This perspective paper delves into the concept of foundation intelligence that shapes the future of smart infrastructure services as the transportation sector transitions into the era of Transportation 5.0. First, the discussion focuses on a suite of emerging technologies essential for foundation intelligence. These technologies encompass digital twinning, parallel intelligence, large vision-language models, traffic simulation and transportation systems modeling, vehicle-to-everything (V2X) connectivity, and decentralized/distributed systems. Next, the paper introduces the present landscape of Transportation 5.0 applications as illuminated by the foundational intelligence, and casts a vision towards the future including cooperative driving automation, smart intersection/infrastructure, parallel traffic management, virtual drivers, and mobility systems planning and operations, laying out prospects that are poised to redefine the mobility ecosystem. Last, through a comprehensive outlook, this paper aspires to offer a guiding framework for the intelligent evolution in data generation and model calibration, digital twinning and simulation, scenario development and experimentation, feedback loop for management and control, and continuous learning and adaptation, fostering safety, efficiency, reliability, and sustainability in the future smart transportation infrastructure.

Autonomous driving (AD) and Cooperative Driving Automation (CDA) hold great promise for transforming mobility. However, current off-the-shelf AD or CDA platforms such as Autoware, Apollo, and CARMA, are subject to gaps between simulation and the real world and do not offer integrated pipelines for CDA research, development, and deployment. In this letter, we conceptualize OpenCDA-ROS, building on the strengths of an open-source framework OpenCDA and the Robot Operating System (ROS), to seamlessly synthesize ROS's real-world deployment capabilities with OpenCDA's mature CDA research framework and simulation-based evaluation to fill the gaps aforementioned. OpenCDA-ROS will leverage the advantages of both ROS and OpenCDA to boost the prototyping and deployment of critical CDA features in both simulation and the real world, particularly for cooperative perception, mapping and digital twinning, cooperative decision-making and motion planning, and smart infrastructure services. By offering seamless integration of simulation and real-world CDA, OpenCDA-ROS contributes significantly to advancing fundamental research, development, testing, validation, prototyping, and deployment of autonomous driving and CDA.

In the era of future mobility within Transportation 5.0, autonomy and cooperation across all road users and smart infrastructure stand as the key features to enhance transportation safety, efficiency, and sustainability, supported by cooperative perception, decision-making and planning, and control. An accurate and robust localization system plays a vital role in enabling these modules for future mobility and is constrained by environmental uncertainties and sensing limitations. To achieve precise and resilient localization in this new era, this paper introduces emerging technologies including edge computing, hybrid data-driven and physical model approaches, foundation models as well as parallel intelligence, that are beneficial for next-generation localization systems. On top of these key technologies, by integrating real-world testing and digital twin technology, we further put forward a Decentralized Autonomous Service (DAS)-based cooperative localization framework for future mobility systems to enhance the resilience, robustness, and safety of transportation systems.

In the field of autonomous driving, camera-based perception models are mostly trained on clear weather data. Models designed to handle specific weather conditions often lack generalization to dynamically changing environments and primarily focus on weather removal rather than robust perception. This paper proposes a semantic-enabled network for object detection under diverse weather conditions. Semantic information enables the model to generate plausible content in missing regions and accurately delineate object boundaries. It also preserves visual coherence and realism across both restored and original image areas, thereby facilitating image transformation and object recognition. Specifically, our architecture consists of a Preprocessing Unit (PPU) and a Detection Unit (DTU), where the PPU utilizes a U-shaped network enriched with semantics to refine degraded images, and the DTU integrates this semantic information for object detection using a modified YOLO network. Extensive experiments demonstrate that the proposed method achieves mAP improvements ranging from 1.49% to 8.78% compared with existing approaches across multiple benchmark datasets under various weather conditions. These results demonstrate the effectiveness of semantic guidance in image enhancement and object detection, providing a comprehensive framework for improving detection performance. The source code will be made publicly available.