Hello,

I have a question regarding software architecture in ROS for humanoid robots.

Let say I have a humanoid robot that can do various tasks. Ultimately each task publishes motor control commands in order to move the robot in a certain way. If the robot tries to do 2 tasks at the same time, that means 2 nodes publish motor commands on the same topic, the commands will be mixed and the robot will fail at both tasks. Is there a common used way to organize this in a higher level architecture ?

What I tried so far was :

• Using a "conductor" node to allow only one task at a time to execute.
• Each task is a set of nodes linked to a service, and the "conductor" can send a request to that service which is turning ON or OFF this task.
• As each complex task is divided into multiple small steps, the "conductor" then acts as a state-machine.

Does anyone has experience/knowledge on software architecture for robots, or is there any package that would help doing so ?

Thanks!