ROS Resources: Documentation | Support | Discussion Forum | Service Status | Q&A

Robot As a Service, anyone?


We’ve gotten interested in collaborating with our University Library to create a “robot as a service” facility. I am in the process of writing a grant and was curious whether anyone here had done something like that who would want to exchange ideas.

p.s. I’ve found several related papers which I would attach here if only I knew how.


Can you define “robot as a service”? E.g., are you referring to situations where someone can get ephemeral and anonymous access to a remote robot, in the style of VM instances in cloud computing?


you might not be permitted to attached PDFs (security risk), but you can list citations in plaintext.

I started a list of related work on this topic at


Yes. Its kind of like the idea of a virtual remote laboratory, but the lab contains a set of (physical) robots that can be scheduled remotely and used to try different things. Initially they wouldn’t ever leave the lab. Also initially you would have the choice of one hardware configuration. The starting point can be thought more of as a teaching and experimentation environment. At least that’s how we’re thinking.

Pito Salas
Brandeis Computer Science
Volen 134



I am starting a company that provides exactly this service ( We have not made any big announcements yet, but here are two cases we are targeting that might align with your interests:

  1. physical robots in a well documented workspace to which users get remote access, but to which they (users) can never physically go themselves.

  2. easy sharing of experimental robots from one lab to another, peer-to-peer, where the documentation depends on what users (peers) provide.

The main goal of #1 is to support reproducible research. The environments are “well documented”, i.e., you can build them yourself if you do not want to use the remote access service. The case of #1 includes both live interaction and non-interactive experiments. We have homing code that returns the robots to a set of known initial states at the end of each instantiation, which supports reproducibility.

The main goal of #2 is to facilitate remotely sharing of heterogeneous robots from research labs, something which has been re-invented by different groups over the past 20 years.



I’m just wondering what’s the advantage between this approach, versus using a simulation? I know a simulated world is never the same as the real world, but it’s easier, and could provide (almost) as much data as required?



Hi there.

I’m involved in development a platform allows to create a kind of “robot as a service”: First platform user created a platform (sic!) for direct drone hiring ( for example. I will be glad to exchange ideas too.


One of the papers @pitosalas wanted to attach is probably this one:

I also vaguely recall that someone, possibly Bosch, had a similar service available way back when the PR2s were first distributed, but I could be mixing it up with the lab from that paper.


Fair question @arifrahman: The feeling is that, once you’ve gotten “it” to work with the simulator, you need a real robot. And you may not have the facility or the space or the money to do so. Also depending on your software/technical level you may not have the ability. What I am envisioning (imagining?) is a way to program the “behavior” of the robot with less than the total flexibility of ROS, but still very powerful, and for a specific more limited domain, at least at first.



I have two link that can be interesting for you : , a old “game” where you reserve a robot for some time and control it to do some pick and place task. , an educational robot arm that can be programmed with snap! from a web interface.
I actually use Blockly (more or less an old version of snap) to let people (non-developer) program robots arm at work, it work pretty well.

Is your vision something like a mix of these 2 links ? Like a ROS controlled robot from snap! with a webcam and reservation system ?


the first two authors of the paper that you cited were with Bosch.


as an example with less sophistication than Snap!, there is


The earliest work of which I am aware is documented in the excellent book Beyond Webcams: An Introduction to Online Robots. Nov 2001. Eds. Ken Goldberg and Roland Siegwart. MIT Press. ISBN: 9780262072250.

It describes remote access to lab robots from around the years 1995, 1997, so some screenshots show the Mosaic Web browser, and some of the authors discuss using Java applets and clever tricks to make user’s Web pages refresh (because modern JavaScript did not exist yet).