CANCUN

計畫說明 計畫期程：2025.1 ~ 2027.12

(本計劃為團隊與法國 Orange Research Lab, 法國巴黎薩克雷大學中央理工學院, ZettaScale, 以及南巴黎電信學院之國際合作計畫。)

下一代行動網路應能實現工業物聯網系統(IIoT)的無縫運行，包括工業設備的時間 敏感控制、自主車輛的分散式感知/決策以及機器人的協同運作。本計畫的目的是 為無線通訊和IIoT應用提供一個整合控制框架。預期的貢獻包括:在網路方面，完 成QoS預測、5G和TSN整合以及網路資源重新配置和控制;在IIoT應用方面，完成 協同感知和分散式路徑和任務規劃。 CANCUN架構將從IIoT應用目標和中央控制 器提供的高品質任務規劃開始，推導出相應的網路配置/重新配置的規劃，預計將 可以有較大的機率滿足應用環境的需求。一旦系統服務開始，自主機器人/車輛開 始執行任務，無線網路就會重新配置，滿足QoS要求。改善傳輸容量和訊號覆蓋的 關鍵在於無線電頻寬資源分配、及無線網路介面和毫米波(mmWave)次系統的彈性 整合，而增強式學習(特別是多臂吃角子老虎機測試, Multi-Armed Bandit Testing)將是CANCUN中學習最佳網路配置的主要工具。然而，仍然有某些傳輸 容量需求和行動裝置的所在位置，無法確保應用系統要求的QoS服務 ，CANCUN計畫則會建議應用系統調整網路效能需求。這種調整取決於應用系統 ，可以對應於自主機器人/車輛的路徑修改和速度降低，以及協同感知的影像編碼 效率的修改。 CANCUN計畫的主要目標使用情境是自主低速車輛(Autonomous Low-Speed Vehicles, ALSV)的應用場景，如在工業區和校園等場域，提供人員 與 貨物的運送，在此類環境的交通基礎設施不足以及缺乏完善的交通信號系統，但 自動駕駛車輛仍然必須確保行車與人員安全。CANCUN計畫的解決方案將在巴黎 中央理工學院CentraleSupelec (CS)和台灣大學National Taiwan University (NTU) 使用相同的測試平台上分別進行理論研究和測試。選定的使用場景將由自主低速車 輛(ALSV)進行測試，並使用5G通信網路上協調路徑規劃並避免碰撞。

Project Description Project period: 2025.1 ~ 2027.12

(This is an international collaborative project, whose partners include Orange Research Lab, Centrale Supelec (University of Paris-Saclay), ZettaScale, and Telecom SudParis.)

The next generation of mobile networks should enable the seamless operation of Industrial IoT systems, including the time-sensitive control of machines, the distributed perception/decision of autonomous vehicles and the collaboration of robots. The aim of the project is to provide a joint control of the wireless network and the IIoT applications. The envisioned scientific contributions include, on the network side, QoS prediction, 5G and TSN integration and network reconfiguration and control and, on the IIoT applications side, collaborative perception and distributed path and task planning. CANCUN framework will start from the IIoT application goal and a high level task planning provided by a central controller, and deduce a corresponding network configuration/reconfiguration planning that is expected to achieve the requirements of the application with a large confidence. Once the tasks are launched and the mobile robots/vehicles start executing the tasks, the wireless network reconfigures and strives to achieve the stringent QoS requirements for the application. The levers for capacity and coverage provisioning will be the radio resource allocation and the flexible integration of wireless interfaces and mmWave small cells, while reinforcement learning (and in particular Multi-Armed Bandits) will be the privileged tool in CANCUN for learning the optimal network configuration. However, for some traffic intensities and localizations, ensuring QoS might not be possible and CANCUN project advocates application adaptation to the network condition. This adaptation depends on the application, and can correspond to a path modification and velocity reduction for the mobile robots/vehicles and a modification of the video coding rate for collaborative perception. The main targets of CANCUN are scenarios with autonomous low speed vehicles (ALSV), such as industrial areas and ports, where safety has to be ensured despite the harsh environment and the absence of traffic signaling. The solutions of CANCUN will be studied theoretically and tested on testbeds developed at CS and NTU. Selected schemes will be tested with autonomous low speed vehicles (ALSV) coordinated using 5G communications on path planning and collision avoidance.

Project progress summary