SKT가 글로벌 모바일 AP 제조사 미디어텍, AI 경량화 전문 스타트업 노타와 협력해 온디바이스 AI 기반 단말 소비전력 절감과 관련된 초기 단계 기술 개발에 성공했다고 밝혔다.
Research on connection optimization technology during terminal-base station transmission and reception
Predicting and controlling on-device AI-based traffic occurrence
Lightweight AI model related to terminal power efficiency… Real modem implementation
SKT has begun to advance its technology to improve smartphone power consumption issues.
SKT announced that it has successfully developed early-stage technology related to reducing terminal power consumption based on on-device AI in collaboration with global mobile AP manufacturer MediaTek and AI lightweight startup Nota.
The technology developed by these three companies aims to minimize the power used to maintain a connection with a base station for a certain period of time after transmission and reception is complete when a terminal exchanges data with the base station.
SKT said that it has been researching technology for the past several years to optimally control the connection status between the terminal and the base station by predicting whether traffic will occur at a certain time on the terminal itself in order to achieve power savings due to the recent increase in constant smartphone usage.
In particular, we reported that smartphones, which are the source of user data, are more suitable for traffic prediction than the server side, and conducted joint research with a startup (Nota) and a global mobile AP manufacturer (MediaTek) to implement the technology in the form of on-device AI.
SKT, Nota, and MediaTek have selected an AI model suitable for terminal power efficiency and are working on low-latency and low-power smartphones. They revealed that they studied the lightweighting of AI models by utilizing techniques such as quantization and pruning so that they could operate.
He also said that he implemented a lightweight AI model into a modem mounted on an actual smartphone and verified its real-time operation, and secured a certain level of data generation prediction accuracy.
SKT said, “This research is a case of preemptive research and development in the communications field for on-device AI, which has recently become a hot topic, and is significant in that it can substantially improve battery life if combined with manufacturers’ efforts to increase battery capacity and reduce power consumption of components in terms of hardware.”
SKT plans to disclose specific research and development results at MWC 2024, which will be held in Barcelona at the end of this month, and plans to further enhance technological perfection through advanced AI models and network linkage tests in the future.
Meanwhile, in addition to this research, SKT is actively promoting the advancement of AI-based communication infrastructure in preparation for the evolution of 5G and 6G. Representative examples include Smart Paging technology for core networks, Link Adaptation technology for wireless networks, and virtualized base station power consumption reduction technology.
Ryu Tak-ki, head of SK Telecom’s infrastructure technology division, said, “This is a meaningful achievement that has confirmed the potential for the evolution of on-device AI-based communication technology,” and added, “We plan to accelerate our research and development to apply AI to each area of mobile communications to improve customer experience going forward.”