The 6G team at Samsung R&D Institute China - Beijing (SRC-B) has achieved distinction by emerging as the champion in the AI+wireless communications track of the renowned National Artificial Intelligence Competition (NAIC). NAIC, a globally recognized event focused on forefront AI research and industry challenges, takes place in China and receives support from the Shenzhen Municipal People’s Government and Peng Cheng Laboratory. With the participation of over 7,293 teams from leading enterprises, academia, and scientific research institutions, the competition serves as a formidable arena where participants compete for the coveted championship.

The AI+wireless communications track posed a significant challenge, requiring participants to employ AI technologies for vacuum overhead communication—an essential focus in future wireless networks. The SRC-B team designed a systematic deep learning solution featuring a lightweight model architecture, a wireless-awareness loss function, a location-specific pretrain-finetune strategy, a target-oriented data extraction solution, and a multimodel cooperative training mechanism. After four rigorous competition rounds, this innovative solution emerged as the champion, earning high commendation from the competition’s technical committee, which comprises esteemed academicians from the Chinese Academy of Sciences and the Chinese Academy of Engineering.

SRC-B’s proposed solution delivers a stable model capable of obtaining super-stable precoding matrices in highly dynamic environments, excelling in optimizing spectrum efficiency and outperforming competing methods. In the absence of channel information feedback, this method globally optimizes spectrum efficiency by orchestrating cooperative functions such as beamforming, power allocation, coherent transmission, and interference cancellation. Effective in feedback-free communication, this method extends its applicability to various scenarios, including 5G/6G fixed wireless access (FWA) scenes, semi-static Internet of Things (IoT) scenes, integrated sensing and communication (ISAC) scenes, and coordinated multipoint (CoMP) scenes, among others.

In addition, SRC-B’s team, consisting of similar members, secured 3rd place in the cell-free scheduler track at the 6GANA Network AI Competition—an international arena addressing cutting-edge issues in network AI. This track, centered on 6G typical network scenes with massive access in a cell-free scenario, witnessed SRC-B’s deep reinforcement learning solution with an input length-insensitive encoder-decoder structure gaining prominence. Beyond delivering close-to-optimal performance, this solution’s inherent scalability allows for seamless adaptation between inference time and performance—a crucial advantage for practical implementation.

Drawing on a rich legacy in wireless communication and AI, SRC-B’s extensive history of research and development has produced numerous AI-based solutions for wireless communication, earning multiple championship titles in international competitions. Encompassing the entire communication system stack, from radio frequency and physical layer design to network optimization and source coding, SRC-B remains poised to advance the wireless industry through ongoing participation and groundbreaking contributions to the intersection of wireless communication and AI.

The members of the Samsung Team at Wireless Communication Competitions