Lantu Automotive Unveils New Smart Driving Platform at the 2025 Advanced Driver Assistance Conference

Lantu Automotive Technology Co., Ltd., a high-end smart new energy brand under Dongfeng Motor, is committed to producing high-quality and safe products while exploring innovative models for autonomous business. At the 2025 8th Advanced Driver Assistance Conference, Liu Huikai, the software development expert for driver assistance and perception development director at Lantu, focused on the Lantu Kunpeng Smart Driving Platform and its end-to-end practices, highlighting technological innovations and achievements in driver assistance. The aim is to enhance user experience through multimodal large models and intelligent safety strategies. Liu stated that the Kunpeng Smart Driving Platform, through end-to-end technology and safety strategy models, guarantees both upper and lower safety limits for the driver assistance system, facilitating a transformation in the industry from highly focused driving to occasional supervision. Since the brand's official logo launch on July 17, 2020, Lantu has rapidly completed its strategic layout of 'three categories in three years, four models in four years,' covering a full product matrix across MPVs, SUVs, and sedans. The Dreamer MPV has consecutively won the title of best-selling new energy MPV, while the FREE SUV and Zhiyin SUV have received accolades for their innovative designs and performance. The products incorporate advanced safety features, including a ten-layer protective system, ensuring a solid foundation for the brand's high-end development. The Kunpeng Smart Driving Platform, built on the Tianyuan Intelligent Architecture, encompasses three modules: the Qingyun Platform, the Kunpeng Smart Driving, and the Xiaoyao Cabin. It aims to achieve a user experience characterized by 'better driving, easier long-distance travel, and no blind spots in visibility.' The platform is designed for four generations of product evolution. Lantu's end-to-end large model technology addresses the challenges of traditional driver assistance systems in dealing with out-of-distribution (OOD) unknown scenarios. The technology employs a VLA multimodal architecture to break through scene cognition boundaries and enhance safety decision-making. The system has demonstrated capabilities to interpret complex traffic signs, unusual vehicle states, and hand gestures, enabling conditional L3 driver assistance that allows for hands-free and eyes-free operation. Looking ahead, Lantu aims to continue optimizing the World Model for virtual data generation and the visual CoT for spatiotemporal modeling, collaborating with the industry to tackle challenges in multimodal data costs and long-tail scene generalization to enhance the global competitiveness of China's driver assistance technology.