Artificial intelligence (AI) plays a critical role in improving safety across modern transportation systems, enabling collision prediction, cooperative perception, autonomous navigation, and real-time traffic control. However, many AI-enabled safety functions rely on closed-box models that create a significant responsibility gap in which automated decisions cannot be explained or justified. This gap is especially concerning in the safety applications of autonomous vehicles (AVs) and smart public transit systems, where incorrect or unjustified AI decisions may result in severe accidents or loss of life. Traditional interpretability approaches provide limited post hoc explanations that are insufficient for transportation safety certification or incident analysis. This article proposes a causal AI-driven framework for responsible, auditable, and safety-aligned intelligence in transportation systems. Using structural causal models (SCMs), counterfactual reasoning, and causal intervention analysis, the proposed framework provides root-cause attribution, fairness assurance, and audit-ready explanations across land, maritime, and air transportation systems. Two representative use cases are presented: the causal evaluation of vehicle-to-everything (V2X)-assisted collision avoidance and the causal justification of autonomous driving policies. Experimental validation confirms that the proposed causal framework reduces collision rates by 80.6% and maintains safety even under severe communication latencies and sensor noise. Furthermore, compared to existing probabilistic causal approaches, our proposed method achieves an optimal balance between safety and travel efficiency while operating at submillisecond latency. These results demonstrate that causally grounded intelligence is essential for public safety, system reliability, and regulatory compliance in next-generation transportation systems.
Low-Earth-orbit (LEO) satellite constellations are poised to become a cornerstone of the 6G Internet of Things (IoT), providing truly global coverage and ubiquitous connectivity. This article presents a holistic 2D system architecture for 6G LEO satellite IoT that incorporates composition and functional perspectives to facilitate the seamless integration of LEO satellites and terrestrial networks. Building upon this architecture, we evaluate three pivotal enabling technologies that target uplink, downlink, and inter-satellite links (ISLs). Specifically, we analyze massive grant-free random access (GF-RA) for efficient uplink connectivity, investigate deep-learning (DL)-based multibeam precoding for robust downlink transmission, and examine distributed cooperative routing for resilient ISL data delivery. Furthermore, we present an on-orbit verification platform that validates the real-world feasibility and performance of the proposed solutions. Finally, we outline key open challenges and future research directions to guide the realization of future LEO satellite IoT.
Direct handset-to-satellite (DHTS) systems represent a key enabler of nonterrestrial networks (NTNs), allowing mass-market handheld devices to directly access low-Earth-orbit (LEO) satellites. This article provides a tutorial perspective on the architecture-aware roles and system-level challenges of signal detection in DHTS systems. We show that regenerative payload configuration, multisatellite cooperation, and control plane (CP)/user plane (UP) separation fundamentally reshape where and how detection is performed. Meanwhile, severe Doppler shifts, asynchronous multilink superposition and onboard processing constraints invalidate many terrestrial assumptions, motivating waveform-aware, structure-exploiting, and cooperative detection strategies. These challenges also drive the adoption of soft-output detection schemes to suppress error propagation in iterative decoding. Such emerging detection requirements motivate tighter integration with access control mechanisms and the development of onboard intelligence. Overall, the evolving role of detection highlights the need for scalable, reliable, and resource-efficient solutions in future DHTS deployments.
High-altitude platform stations (HAPSs) as international mobile telecommunications (IMT) base stations (HIBSs) have the advantages of wide coverage, flexible deployment, low latency, and low construction cost and can be used as a supplement and extension of the existing network to promote the realization of full coverage of the space-air-ground integrated network. However, in the context of limited low-frequency resources, promoting compatibility and coexistence between HIBSs and existing ground IMT networks to enhance system capacity and quality of service is challenging, and some issues should be addressed. Thus, this article focuses on studying and analyzing the compatibility of HIBSs and ground IMT systems in the 694-960-MHz ultrahigh frequency band and then showing coexistence conditions based on simulation results.
Driven by the rapid development of low-Earth orbit (LEO) satellite networks, machine learning (ML) model training and data processing face significant challenges owing to constrained communication bandwidth (BW) and privacy concerns. Federated learning (FL) has emerged as a promising solution for in situ data processing; however, centralized FL techniques are vulnerable to single-point failures and malicious attacks. To address these issues, we propose SpaceChain, a blockchain-enhanced FL framework tailored to LEO satellite networks. The SpaceChain framework includes a decentralized validation system for assessing local model updates. The system implements a proof-of-stake (PoS)-inspired consortium consensus mechanism in a permissioned blockchain system, thereby giving precedence to satellites based on cumulative learning contributions for validation, leading to a reduction of the energy expenditure when compared with a proof-of-work (PoW) protocol. FL maintains the privacy of data, with raw data stored on satellites, and the blockchain maintains tamper resistance in coordinating model updates. Satellites are assigned specific roles during each communication round, including ML nodes for local updates, consensus verifiers for validation, and blockchain stake validators for the final model integration. This role-based architecture ensures dynamic node participation and enhances the reliability and efficiency of the system. The simulation results demonstrate that the SpaceChain framework enhances the security, reliability, and performance of ML training in LEO networks. This article outlines the principles, architecture, and potential applications of SpaceChain, providing a foundation for secure and efficient ML training in next-generation satellite communication systems.
