The space-air-ground integrated network (SAGIN) integrates satellites, unmanned aerial vehicles (UAVs), and terrestrial remote clouds to provide seamless network access and high-volume computing services for remote In...
详细信息
The space-air-ground integrated network (SAGIN) integrates satellites, unmanned aerial vehicles (UAVs), and terrestrial remote clouds to provide seamless network access and high-volume computing services for remote Internet of Things (IoT) devices, thus alleviating geographic and resource constraints. Existing methods typically focus on the network dynamics while overlooking the comprehensive consideration of device dynamics, namely, the time-varying task performance weights, task sizes, and task processing demands. Moreover, the centralized learning-based offloading schemes often lead to substantial signaling overhead. To bridge these gaps, this article proposes a distributed dynamic task offloading mechanism with game-theoretic multiagent stochastic learning (MASL). Technically, a stochastic game is formulated with each device as a player minimizing its weighted sum cost of latency and energy. We prove the existence of Nash equilibrium (NE) for our proposed game and propose a multiagent entropy-enhanced stochastic learning (MESL) algorithm in a fully distributed manner with no information exchange among IoT devices. By introducing the entropy of decision probability for each device, MESL increases decision dimensions, accelerates convergence, and facilitates optimal strategy achievement. Experimental results show that the MESL algorithm significantly reduces the overall cost and greatly enhances the convergence speed in dynamic SAGIN environments compared to existing algorithms.
With the widespread adoption of resource-intensive mobile applications, mobile edge computing (MEC) has emerged as a solution to enhance the computational power of mobile user equipments (UEs) and minimize their compu...
详细信息
ISBN:
(纸本)9798350329285
With the widespread adoption of resource-intensive mobile applications, mobile edge computing (MEC) has emerged as a solution to enhance the computational power of mobile user equipments (UEs) and minimize their computational delay by offloading tasks to edge servers (ESs). This paper delves into the computingoffloading challenge for multiple UEs in dynamic Internet of Things (IoT) networks with partial information-sharing. In such settings, the transmission bandwidth for each UE varies over time, and they can only access the historical data of their peers. Since UEs are self-interested in offloading computational tasks to ESs that possess limited computational resources, we model the UEs' offloading decision-making in this dynamic, privacy-bound scenario as a game. Subsequently, this game is further formulated as a multi-agent Partially Observable Markov Decision Process (POMDP). To address the POMDP and attain a near-optimal Nash equilibrium (NE) of the structured game, we introduce an algorithm grounded in multi-agent reinforcement learning, integrating Differentiable Neural Computer and Advantage Actor-Critic framework (abbreviated as DNA). Through this method, each UE autonomously decides the optimal computingoffloading strategy based on its game history, without obtaining the detailed offloading policies of other UEs. Experimental outcomes reveal that DNA surpasses the state-of-the-art benchmark methods by at least 8.3% in computingoffloading utilities and 3.98% in convergence rate, highlighting its effectiveness in a dynamic IoT environment with partial information-sharing between UEs.
暂无评论