Abstract Information and communication technology has many promising benefits including improvement the traffic network capacity, efficiency, and stability. However, to date, most of the improvements in signal management and interactions in connected vehicle environments focus solely on the vehicular side. This has led to a massive gap for non-motorized users and vulnerable road users. Specifically, deficit perception capability, inconsistent dissemination, obsolescent acquisition techniques, and ignorance of equality make the current experience of the active non-motorized users inconvenient and risky, especially for those with disabilities. To serve the users in an unbiased and automated way, a novel cooperated signal phase and timing (SPaT) services infrastructure — Vision Enhanced Non-motorized Users Services (VENUS) smart node is proposed. With customized up-to-date computer vision algorithms and artificial intelligence pipelines on the edge, VENUS smart node can collect necessary active-user information (including location, class, pose direction and mobility status), and generate directional crossing request for every pedestrian and cyclist in real time. Meanwhile, the improved communication system makes the VENUS node a reliable information hub to share the SPaT messages and carry interactions to/from the signal controller, connected vehicles and user personal information devices (i.e., cell phones, wearable devices) through various protocols. Based on extensive experimentation, 1076 testing users from six intersections, the VENUS sensing achieves 90.24% accuracy on directional-aware crossing trigger generation and 89.87% accuracy on mobility status estimation for normal users and four types of disabled persons. Furthermore, the VENUS smart node is fully compatible with the connected vehicles environment, and improves the signal system at low cost, mainly due to its flexibility and adaptability with existing infrastructure. The VENUS smart node is the first connected infrastructure architecture that integrates traffic sensing, data processing and information dissemination together for the self-operating indistinguishable signal services based on edge computing.

    Highlights A novel cooperated cyber–physical infrastructure system for non-motorized users in intersection areas, Vision Enhanced Non-motorized Users Services (VENUS) smart node is developed. The proposed solution can comprehensively support active SPaT services for pedestrians, cyclists and users with a variety of disability types via information collection and interaction. Exhaustive and precise non-motorized user perception pipelines based on edge artificial intelligence and computer vision technologies, including object recognition, user localization, pose direction and mobility status estimation, are integrated into the VENUS smart node workflow. A smart gateway is incorporated into VENUS that can handle bidirectional communications among controllers, Traffic Management Centers (TMCs), and cellphones (even without cellar services) through multiple protocols. Extensive experiments are conducted to evaluate VENUS. Among all 1076 testing users (including pedestrians, bikers, persons in a wheelchair, pedestrians pushing a person in a wheelchair, person using crutches, person using a walking frame), the VENUS sensor achieves 88.14% detection accuracy, 90.24% of directional touch-free crossing request generation, and 89.87% mobility status estimation.


    Zugriff

    Zugriff prüfen

    Verfügbarkeit in meiner Bibliothek prüfen

    Bestellung bei Subito €


    Exportieren, teilen und zitieren



    Titel :

    Cooperative traffic signal assistance system for non-motorized users and disabilities empowered by computer vision and edge artificial intelligence


    Beteiligte:
    Yang, Hao (Frank) (Autor:in) / Ling, Yifan (Autor:in) / Kopca, Cole (Autor:in) / Ricord, Sam (Autor:in) / Wang, Yinhai (Autor:in)


    Erscheinungsdatum :

    2022-09-14




    Medientyp :

    Aufsatz (Zeitschrift)


    Format :

    Elektronische Ressource


    Sprache :

    Englisch





    Edge Intelligence Empowered Urban Traffic Monitoring: A Network Tomography Perspective

    Pan, Shengli / Li, Peng / Yi, Changsheng et al. | IEEE | 2021




    Artificial Intelligence Empowered Models for UAV Communications

    Pradhan, Nilanjana / Sille, Roohi / Sagar, Shrddha | Springer Verlag | 2022