• DepthAI-v2Hardware

  • Inaccessible after simultaneously restarting OAK-D POE in Standalone and router

I have a robot that has a Jetson connected via ethernet to OAK-D-Pro-W-PoE-97's running flashed in standalone mode. The pipeline exposes a port and frames are streamed over TCP.

When the robot is powered down and then powered up again, the OAK-D's, Jetson, switch and other devices are simultaneously being powered on. 75% of the time the devices get stuck and are inaccessible over the network. They do not respond to pings and cannot be discovered via depthai. The only remedy is to power the OAK-D on and off. This fixes the issue 100% of the time.

I've been able to reproduce this off the bot as well, connected to tp-link switch, powering the OAK-D off and the switch off and powering them both on simultaneously.

We only started seeing this issue when the devices were running in standalone mode. We've been running for months with them connected and provisioned at runtime via the host.

Running depthai python 2.30.0.0.

The only other relevant detail is that the OAK-D's are not powered over ethernet but via 5V through the M8.

Has anyone seen this? What can be done?

    mbrevoort
    Interesting. In the off the bot test you ran, was the device still powered via M8? What if you add a 5s delay between switch power on and OAK power on?

    Thanks,
    Jaka

      12 days later

      Yes, in the test it was powered by the programming board. What’s happening is that the camera device comes up before the switch’s DHCP service and I think falls back to its default IP which is not routable on our network. So the only reliable solution was to assign static IPs. Orchestrating power sequencing and delays so the switch comes up reliably first was not feasible in our design at this point.