With the development of Internet of Things (IoT) and Artificial Intelligence (AI) technologies, smart wearable devices (SWDs) are regarded as a promising approach for sports training monitoring. However, such monitoring generates a large number of computation-intensive tasks, which SWDs struggle to process in real-time due to limited battery capacity and computing power. Fortunately, the emergence of multi-access edge computing (MEC) offers an effective solution, allowing smart wearable devices to offload tasks to edge servers at the network edge for low-latency, energy-efficient processing. This paper investigates the dynamic task offloading problem in MEC-assisted wearable device systems. By jointly optimizing the offloading decisions, CPU frequency, offloading power of SWDs, and the CPU frequency of edge servers, we aim to minimize the total energy consumption of SWDs while maintaining the queue backlog. Through the Lyapunov drift-plus-penalty optimization framework, the long-term stochastic optimization problem is decoupled into a series of deterministic single-slot subproblems. A lightweight server selection algorithm is proposed to enable adaptive switching to alternative servers during overloads with negligible computational and signaling costs. Then, the problem is further decomposed into multiple sub-problems that can be solved in parallel. Based on this, we propose the Sports Training Monitoring (STM) algorithm to achieve efficient online solutions. Theoretical analysis and experiments indicate that STM can effectively reduce the energy consumption of SWDs while maintaining system performance.
Zhang et al. (Sun,) studied this question.