开发者Lakshay和Raveen在Hacker News上发布了名为Cerulion的开源项目,这是一个基于Rust编写的、兼容ROS 2的机器人运行时 1。该项目旨在解决ROS系统中进程间通信延迟高及难以调试的问题,通过低延迟IPC和多进程图实现了确定性的重放功能 1。Cerulion允许现有的ROS 2节点在不修改代码的情况下与原生Rust/Python代码并行运行,并提供了从运行、移动数据到捕获和诊断的完整数据循环工具链 1。
在性能方面,Cerulion使用iceoryx2实现共享内存,无需序列化和负载复制,使得1MiB的sensor_msgs/Image在两个进程间的往返延迟p50为4.08µs,而默认ROS 2 Jazzy版本的该指标为10.95 ms 1。该运行时支持Foxy, Humble, Jazzy和Lyrical多个版本的ROS 2 1。为了辅助调试,项目提供了cerulion bag play --resim all命令用于确定性重放验证,以及类似行车记录仪的“Flashback”功能,可保留约30秒日志并在崩溃时自动保存 1。项目源码托管于GitHub仓库:https://github.com/cerulion-inc/cerulion 1。
Developers Lakshay and Raveen have introduced Cerulion, an open-source robotics runtime written in Rust that is compatible with ROS 2 1. The project aims to address common challenges in ROS systems, such as high inter-process communication latency and debugging difficulties, by implementing a multi-process graph structure that enables deterministic replay capabilities 1.
A key performance highlight of the runtime is its ability to handle sensor_msgs/Image data at 1 MiB size with a round-trip latency of just 4.08 microseconds (p50) between two processes, significantly outperforming the default ROS 2 Jazzy version’s 10.95 milliseconds 1. This low-latency communication is achieved through the use of iceoryx2 for shared memory, which eliminates the need for serialization and payload copying during data transfer 1.
Cerulion allows existing ROS 2 nodes to run alongside native Rust or Python code without requiring any modifications to the original node source code 1. The runtime supports multiple versions of ROS 2, including Foxy, Humble, Jazzy, and Lyrical 1. To facilitate robust testing and verification, the toolchain provides a specific command, cerulion bag play --resim all, which allows users to perform deterministic replay validation 1. Additionally, the system includes a feature called "Flashback," which functions similarly to a dashcam; it maintains a rolling log of approximately 30 seconds of data and automatically saves this information if a crash occurs 1. The complete project, encompassing tools for running, moving data, capturing, and diagnosing, is available on GitHub at https://github.com/cerulion-inc/cerulion 1.
评论
还没有评论,欢迎留下第一条。