Research on Key Technologies and Platforms for Collaborative Intelligence Driven Autonomous Driving Vehicles
協同智能驅動自動駕駛關鍵技術與平台

Macao’s first Autonomous Driving Bus Testing Platform has launched in 2020, marking the start of a new chapter of intelligent transportation system in Macao. This research project is the key R&D project funded by the Science and Technology Development Fund (FDCT).  It has been working on creating a swarm intelligence based first-class vehicle platform and an experimental base of vehicle-road coordination for autonomous driving in the Guangdong-Hong Kong-Macao Greater Bay Area. The project is led by Prof. Xu Cheng-Zhong, Dean of the Faculty of Science and Technology of the University of Macau (UM) and his team and it is a collaborative effort between UM and leading institutions in mainland China, including Shenzhen Institutes of Advanced Technology under the Chinese Academy of Sciences, Baidu and Shenzhen Haylion Technologies. Desired outcomes and progress have been achieved since the start of the project.

At the level of key technologies, the team has achieved breakthroughs in robust autonomous driving against adversarial attacks, scene adaptability via transfer learning, highly-efficient inference for perception and decision models. These research outcomes have been published in top-tier AI conferences, such as CVPR, ICML, ICLR, and NeurIPS. The team participated into an international algorithm competition recently, namely “CVPR Security AI Challenger”, competed with total 1,681 teams around the world and won the 2nd prize. Heretofore, 22 journal papers, 19 conference papers, and 15 patents have been generated from this project. At the level of research and testing platform, the team successfully constructed the “Future Urban Intelligent Transportation Augmented Emulation Platform” through the independently developed digital twin technologies. This platform covers a wide spectrum of common scenarios and unusual ones, enabling the testing of autonomous driving algorithms, mixed operations of autonomous and manned driving, cooperative control, competitions of autonomous driving algorithms, etc. Besides, since the launch of MoCAD Testing Platform in October 2020, the autonomous bus has been constantly improving in terms of stability and safety, capable of responding to emergencies, avoiding risks by making lane change, slowing down, etc. The autonomous bus has been tested in multiple paths on UM campus, some of which are made available to the public. In the future, more testing paths will be opened at a regular basis.

At the level of industrialization, the team was invited to compete for the autonomous shuttle bus project in HK airport (currently under review). It was also planned to schedule a direct autonomous bus route from Macau to Hengqin, playing a leading demonstration role for the scientific development in Guangdong-Hong Kong-Macao Greater Bay Area. At the level of talent development, the project has nurtured more than 20 postdocs and doctoral students. To meet the needs of relevant talents, the Macao first autonomous driving course which held recently. More than 70 students completed the course which could provide a foundation for the future development of the industry and accelerate the development of a new generation of autonomous driving.

澳門大學自動駕駛測試平台於2020年10月正式啟動，為澳門智慧交通寫下新篇章。作為澳門特區科學技術發展基金的「重點研發專項資助計劃」之一，項目一直以打造粵港澳大灣區一流的自動駕駛群體智慧車載平台和車路協同自動駕駛試驗基地為目標。由澳大科技學院院長須成忠教授及其團隊進行研究，並結合澳大智慧城市物聯網國家重點實驗室以及內地眾多單位的科技力量，包括中科院深圳先進技術研究院、百度及深圳海梁科技。自項目啟動以來，皆取得不俗的成果及進展。

首先在關鍵核心技術上，項目組在自動駕駛中基於攻防對策的的魯棒性、基於遷移學習的場景適應性、感知和決策模型的高效推理等重要領域取得原創性的突破，成果獲發表於CVPR、ICML、ICLR、NeurIPS等人工智慧頂級會議。此外，在「CVPR安全AI挑戰者賽」上，與全球1681個團隊相互切磋並獲亞軍。目前已發表期刊論文22篇，會議論文19篇及專利15項。另一方面，在研發和測試平台上，項目運用自主研發的數字孿生技術成功搭建了「未來城市智慧交通增強模擬實驗平台」，此平台涵蓋了自動駕駛中的大部分常見及邊緣場景。除了能夠實現自動駕駛算法測試、非自動與自動駕駛的混合運行、協同控制還有自動駕駛算法挑戰競賽等功能。此外，項目啟動以來，經過不斷的改進及研發，巴士的穩定性和安全性皆獲得顯著提升，能夠更好地應對突發性事件，並進行變道、減速以避開風險。至今已於澳大校內多條路徑進行測試，其中部分開放給師生市民體驗，並將在未來定期開設試乘路線。

在科技成果產業化層面上，項目獲邀投標香港機場穿梭巴士專項，目前評標仍在進行中。同時，未來將計劃開設澳門橫琴無人直線巴士，為粵港澳大灣區科技發展做出引領示範作用。在人才培養層面上，項目吸引20多位博士和博士後參加。為滿足相關人才的需求 ，早前舉辦的本澳首屆自動駕課程反應熱烈，超過70名學員完成課程，期望能在自動駕駛領域積累人才，為產業發展提供紮實的基礎，加快推動新一代智慧交通發展。