Time-efficient route planning is a significant research area of Internet of Vehicular Things. Optimal route selection is important to reach the destination in minimal time. Further, energy efficiency is vital for route planning in a sustainable environment. To address these issues, this paper proposes a federated learning and genetic algorithm-based green edge computing framework for optimal route planning in Internet of Vehicular Things. The vehicles are connected to the road side unit. The road side unit processes the image and video of the road, and predicts the number of vehicles on the road. For video processing Region-based Convolutional Neural Network is used. The road side units send the result and the local model parameters to the regional server. The regional server determines the optimal route using modified genetic algorithm, and sends it to the vehicles and the cloud. Also, the regional server updates its model and sends the updated model parameters to the road side units. The road side units update their local models accordingly. The regional server also sends the model parameters to the cloud, and the cloud updates the global model. The cloud sends the updated model parameters to the regional servers. The regional servers update their models accordingly. The results present that above 90% accuracy is achieved by the proposed model. The results also present that using modified GA the proposed approach reduces time and power consumption to find the optimal route by ∼62% and ∼66% than the cloud-only model.

中文翻译：

---

FedGen：基于联合学习的绿色边缘计算，在车联网中使用遗传算法实现最佳路线选择


高效的路径规划是车联网的一个重要研究领域。最佳路线选择对于在最短的时间内到达目的地非常重要。此外，能源效率对于可持续环境中的路线规划至关重要。为了解决这些问题，本文提出了一种基于联邦学习和遗传算法的绿色边缘计算框架，用于车联网中的最优路线规划。车辆连接到路边单元。路侧单元处理道路的图像和视频，并预测道路上的车辆数量。对于视频处理，使用基于区域的卷积神经网络。路侧单元将结果和本地模型参数发送至区域服务器。区域服务器使用改进的遗传算法确定最佳路线，并将其发送到车辆和云端。此外，区域服务器更新其模型并将更新的模型参数发送到路侧单元。路边单位相应更新本地模型。区域服务器还将模型参数发送到云端，云端更新全局模型。云端将更新后的模型参数发送到区域服务器。区域服务器相应地更新其模型。结果表明，该模型的准确率达到 90% 以上。结果还表明，使用改进的遗传算法，所提出的方法比纯云模型减少了寻找最佳路线的时间和功耗约 62% 和约 66%。