Revision history [back]

Well, I guess that the http://www.rosindustrial.org/ website is the best proof that there is at least attempts to use ROS on an industrial scale (I don't know if there is systems currently in production however).

But again, the usual ambiguity appears in your question. If you ask if the roscomm layer and the ROS core tools are mature enough to support an use on an industrial case, I would say yes. There is, of course, still room for improvements but this seems definitively reasonable.

On the "pro" points, I would say that the despite its relative young age, the huge adoption of ROS by the community made it so that this low-layer has been intensively used in the last past year which makes it something one can rely on. At least, if you compare it to other general-purpose robotics middleware.

There is of course limitations. There is no formal verification, for instance, of the ROS core layer which makes it unsuitable for critical operations. But, if you have this kind of requirements, the amount of software you can choose from becomes severely limited anyway...

Again, on the "con" point, if you consider not only the ROS communication layer but also all the other associated package then the answer is more "no". I mean, some packages are stable and well-tested (the PR2 ones for instance) but most of them are research code which have been packaged and may or may not work depending on your setup. These packages are the work of volunteers and suffer from very little workforce available and the usual research syndrom (aka code it, publish it, throw it). It may not be the case for all packages but it is an usual problem.

There have been discussion on how to split the packages between "mature" packages one can rely on and unstable packages but so far, except the fact that some are available as Debian package, there is little hint to help one distinguish between these two kind of packages. Also, the ROS Industrial SIG have been discussing some kind of LTS support for ROS which would be more suited to plant automation than the current release cycle.

Well, I guess that the http://www.rosindustrial.org/ website is the best proof that there is at least attempts to use ROS on an industrial scale (I don't know if there is systems currently in production however).

But again, the usual ambiguity appears in your question. If you ask if the roscomm layer and the ROS core tools are mature enough to support an use on an industrial case, I would say yes. There is, of course, still room for improvements but this seems definitively reasonable.

On the "pro" points, I would say that the despite its relative young age, the huge adoption of ROS by the community made it so that this low-layer has been intensively used in the last past year which makes it something one can rely on. At least, if you compare it to other general-purpose robotics middleware.

There is of course limitations. There is no formal verification, for instance, of the ROS core layer which makes it unsuitable for critical operations. But, if you have this kind of requirements, the amount of software you can choose from becomes severely limited anyway...

Again, on the "con" point, if you consider not only the ROS communication layer but also all the other associated package then the answer is more "no". I mean, some packages are stable and well-tested (the PR2 ones for instance) but most of them are research code which have been packaged and may or may not work depending on your setup. These packages are the work of volunteers and suffer from very little workforce available and the usual research syndrom (aka code it, publish it, throw it). It may not be the case for all packages but it is an usual problem.

There have been discussion on how to split the packages between "mature" packages one can rely on and unstable packages but so far, except the fact that some are available as Debian package, there is little hint to help one distinguish between these two kind of packages. Also, the ROS Industrial SIG have been discussing some kind of LTS support for ROS which would be more suited to plant automation than the current release cycle.

Well, I guess that the http://www.rosindustrial.org/ website is the best proof that there is at least attempts to use ROS on an industrial scale (I don't know if there is systems currently in production however).

But again, the usual ambiguity appears in your question. If you ask if the roscomm layer and the ROS core tools are mature enough to support an use on an industrial case, I would say yes. There is, of course, still room for improvements but this seems definitively reasonable.

On the "pro" points, I would say that despite its relative young age, the huge adoption of ROS by the community made it so that this low-layer has been intensively used in the last past year which makes it something one can rely on. At least, if you compare it to other general-purpose robotics middleware.

There is of course limitations. There is no formal verification, for instance, of the ROS core layer which makes it unsuitable for critical operations. But, if you have this kind of requirements, the amount of software you can choose from becomes severely limited anyway...

Again, on the "con" point, if you consider not only the ROS communication layer but also all the other associated package then the answer is more "no". I mean, some packages are stable and well-tested (the PR2 ones for instance) but most of them are research code which have been packaged and may or may not work depending on your setup. These packages are the work of volunteers and suffer from very little workforce available and the usual research syndrom (aka code it, publish it, throw it). It may not be the case for all packages but it is an usual problem.

There have been discussion on how to split the packages between "mature" packages one can rely on and unstable packages but so far, except the fact that some are available as Debian package, packages, there is little hint to help one distinguish between these two kind of packages. Also, the ROS Industrial SIG have been discussing some kind of LTS support for ROS which would be more suited to plant automation than the current release cycle.