Everyone interested in safety-critical applications of ROS has had some time to stew and a few more people have risen up to offer their services, so I’d like to start organising our next meeting. Please answer the poll below for your preferred meeting times.
We will soon need to start focusing on something concrete that we can achieve for safety-critical applications. Ideally I’d like us to find something technical to contribute because that will be the most beneficial to all, but document-related contributions, such as the proposed sample safety case, are also useful.
Based on the results of the survey, I’ve set the next meeting time to Thursday, August 8, 2019 2:00 PM. The information for joining is below.
I do not yet have much of an agenda, because we still seem to be attracting new people each time. However there are more people with expertise in functional safety joining now so I believe we can start to get some firm ideas about what we can do that will be a useful contribution to the ROS community.
Please post items you would like to discuss in the meeting below. I intend to talk a bit about what is necessary to produce a sample safety case so we can start figuring out who can provide the work.
ROS 2 Safety Working Group
Thu, 8 Aug 2019 23:00 - 00:00 JST
Pity! I will be traveling (via train) at that time . I may join via phone then. Let’s hope the connection allows me to do so.
Just out of curiosity @gbiggs, I checked the survey quickly and saw that the 6th of August slot was rather popular, maybe it wasn’t convenient for you?
Yes, unfortunately I had another event come in on that date. If you would like me to shift it to that time, then I am happy to do so, but I will need someone else to run the meeting in case I can’t make it in time.
What can we do, given our small number of engineers who have experience in functional safety?
List of best practices for things that a safe robot should do
e.g. what should be on the start up checklist?
e.g. how do you set up an estop that works safely?
Define the minimum core set of packages to be safe?
Could we define a minimum example, such as collision avoidance, and do the full vertical stack that is necessary to achieve doing that safely?
Difficult because ROS itself is not certifiably safe, but we could do it on the assumption that ROS is certifiably safe. Or we could make a non-ROS monitor to watch over the system and do the safety analysis necessary to show that the safety monitor is free from interference from ROS.
A generic safety monitor for use with a ROS system?
. I may join via phone then. Let’s hope the connection allows me to do so.