Microsoft announced a collaboration with Kaohsiung University of Science and Technology, Turing Drive (Taiwan Intelligent Driving), and ADLINK Technology to jointly build self-driving cars powered by the Azure AI service platform. The collaboration emphasizes a complete deployment environment and highly integrated development resources to reduce R&D labor costs and self-driving car model training time, thereby accelerating the development of self-driving cars.
In response to the global trend of digital transformation, Kaohsiung University of Science and Technology has focused on self-driving car research and development and cloud technology in recent years. This combines the fuel cell expertise of Kaohsiung University of Science and Technology's Director of Research, Guo Junxian, with the cloud computing resources of Microsoft's Azure AI service platform to promote unmanned vehicle research and development in Taiwan's academic community.
This collaboration uses Azure AI service platform computing resources, which are as follows:
• Create an efficient and user-friendly development environment through Azure Machine Learning:NHTU is placing ADLINK's open-source self-driving car community software, Autoware, on the Azure Machine Learning platform for testing and simulation, including controlling driving direction and trajectory. Azure Machine Learning provides various tools and services to accelerate model training. For example, automated labeling tools reduce labeling time by over 60%, allowing for rapid accumulation of training data sets.
In addition, by integrating Azure Machine Learning with MLOps capabilities, team members of more than 30 people can simultaneously adjust the same model, effectively reducing the thorny issues encountered in actual road conditions.
• Quickly identify obstacles with Custom Vision:Through Azure Cognitive Services' Custom Vision feature, Kaohsiung University of Science and Technology can quickly identify road images, accurately determine the location of obstacles, and safely avoid them.
• Efficient computing resources support 3D LiDAR sensors to increase driving safety:Placing 3D lidar sensors on Azure virtual machines, leveraging the scalable and flexible computing resources of Azure cloud services, supports real-time lidar sensor scanning of the surrounding environment, calculating distances to obstacles, and integrating 3D mapping, object detection, and various recognition applications. Furthermore, integration with Autoware's navigation and control modules allows for full vehicle system operation simulation, providing reliable surrounding object perception, navigation, and control systems to ensure the safety of self-driving vehicles.
In the past, the transition from development to road use for self-driving cars required overcoming complex data storage, value-added, operationalization, and testing processes, as well as integrating multiple disparate self-driving software packages. This resulted in a relatively high barrier to entry. However, the comprehensive services offered by the Microsoft Azure platform significantly reduce development time, achieving the advantages of a "single platform, easy development, and what you get is what you see" approach. Combining the cloud platform's efficient simulation and computing capabilities with self-driving vehicle technology, this approach lowers the learning curve and user experience through user-friendly services and user interfaces, thereby encouraging more students to participate, gain a deeper understanding of the cloud, and ultimately embrace it.
Integrating Turing Drive with ADLINK Technology's hardware and software demonstrates Taiwan's synergy in self-driving car development.
To improve the overall efficiency of self-driving car research and development, Kaohsiung University of Science and Technology not only introduced Microsoft Azure AI services during the self-driving car development process, but also cooperated with Turing Drive and ADLINK Technology. It hopes to leverage the R&D talents of the National Kaohsiung University of Science and Technology and the Microsoft Azure AI service platform, combined with the software and hardware strengths of Turing Drive and ADLINK Technology, to leverage the advantages of mutual cooperation.
Turing Drive is the first to use Microsoft Azure AI services to develop its self-driving car module. By leveraging Turing Drive's development experience in vehicle platforms such as small and medium-sized buses and slow-moving vehicles, as well as its self-driving R&D results in various scenarios such as closed, semi-open, and open urban roads, coupled with ADLINK's edge computing services and the Azure cloud platform's high integration, ease of use, high efficiency, and local support, the development time was significantly shortened, allowing the entire R&D process to be completed in just two months, delivering excellent results.
