ABSTRACT High‐volume data transmission in Internet of Things (IoT) networks demands the establishment of high‐quality multi‐hop communication paths at the network layer. To address this challenge, this paper proposes a joint routing and resource allocation framework that simultaneously optimizes relay node selection and transmit power allocation under decentralized manner. The proposed approach leverages deep reinforcement learning (DRL) to enable hop‐by‐hop decision‐making with only local observations. To mitigate the partial observability stemmed from limited perception range of node, a transformer‐based architecture is integrated into the DRL agent to enable the fusion of observations and actions collected along the established path. Specifically, the encoder module aggregates historical and frontier observations along the forwarding path, while the decoder module models temporal dependencies among historical actions. The fused representation is utilized to generate optimal decisions for next‐hop node selection and power allocation. After each action execution, a one‐hot encoded placeholder of the chosen action is fed back into the decoder module for subsequent decisions. Finally, numerical simulations demonstrate that the proposed transformer‐enhanced DRL framework significantly outperforms state‐of‐the‐art baselines in terms of end‐to‐end path quality, and robustness under decentralized IoT network configuration.

The emergence of 6G networks demands realtime autonomy and semantic intelligence in dynamic, missioncritical environments such as UAV swarms. Traditional control paradigms, including rule-based and black-box AI systems, struggle to interpret high-level intents, adapt to disruptions, and coordinate distributed agents effectively. In this paper, we present the Knowledge-Driven Autonomous Network Lite (KAN-lite), a lightweight semantic reasoning module designed to support intent-aware task coordination in multiagent networks. KAN-lite integrates ontology-guided inference, contextual policy synthesis, and decentralized negotiation to enable interpretable, resilient adaptation under uncertainty. We validate our approach in a simulated swarm optimization scenario involving 100 UAVs operating under communication failures. Compared to baseline methods, KAN-lite achieves faster recovery, higher throughput, and improved task robustness. These results highlight the value of structured semantic reasoning in enabling scalable and explainable autonomy for next-generation wireless systems.

Unmanned Aerial Vehicles (UAVs) face significant challenges in autonomous navigation due to their limited energy and computational resources. This paper introduces a knowledge-driven meta-learning framework specifically designed for ultra-fast path planning in lightweight UAVs. The proposed approach integrates domain-specific knowledge across three core domains-environment, network, and behavior-with visual data to enable adaptive learning from unlabeled data and rapid model retraining in various scenarios. To evaluate this framework, we created the Meta-UAV Optimal Path Dataset, a unique dataset tailored for complex, multi-domain path planning tasks. Additionally, a knowledge-driven loss function incorporating physics-based constraints ensures that the model's predictions align with real-world conditions. Experimental results demonstrate that our model achieves superior path efficiency, cross-domain adaptability, and lower resource consumption compared to traditional models, making it a suitable choice for real-world UAV applications.

The collision of ranging signals in the process of massive concurrent access ranging seriously affects the networking speed of satellite communication networks. Random access schemes are usually used to reduce the collision probability, but their effects are limited due to the limitation of ranging access rules and the network size. To further improve the performance of access ranging, the idea of optimal access ranging parameters is introduced in this article. A multiobjective anti-collision algorithm (MOACA) is proposed to optimize the concurrent access ranging process of high-capacity nodes of Internet of Things (IoT) to multifrequency time-division multiple-access (MF-TDMA) satellite communication systems. The model of the ranging measurement process is put forward and the expressions of ranging measurement time, collision probability, and the number of ranging channels are derived. On this basis, a multiobjective optimization problem (MOP) of the access ranging process is established. In MOACA, an ideal point database is formed based on the solutions of the MOP problem with typical inputs, and the negotiation curves of objectives are introduced to obtain the most appropriate ideal point from the database under certain requirements and system states. Besides, MOACA also contains a single-objective problem (SOP) for achieving the ideal point in the situation where no matching pattern can be found in the database. The simulation result shows that MOACA can improve the performance of the concurrent access ranging process.