基于Ray的服务于自动驾驶的远程分布式计算系统  

A Ray-Based Remote Distributed Computing System for Autonomous Driving

作　　者：邹星宇 文军[1] 陈波 

机构地区：[1]电子科技大学信息与软件工程学院,四川 成都 [2]智能终端四川省重点实验室,四川 宜宾

出　　处：《计算机科学与应用》2021年第5期1334-1340,共7页Computer Science and Application

摘　　要：随着自动驾驶技术的飞速发展,自动驾驶系统需要采集和处理更多的信息,车载的计算平台难以支撑高级别自动驾驶任务的算力需求。本文提出一种使用Ray框架来处理自动驾驶任务的远程分布式计算系统设计方案,克服单一计算平台的算力限制,为更高级别的自动驾驶提供支持。实验表明,Ray框架实现的分布式计算系统可以大幅减少自动驾驶任务的处理时间,提高系统吞吐量,保证自动驾驶任务的实时性。With the high development of autonomous driving technology, autonomous driving systems need to collect and process more information, and the onboard computing platform has been unable to support the computing requirements of high-level autonomous driving tasks. Thus, we propose a remote distributed computing scheme that uses the Ray framework to implement such kinds of tasks to break through the computing power limitation of a single computing platform, and provide the support for higher-level autonomous driving. Experiments show that the distributed computing system implemented by the Ray framework can greatly reduce the processing time of autonomous driving tasks, increase system throughput, and ensure the real-time performance of autonomous driving tasks.

关 键 词：深度学习 物联网 人工智能 自动驾驶 RAY 

分 类 号：U46[机械工程—车辆工程]