Dynamic Resource Allocation and Energy Consumption Optimization for Computational Tasks in Wearable Devices

This study addresses the constraints of computational power and energy sensitivity in wearable devices operating in edge computing environments by proposing a Lightweight Edge Resource Scheduling (LERS) algorithm. The LERS algorithm introduces a four-dimensional state-awareness mechanism—comprising device utilization, remaining battery level, temperature status, and task queue—to achieve accurate modeling of the devices dynamic states. Compared with Greedy and Heuristic methods, it demonstrates a stronger ability to respond to state variations. A lightweight strategy is employed to optimize the structure of the standard Deep Q-Network (DQN), significantly reducing computational overhead and making it more suitable for resource-constrained wearable device environments. Although Priority-DQN incorporates task priority, it fails to balance energy consumption control and temperature management, making it difficult to achieve multi-objective scheduling optimization under complex conditions. Experiments were conducted across six representative device states (S1–S6) and five offloading ratios (10%–90%). The results demonstrate that LERS achieves superior task completion rates (TCR) compared to both Heuristic and Priority-DQN methods. Under extreme conditions (S6) and high offloading scenarios (90%), LERS maintains a TCR of 92.0%, reduces energy consumption by 8.3 mWh compared to the Greedy approach, and lowers peak temperature by 4.5∘C relative to standard DQN, thereby significantly mitigating resource bottlenecks and overheating issues. The proposed method not only ensures task completion but also achieves coordinated optimization of energy consumption and thermal conditions, confirming the practical value of lightweight strategies in resource-constrained wearable devices.