Adaptive Edge-to-Cloud Orchestration Pipelines for Autonomous Vehicle Intelligence

Abstract

The increasing complexity of autonomous vehicle systems demands robust data processing pipelines capable of handling
massive, heterogeneous data streams generated in real-time by onboard sensors and external infrastructure. This paper
proposes an adaptive edge-to-cloud orchestration pipeline designed to optimize the processing and analysis of autonomous
vehicle intelligence. By dynamically distributing computation tasks between edge nodes (such as vehicles and roadside
units) and cloud servers, the pipeline balances latency requirements, computational resource constraints, and bandwidth
limitations.
The proposed pipeline leverages containerized microservices orchestrated via Kubernetes, integrated with AI models for
perception, prediction, and decision-making tasks relevant to autonomous driving. Adaptive orchestration algorithms
monitor system performance metrics, including network conditions and workload, to adjust task placement and resource
allocation continuously. This approach ensures low-latency inference for safety-critical operations while harnessing the
cloud’s scalability for complex analytics and long-term learning.
Evaluation using real-world autonomous driving datasets and simulated urban scenarios demonstrates that the adaptive
orchestration pipeline reduces end-to-end latency by up to 35% compared to static cloud-only or edge-only deployments.
It also improves resource utilization efficiency by dynamically scaling edge and cloud resources based on demand.
The pipeline supports heterogeneous hardware and networking environments, enhancing its applicability in diverse
autonomous vehicle ecosystems.
This research contributes to the evolution of intelligent transportation systems by providing a flexible, efficient, and
scalable solution for distributed data processing and AI inference in autonomous vehicles. Future work will explore the
integration of federated learning for privacy preservation and the incorporation of 5G/6G connectivity to further enhance
adaptive orchestration.