Core Concepts
Genie introduces a transparent caching technique in ROS to address latency issues in autonomous vehicles, enhancing performance and data quality significantly.
Abstract
Genie is a novel encapsulation technique that enables transparent caching in ROS, improving latency by 82% on average. It addresses key limitations of edge computing for autonomous vehicles and enhances object reusability and confidence in object maps.
The paper discusses the challenges of latency in autonomous vehicles due to SWaP constraints and proposes Genie as a solution. By leveraging edge servers equipped with GPUs, Genie improves computational efficiency and data reuse effectively. The distributed cache construction allows for collaborative caching among vehicles, enhancing information sharing and reusability.
Furthermore, the study evaluates Genie's performance in terms of tail latency, image reusability, object reusability, and confidence boost. Results show that Genie outperforms local and remote execution methods, providing substantial improvements across various scenarios. The case study on vision-assisted driving demonstrates the potential benefits of shared data among vehicles using Genie.
Overall, Genie presents a promising approach to address latency challenges in autonomous vehicles through innovative caching techniques and collaborative sensing.
Stats
"Genie can enhance the latency of Autoware Vision Detector by 82% on average."
"Object reusability reaches up to 67% for incoming requests."
"Confidence score measures the quality of gathered cache over time."