You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We observe on our systemtests machine (Threadripper with 32 cores) that the continuously monitored unicast latency is around 20ms, while on older i7 Laptops the same latency is typically below 10ms.
Tests so far:
Same behavior for Crazyradio PA and Crazyradio2
comCheck reports similar values on both machines (3ms latency, 1300 packets/s)
Changing the motion capture frequency to 50Hz (from the default 100Hz), lowers the latency to 4ms. There is a sudden jump from 90 Hz (10 ms) to 100 Hz (20ms). Higher values result in higher latency (30ms for 240 Hz).
Overall this seems to be some issue with the multithreading, either on the ROS side or the crazyflie-link-cpp side.
The text was updated successfully, but these errors were encountered:
On this particular machine the vendor calls (changing address and enableAck) take 5ms, limiting the throughput to 200 packets/s if switches between broadcast/non-broadcast are needed. Arnaud mentioned that this could be USB chipset related
Current workaround is to use 90 Hz for the mocap.
There seem to be other related issues related to the ack-rate, which sometimes drops to 0.5.
We observe on our systemtests machine (Threadripper with 32 cores) that the continuously monitored unicast latency is around 20ms, while on older i7 Laptops the same latency is typically below 10ms.
Tests so far:
comCheck
reports similar values on both machines (3ms latency, 1300 packets/s)Overall this seems to be some issue with the multithreading, either on the ROS side or the crazyflie-link-cpp side.
The text was updated successfully, but these errors were encountered: