ROS Resources: Documentation | Support | Discussion Forum | Service Status | Q&A answers.ros.org

Bandwidth for tele-op

Hi, there was some discussion recently about bandwidth requirements for tele-op. This has been a topic of interest for our customers as well. I thought I’d share more information and some demo videos. You can check out my blog at https://blog.bthere.ai/bandwidth-for-tele-op/. I’d love to hear questions and observations from the community here too.

In the blog I included demo videos to show generally what tele-op video looks like at various bitrates, resolutions, and framerates. At bThere we generally try to stream video at 1 Mbit because it provides a good image at a good framerate. It’s not as good as what we see on TV or on Netflix, but it works well for driving robots.

We’ve also been able to reduce the bitrate much lower in order to work well on poor network connections. Lower bitrates can also reduce cell connection expense. The downside of low bitrates is lower fidelity on what the operator is seeing as well as higher glass-to-glass latency. Robots must have safety measures built in with sensors and local logic to avoid damaging collisions, but better video still makes it easier for a remote operator to drive a robot safely.

Another interesting aspect is variation of bitrate depending on scenario. A dashboard showing thumbnail videos from a dozen robots can (and should) use low resolution and low bitrate. However, when remotely driving one robot in a tight environment it is better to increase the bitrate and get better video. Additionally, having multiple cameras providing more environmental awareness is super helpful.

Happy to field any questions from folks. I’d love to hear from people about what works well (or badly) for them and their robots.

Stuart

3 Likes

As an example on how to reduce the consumed bandwidth, I saw cases where secondary cameras frames were minimized and put on top of the main video frame, this way sending a single “merged” frame instead of multiple.

1 Like

Definitely. Picture in picture can reduce the number of streams at the expense of obscuring part of the main image. It’s a bit like a rear view mirror in a car. But you can also turn it on and off as desired.