[kyle_kepley]

I wanted to work with this for a while to get a better idea of what the problem is. It appears that there is a buffer where RoboDK is putting the commands that get sent to the Fanuc when running simulations in the "Run on Robot" mode. When there is a problem on the Fanuc side such as DCS error, user E-stop, light curtain break or anything that causes an alarm and haults the robot, RoboDK does not know and continues to try and run the program. Once the robot is brought back online on the Fanuc side, then commands from the buffer on the PC resume being sent to the robot. The user can avoid this by closing the connection and stopping the RDK_3 driver program before putting the robot back online, which will kill the cache. However, if the robot becomes active again without having first manually cutting ties with the PC, then the robot will start executing commands and begin moving unprovoked wherever it happens to be. Ideally RoboDK should have a way to know when the robot has an error or is E-stopped and then terminate the program on the PC side and empty the command cache.

[Albert]

The buffering of commands happens when rounding/blending is activated. Can you try disabling the rounding/blending feature?

The rounding/blending are the CNT movements for Fanuc instead of FINE movements.

To disable rounding/blending you can simply not use the set rounding instruction (or set it to -1).

[kyle_kepley]

Is the rounding/blending a Fanuc setting or a RoboDK setting? I am not using a "set rounding instruction", assuming that is an instruction that gets used in RoboDK programs.

[Albert]

You can learn more about the rounding instruction here:
https://robodk.com/doc/en/Robot-Programs.html#InsSmooth

If you have a robot machining project or curve follow project you can define the rounding with your program events.