FogROS brings robotic cloud computing to a robot operating system

FogROS brings robotic cloud computing to a robot operating system

On my recent trip to the Bay Area, I took a few hours to visit the Berkeley Lab for Artificial Intelligence Research (BAIR). Professor Ken Goldberg took me around the lab and introduced me to a couple of projects the students were working on. Fougeres immediately caught my attention – and it’s not just because it has a similar name to French cuisine that it is a problem.

Image credits: Unlock the bots

The offering arrives as part of the latest release of the open source robotic operating system, ROS 2 Humble Hawksbill – ROS 2 v8. In short, it offers a way to offload automated tasks to a remote server, using a cloud computing platform such as Amazon Web Services. Advances in server-side computing that have made things like cloud gaming possible with minimal latency can also be applied to bot operations.

“Robots are often limited in their onboard computing capabilities due to weight and power requirements,” Jeff Ichnowski, a postdoctoral student at Berkeley who led the project, told TechCrunch. They also rarely have hardware accelerators such as GPU, TPUS or FPGA. But many bot algorithms and recent advances (such as deep learning) take advantage of high-end computers and hardware accelerators. We envision that using cloud computing to speed up slow computations could enable bots to do more things in the same amount of time.”

The platform announced today as part of the new version of ROS is actually FogROS 2. The first version, introduced last summer, was an early proof of concept. In March, the teams quietly previewed FogROS 2 available through GitHub, and today it’s rolling out to everyone, with a number of improvements designed to improve cloud-based performance.

Image credits: Russians

Much like playing Xbox games on a smartphone, the basic principle here is to provide a way to perform complex tasks on a bot that does not require equally complex on-board processing. If you can complete the task via a remote server, you can save on size, weight, and – perhaps most importantly – cost. The team notes in a recently published research paper:

We show in application examples that the performance gained using cloud computers can beat network latency to dramatically speed up bot performance. For example, FogROS 2 reduces SLAM latency by 50%, reduces planning time from 14 seconds to 1.2 seconds, and speeds up motion planning by 28x. When compared to the alternatives, FogROS 2 reduces network usage by up to 3.8x.

Goldberg notes that such a platform could open up more possibilities for bots than those listed above. “Other computing-intensive tasks for bots such as stochastic planning and facilitating supervised and unsupervised deep learning can benefit tasks from multiple bots.”

Future versions of the software could open things up for additional platforms, including Google Cloud and Azure. The team notes:

In future work, we will continue to add support for additional cloud computing providers and services. We’ll explore additional models of computing, such as no-server, positional states, and more. We’ll also explore extending the networking capabilities of FogROS 2 to allow bots to communicate, collaborate, and share data more easily.

Humble Hawskbill includes a number of additional features and a number of extras outside of the cloud computing platform. Per Open Robotics, tasked with maintaining ROS, updates include,

“The robot operating system debuted at ICRA thirteen years ago this month and Open Robotics celebrated its tenth anniversary in March,” the Open Robotics CEO said in a statement linked to the news, so the release of ROS 2 Humble Hawksbill is a perfect opportunity to thank The global community of thousands of developers and millions of users who contribute to the platform and work to improve it.”

2022-05-23 15:01:33

Leave a Comment

Your email address will not be published.