In the rapidly evolving and maturing field of robotics, computer simulation has become an invaluable tool in the design process. Webots, a state-of-the-art robotics simulator, is often the software of choice for robotics research. Even so, Webots simulations are often run on personal and lab computers. For projects that would benefit from an aggregated output dataset from thousands of simulation runs, there is no standard recourse; this project sets out to mitigate this by developing a formalized parallel pipeline for running sequences of Webots simulations on powerful HPC resources. Such a pipeline would allow researchers to generate massive datasets from their simulations, opening the door for potential machine learning applications and decision tool development. We have developed a pipeline capable of running Webots simulations both headlessly and in GUI-enabled mode over an SSH X11 server, with simulation execution occurring remotely on HPC compute nodes. Additionally, simulations can be run in sequence, with a batch job being distributed across an arbitrary number of computing nodes and each node having multiple instances running in parallel. The implemented distribution and parallelization are extremely effective, with a 100\% simulation completion rate after 12 hours of runs. Overall, this pipeline is very capable and can be used to extend existing projects or serve as a platform for new robotics simulation endeavors.
翻译:在机器人的迅速演变和成熟领域,计算机模拟已成为设计过程中一个宝贵的工具。Webot,即最先进的机器人模拟模拟器,往往是机器人研究的首选软件。即使如此,Webot模拟器经常在个人计算机和实验室计算机上运行。对于将从数千个模拟运行的综合产出数据集中受益的项目来说,没有标准追索方法;这个项目通过开发一个正式的平行管道,在强大的HPC资源上运行Webots模拟序列来缓解这一点。这样一条管道将使研究人员能够从其模拟中生成大量数据集,为潜在的机器学习应用和决策工具开发打开大门。我们开发了一个能够运行Webots无头和图形化模拟器的管道,在HPC的模拟运行中远程进行模拟执行。此外,模拟可以按顺序进行,将批量的工作分布在任意数量的计算节点和每个节点同时运行的多个实例中进行。执行的分布和平行化是极其有效的,能够运行Webotos的模拟器模拟器模拟,在现有的100小时后,可以运行一个全面的模拟平台,可以运行一个非常有效的模拟,在12小时后进行。