Abstract: As a new model,edge computing can effectively solve the problem of insufficient computing power of nearby user equipment.Due to the complex situation in the real world,the location distribution of users in each time period is difficult to be predicted,and the area covered by an edge server equipped with limited resources will be difficult to carry an uneven and uneven number of users in each time period,resulting in users being unable to request services.In addition,unreasonable allocation strategies will reduce the capacity of users in the area and may cause waste of resources.In response to the above problems,an adaptive mobile pathaware user allocation algorithm was proposed.The user's location information and road network data were used to determine the user's travel status through an improved map matching method,and predict a future travel path of the user.Based on the user's expected path,an allocation strategy was proposed to ensure the longer stable connection and less connection loss due to exceeding the signal range by using the expected stay time of server range as the adaptation value.A method for adjusting the allocation strategy based on best-fit was proposed.Through migrating some users from fully loaded servers to nearby servers with free space,the total user capacity of the servers in the area was indirectly increased,and the idleness time of adjacent servers was reduced,thereby the resource utilization was improved.Comparative experiments based on real user trajectory data sets showed that the proposed algorithm was significantly better than existing algorithms in terms of user coverage and resource utilization.

Key words: edge computing, adaptive mobile path-aware, user allocation, path prediction

摘要： 边缘计算作为一种新模式,可以有效解决附近移动设备运算能力不足的问题。然而,由于现实世界中复杂的状况,各个时间段内用户的位置分布难以预测,由配备有限资源的边缘服务器覆盖的区域内将难以承载各时间段内不均衡不恒定数量的用户,导致用户无法请求服务。另外,不合理的分配策略将降低区域内承载用户的能力,并可能造成资源浪费。针对上述问题,提出一种自适应移动路径感知的用户分配算法。该算法首先利用用户的位置信息和路网数据,通过改进的地图匹配方法确定用户行进状态,并对用户未来移动路径进行预期;然后,基于用户预期路径提出一种分配策略,以服务器范围内用户预期停留时长作为分配策略的适应值,确保用户拥有更久的稳定连接以及更少的因超出信号范围而丢失连接;最后,提出一种基于最佳适应(Best-Fit)的分配策略重构方法,通过将满载服务器中部分用户迁移至邻近有空闲空间的服务器,间接增加了区域内服务器的总用户容量,降低了周边服务器的空闲时间,从而提升了资源利用率。通过基于真实用户轨迹数据集的对比实验表明,本算法在用户覆盖率及资源利用率上明显优于现有算法。

关键词: 边缘计算, 自适应移动路径感知, 用户分配, 路径预测