Collaborative Edge Intelligence for Autonomous Vehicles: Opportunities and Challenges

In this paper, we introduce collaborative edge intelligence (CEI), a novel distributed computing paradigm, to support ultra-low latency and large-scale deployment of autonomous vehicles. Existing computing paradigms are inadequate in supporting autonomous vehicles effectively. Specifically, vehicle-based and edge-based technologies are limited by the computational capacities of individual devices, making it challenging to handle resource-greedy AI models and massive sensing data. Cloud-based approaches, while resource-abundant, suffer from unstable communication links. The cloud-edge collaborative method often neglects the collaboration among edge devices, resulting in limited scalability, and may still suffer from unpredictable cloud-edge communication. Collaborative edge intelligence is an emerging paradigm where geo-distributed computing nodes collaborate to form a federated resource pool, collectively providing computational services. CEI not only facilitates vertical collaboration among cloud, edge, and end devices but also promotes horizontal collaboration among edge devices. It has outstanding benefits, including low-latency data processing, enlarged resource pools, and expanded service regions. In this paper, we first briefly review existing computing paradigms. We then clarify the concept of collaborative edge intelligence and its benefits. Finally, we discuss the opportunities and challenges of CEI.