On Tue, Apr 13, 2010 at 1:22 AM, Rosen Diankov wrote: > If processes immediately die without properly calling the ros shutdown > methods, the node will stay registered with the core forever. the > result is that the rxgraph slowly gets cluttered with dead nodes and > makes it really inconvenient. we've tried 'rosnode kill', but that > doesn't seem to un-register the node. because we want to keep a > roscore constantly running on every robot, we cannot just restart the > roscore.... hi Rosen, You can try the semi-hidden 'cleanup' feature of rosnode: rosnode cleanup It will ping all nodes, then offer to delete entries from the master that correspond to the nodes that don't respond. brian.