We consider a discrete-time multi-channel network where the destination collects time-sensitive packets from multiple sources with sided channel information. The popular metric, Age of Information (AoI), is applied to measure the data freshness at the destination. Due to the interference constraint, only disjoint source-channel pairs can be chosen for transmission in each time slot, and the decision maker should choose the optimal scheduling pairs to minimize the average AoI at the destination. To learn the optimal channel selection, we apply the linear contextual bandit (LCB) framework by utilizing the sided information provided by pilots. Concretely, we establish the relationship between AoI regret and sub-optimal channel selection times and propose both age-independent and age-dependent algorithms. The former method is proven to achieve the sub-linear AoI regret but is outperformed by the latter algorithm both in the linear and non-linear contextual model in simulation.

翻译：我们考虑一个离散的多通道网络,即目的地从多个来源收集具有时间敏感性的包包,并附带频道信息。流行的衡量标准“信息年龄(AoI)”用于测量目的地的数据更新程度。由于干扰限制,只能选择不连接的源通道配对在每一个时段中传输,决策者应该选择最佳的排期配对,以尽量减少目的地的平均AoI。为了学习最佳的频道选择,我们利用飞行员提供的侧端信息来应用线性背景带(LCB)框架。具体地说,我们建立了AoI遗憾和次最佳频道选择时间之间的关系,并提出了独立年龄和基于年龄的算法。前一种方法已证明可以实现亚线性AoI遗憾,但后一种算法在线性和非线性背景模拟模型中都比后者差。