Building a V2X Simulation Framework for Future Autonomous Driving

Collecting surrounding vehicles' motion information is one of the key issues for accident prevention and autonomous driving. Although multi-vehicle simulation frameworks are widely provided, We need a platform that enable inter-vehicle V2X communications. In this work, based on the open source simulation platform, CARLA, we extend and implement several modules to build a V2X simulation framework. In the proposed framework, vehicles are allowed to share their profiles and sensory data through V2X communications. With the motion information of other vehicles, a car can thus make more intelligent decisions. To validate the effectiveness of the framework, we run simulations in variose scenarios. Each time, a primary vehicle is selected and then both its sensory data and received surrounding vehicles' information are output and recorded in a simulated dataset. It is shown that with the dataset and our multi-vehicle data fusion algorithm, the primary vehicle can visually see the driving status of surrounding cars, which can greatly help a vehicle to choose a better driving strategy. This work not only proposes a V2X communication-enabled multi-vehicle simulation framework based on CARLA, but also provides a low cost way to generate simulated V2X datasets.