AVEC: A Statistical Framework for Adaptive Vehicular Edge Data Cleaning

Abstract

In Vehicle-to-Vehicle and Vehicle-to-Infrastructure (V2X) communication, a large amount of data and information is transmitted over the air by the vehicles. If this data is captured, e.g., by a network of roadside units (RSUs) deployed at strategic locations, cleaned and processed, it may generate an interesting value. The process of cleaning the data involves the removal of data duplicates, as two or more RSUs may capture the same information from the same vehicle. Indeed, a vehicle can be located inside the communication range of multiple RSUs at the same time. The data cleaning process can be achieved through a centralized platform in the backend, where all the deployed RSUs connect and upload their collected data. To avoid overloading the backend, we propose to involve the RSUs in the cleaning process. Ideally, the RSU should be able to detect if any received information from a passing vehicle has not been also received by another nearby RSU. To achieve that, we use an adaptive probability-based splitting of the sensing range. Such a continuous process allows each RSU to adjust the probability distribution of the communication reliability after a sensing time window and to check parameters of neighbor nodes. Simulation results show the efficiency of our solution and demonstrate its ability to adapt with the network dynamicity, by adjusting the algorithm parameters, until reaching a good level of data cleaning compared to static and random approaches.