Tropical Cyclone (TC) estimation aims to accurately estimate various TC attributes in real time. However, distribution shifts arising from the complex and dynamic nature of TC environmental fields, such as varying geographical conditions and seasonal changes, present significant challenges to reliable estimation. Most existing methods rely on multi-modal fusion for feature extraction but overlook the intrinsic distribution of feature representations, leading to poor generalization under out-of-distribution (OOD) scenarios. To address this, we propose an effective Identity Distribution-Oriented Physical Invariant Learning framework (IDOL), which imposes identity-oriented constraints to regulate the feature space under the guidance of prior physical knowledge, thereby dealing distribution variability with physical invariance. Specifically, the proposed IDOL employs the wind field model and dark correlation knowledge of TC to model task-shared and task-specific identity tokens. These tokens capture task dependencies and intrinsic physical invariances of TC, enabling robust estimation of TC wind speed, pressure, inner-core, and outer-core size under distribution shifts. Extensive experiments conducted on multiple datasets and tasks demonstrate the outperformance of the proposed IDOL, verifying that imposing identity-oriented constraints based on prior physical knowledge can effectively mitigates diverse distribution shifts in TC estimation.

In this paper, we study the unmanned aerial vehicle (UAV)-assisted emergency communication network, where multiple UAVs are dispatched from emergency centers and cruise the target region to provide service. To ensure timely communications, a strict requirement is that the time spent on the tour of each UAV is no greater than a delay threshold. Our optimization task is to minimize the total cost while satisfying the delay constraints, in which the total cost is a weighted sum of deployment cost and cruise cost. We propose a joint deployment and path planning scheme to address the NP-hard task from a load-balancing perspective. Specifically, we first adopt a continuous approximation method to estimate the number of needed emergency centers and locate them by clustering. Then, we assign the ground users to the emergency centers via loadbalancing region partitioning, based on which the path planning of multiple UAVs can be reduced to independent traveling salesman problems. Numerical results demonstrate that our proposal can complete the delay-bounded mission at the lowest cost as compared with other methods, providing an efficient and cost-effective way for practical emergency communications.

Unmanned aerial vehicle (UAV)-assisted communications has emerged as a promising technology in many domains, such as disaster relief and emergency scenarios. However, the limited battery capacity restricts UAVs from performing such persistent missions. In this paper, we consider a general hybrid trajectory planning problem to efficiently provide emergency communications in time-constrained disaster-affected regions, where a ground vehicle carrying backup batteries moves along with the UAV as a “mobile charging platform” to handle the energy issue of the UAV. Our optimization task is to minimize the total cost of the mission. We show that the optimization task is an extension of the traveling salesman problem with soft time window constraints. Due to the NP-hardness of the task, we propose a novel deep reinforcement learning with a sequential model strategy to learn the policy for the UAV's visiting order, based on which the collaborative routes of the UAV and ground vehicle are designed. Numerical results show that our proposed learning-based route planning scheme is effective and efficient.

Nowadays, the compelling applications of the Internet of Things (IoT) bring unexpected economic benefits to our daily lives. But at the same time, it also poses huge challenges to service providers. Diverse proprietary hardware (i.e., firewall and code conversion) have to be deployed in networks for meeting different applications’ requirements. Recently, network functions virtualization (NFV) is considered a promising technique. In the NFV-enabled architecture, network services can be implemented via a set of orderly virtual network functions (VNFs) on standardized compute nodes, which is termed service function chains (SFCs). However, with the explosion of IoT applications, embedding multiple SFCs in a shared NFV-enabled infrastructure becomes a challenging problem. Centralized schemes suffer from the scalability and private issue, while distributed schemes suffer from the nonconvergence problem. In this article, we propose a hybrid intelligent control architecture, which adopts the centralized training and distributed execution paradigm. A centralized critic is introduced to ease the training process of the distributed network nodes. Besides, considering the competitive behavior of users, we formulate the resource allocation problem as a multiuser competition game model. Based on this, we proposed a multiagent reinforcement learning-based SFCs deployment algorithm.

Due to the flexibility and agility, unmanned aerial vehicle (UAV) is a promising way for gathering data generated by wireless sensor networks. However, the limited battery capacity of the UAV restricts its application on many occasions, e.g., the networks deployed in the wild. In this paper, we propose a cooperative trajectory planning scheme to deal with the energy issue of the UAV, where a truck carrying backup batteries moves along with the UAV acting as a “mobile recharging station”. Our optimization task is to minimize the total mission time for gathering data from all the sensor nodes, which can be achieved by solving two problems: First, we need to divide the entire mission area into multiple subregions so that the UAV can hover over each subregion to collect the data of the sensor nodes through just one taking-off and landing under the constraint of battery capacity; second, we should find out the optimal trajectory of the truck so that the UAV can get to the hovering positions of each subregion from the truck and fly back to it before the battery drains considering the road condition in real world. We introduce an efficient clustering algorithm to partition the area into subregions in a load-balanced way to minimize the number of movements of the UAV. The trajectory planning task is formulated as a coordinated traveling salesman problem, which is solved by a three-step trajectory planning algorithm heuristically, and we also give the analysis of the upper bound and lower bound to demonstrate the performance guarantee. Numerical results show that our proposed scheme provides an effective and cost-efficient way for the data collection of large-scale wireless sensor networks in practical scenarios.

Unmanned aerial vehicle (UAV) is a promising way to collect data generated by wireless sensor networks, nevertheless, the battery capacity of the UAV restricts its application on many occasions, e.g., the network deployed in the wild. In this paper, we propose a coordinated route planning scheme to deal with the energy issue of the UAV, where a truck carrying backup batteries moves together with the UAV as a “mobile recharging station”. Our optimization task is to minimize the total mission time for collecting data from all the sensor nodes. We develop an efficient clustering algorithm to divide the entire mission area into multiple subregions in a load-balanced way to minimize the number of movements of the UAV, and formulate the trajectory planning task as a coordinated traveling salesman problem which is heuristically solved by a three-step route planning algorithm. Numerical results show that our proposed scheme provides an effective and cost-efficient way for the data collection of wireless sensor networks in practical application scenarios.

Densely deployed base station (BS) network is one of the important technologies for 5G and beyond mobile communication system, which improves the system throughput by deploying a large number of BSs in the service area. However, such a mobile network has to deal with the consequent power issue since the energy consumption of the BSs generally accounts for a substantial part of the whole system. In this paper, we propose an intelligent BS sleeping scheme to reduce the system energy consumption as much as possible with reasonable signaling overhead while guaranteeing the quality of experience of users. First, we introduce a long short term memory learning method to forecast the traffic distribution in the service area, by which we can determine when the BS sleeping operation is triggered; second, we develop an efficient three-step procedure to determine which of the BSs would sleep or be wakened. Experiment results show that our proposed traffic prediction method works quite well in practical scenarios. The prediction error is not more than 10%, and the energy consumption decreases more than 40% in average for a commercial mobile network with 20 BSs.