In this paper, we propose a novel memory-enabled non-uniform sampling-based bumblebee foraging algorithm (MEB) designed for optimal channel selection in a distributed Vehicular Dynamic Spectrum Access (VDSA) framework employed in a platoon operating environment. Given how bumblebee behavioral models are designed to support adaptation in complex and highly time-varying environments, these models can be employed by connected vehicles to enable their operation within a dynamically changing network topology and support their selection of optimal channels possessing low levels of congestion to achieve high throughput. As a result, the proposed VDSA-based optimal channel selection employs fundamental concepts from the bumblebee foraging model. In the proposed approach, the Channel Busy Ratio (CBR) of all channels is computed and stored in memory to be accessed by the MEB algorithm to make the necessary channel switching decisions. Two averaging techniques, Sliding Window Average (SWA) and Exponentially Weighted Moving Average (EWMA), are employed to leverage past samples and are evaluated against the no-memory case. Due to the high variability of the environment (e.g., high velocities, changing density of vehicles on the road), we propose to calculate the CBR by employing non-uniform channel sampling allocations as well as evaluate it using both simplified numerical and realistic Vehicle-to-Vehicle (V2V) computer simulations. The numerical simulation results show that gains in the probability of the best channel selection can be achieved relative to a uniform sampling allocation approach. By utilizing memory, we observe an additional increase in the channel selection performance. Similarly, we see an increase in the probability of successful reception when utilizing the bumblebee algorithm via a system-level simulator.

翻译：本文提出了一种新颖的、基于记忆的非均匀采样蜂群觅食算法（MEB），用于在编队行驶环境中实现分布式车联网动态频谱接入（VDSA）框架下的最优信道选择。鉴于蜂群行为模型旨在支持复杂且高度时变环境中的自适应能力，这些模型可被网联车辆采用，使其能够在动态变化的网络拓扑中运行，并支持其选择低拥塞程度的最优信道以实现高吞吐量。因此，所提出的基于VDSA的最优信道选择方法借鉴了蜂群觅食模型的基本原理。在本文方法中，计算所有信道的信道繁忙率（CBR）并存储于记忆中，供MEB算法在做出必要信道切换决策时访问。本文采用滑动窗口平均（SWA）和指数加权移动平均（EWMA）两种平均技术来利用历史样本数据，并与无记忆情形进行了对比评估。由于环境的高度可变性（例如，车辆高速运动、道路上车辆密度不断变化），我们提出通过采用非均匀信道采样分配来计算CBR，并分别通过简化数值仿真和逼真的车对车（V2V）计算机仿真进行评估。数值仿真结果表明，与均匀采样分配方法相比，该方法在最优信道选择概率上可取得增益。通过利用记忆机制，我们观察到信道选择性能进一步提升。类似地，通过系统级仿真器采用蜂群算法时，我们观察到成功接收概率的增加。