Autonomous driving (AD) comes with inherent chal-lenges. Specifically, many of the actions taken by the autonomousvehicle are based on increasingly complex algorithms, mainlyapplied from the artificial intelligence (AI) domain such as deepneural networks (DNN). These algorithms are known for theirgreed of computing resources which usually is supported throughhigh performance computing units such as GPUs, FPGAs, ormulti-core CPUs. Such powerful resources are not well suitedfor embedded systems used in an autonomous vehicle due totheir cost and energy consumption. To overcome the lack of per-formance of embedded platforms, in this work we present a newmulti-level heterogeneous framework dedicated to AutonomousDriving (AD) vehicles. The proposed proposed framework makeseasy the design of AD architectures and is composed of threelevels : 1) Vehicle (or Edge), 2) Fog, and 3) Cloud.Index Terms—Intelligent Transport System (ITS), ArtificialIntelligence, Embedded System, Edge/Fog/Cloud Computing