Deep Reinforcement Learning-based Data Offloading Strategies For Wireless Body Area Networks

Wireless Body Area Network(WBAN)is widely used in medical services to provide remote real-time and continuous medical monitoring.With the increasing amount of sensor data,WBAN is limited by its limited storage and computing capacity,resulting in a significant decrease in efficiency and reliability.By placing computing close to the terminal,Mobile Edge Computing(MEC)technology can swiftly respond to user service requests,effectively reducing task processing latency and enhancing user service.In order to address the computational demands of WBAN applications,WBAN,a computing platform with constrained processing power,can be integrated with edge computing.In the actual application of MEC-based WBAN task computing,there are some challenges,the most important of which is how to formulate the optimal computing offloading strategy.This thesis investigates the task offloading decision-making problem in edge networks using deep reinforcement learning.The following is a summary of the key research findings:(1)Research on the joint optimization of computational offloading and resource allocation(JCORA)in MEC under the medical service scenario.JCORA is described as a Markov Decision Process(MDP),and a WBAN offloading strategy research based on deep deterministic policy gradient(DDPG-based WBAN Data Offloading Strategy,DDPG-WOS)is proposed to optimize the overall system delay and energy consumption in the long term.The scheme uses MEC to alleviate the computing pressure of a single WBAN and improve transmission capacity.DDPG-WOS considers factors such as channel conditions,transmission quality,computing power,and energy consumption,optimizing the process of formulating offloading strategies.The suggested approach can successfully reduce the overall system delay and terminal energy consumption,according to simulation findings.(2)Based on the research content of(1),further consider a drawback in the offloading decision-making scheme,which is that the reward value is only considered for the current and next states,which may lead to an incorrect estimation of the optimal value of the actual reward function,thus affecting WBAN’s optimal offloading decision-making.Therefore,this thesis proposes a multi-step deep deterministic policy gradient(MSDDPG)algorithm,which considers the reward values of multiple states comprehensively.The MSDDPG algorithm can more accurately estimate the value of the actual reward function,thereby formulating better offloading strategies.In summary,this thesis mainly studies the application of mobile edge computing technology in the medical service scenario,to solve the storage and computing power limitations faced by WBAN in data processing.It has important practical significance for building low-latency,low-energy,and high-reliability medical service networks.