consists of a sink and multiple sensor nodes, which are used to collect information from the environment (such as temperature, pressure, location, etc.). The collected data is transmitted to the sink through multi - hop communication and then sent to the user by the sink. Since sensor nodes are usually non - recoverable and energy - limited, the lifetime of a WSN is usually defined as the duration before the first or more sensor nodes in the network run out of energy. In addition, the inaccessibility of sensor deployment leads to high charging costs. Therefore, how to maximize the lifetime of WSN has become an important problem in the research field of WSN.
Utilizing the mobility of sinks has been proven to be an effective method to prolong the lifetime of WSN. By allowing mobile sinks to move between different positions in the sensing area, more balanced energy consumption can be achieved, thereby prolonging the lifetime of WSN. However, determining the optimal traversal path of sinks is an NP - hard problem. Existing methods, such as those based on mathematical programming or heuristic algorithms, although they can find feasible solutions to a certain extent, often require huge computational resources or rely on sub - optimal human knowledge, resulting in low performance.
### Solutions Proposed in the Paper
To solve the above problems, the paper proposes a new deep reinforcement learning framework - HGFF (Heterogeneous Graph - based Feature Fusion), which combines heterogeneous graph neural networks and deep reinforcement learning techniques to automatically construct the movement path of sinks. The main contributions include:
1. **Proposing a new deep reinforcement learning framework**: HGFF automatically generates effective heuristic strategies through deep reinforcement learning to guide the movement of sinks, which requires neither a large amount of computational resources nor depends on human knowledge.
2. **Modeling WSN as a heterogeneous graph**: Use graph neural networks to extract node features, aggregate the information of neighboring nodes through the message - passing mechanism, and introduce type embedding to learn the heterogeneous information of different types of nodes.
3. **Introducing a global attention mechanism**: Different from common graph attention networks, HGFF adopts a global attention mechanism, allowing station nodes to pay attention to the information of all sensor nodes in the graph, which improves the expressive ability of the model.
4. **Experimental verification**: By designing a variety of static and dynamic maps to simulate different WSN scenarios in the real world and conducting extensive experiments, the results show that HGFF significantly outperforms existing methods in all types of WSN.
### Conclusion
The HGFF framework proposed in the paper has made significant progress in solving the sink path planning problem in WSN. It can not only effectively prolong the network lifetime, but also has high computational efficiency and generalization ability, providing new ideas and technical support for the practical application of WSN.