Is it possible to setup Jetson Nano in something similar to headless operation without using a wifi connection, but instead something like telemetry?

Headless operation of Jetson Nano is easy utilizing wifi assuming that a local network is accessible. However, there are many cases where a wifi network is not accessible. Is it, therefore, possible to do something similar with telemetry or other method? Obviously this would require external hardware, but insofar as the jetson is concerned, would it be compatible? Such a solution would be helpful in my endeavor. Thank you.

I’m not sure what you mean by telemetry. It is easy for any Linux (or UNIX-like) system to run headless. They all have the “localhost” network even when disconnected such that network applications can talk to each other locally without any network. The biggest problem is that sometimes a GPU-based program needs a “context”…the designation of a GPU environment…or it won’t work. A lot of CUDA programs need a context. One normally gets that context by running a GUI login. If this is not available, and if it is the case of needing a context, one would run a virtual X server instead of an actual local monitor/keyboard/mouse. Programs needing a context don’t care if there is a virtual desktop which is not actually connected to anything or if it is an actual local monitor. Both provide the context. Is that what you mean?

1 Like

Thank you @linuxdev for your reply. After doing some more reading online upon seeing your reply, it is clear to me I had some misunderstandings. I had a misunderstanding on how network applications work in general and was not aware of the differences between LAN and WAN. I initially thought one needs to be connected to the internet for local devices to communicate over a network so I was looking for alternatives, but this is clearly a misunderstanding of mine. It seems that the best solution for me is to have a dedicated router without an internet connection to communicate data collected by the jetson nano to another linux device within a LAN. I assume the nano is compatible for that application? The reason I was asking about telemetry is because I have successfully used telemetry in the past for extending serial communication. These telemetry devices I had used in the past worked well in that area, but I imagine they are not optimized for this application. Thank you; you can close the ticket if my understanding is correct.

Yes, as a matter of fact I have maybe 6 or so Jetsons on one router and local network and PC. They can all talk to each other simultaneously because of the router being on the same subnet. They can all talk to the outside world via the router as well. If the outside world internet has failed, all can still talk to each other.

The question of having some outside world device originate a connection to a Jetson would have to be set up in the router for port forwarding (the router knows what addresses are replying to something inside the router, but if no Jetson or other internal device originates a connection, and something from the outside world wants in, then the router will reject this unless specifically told to allow it, e.g., you’re on vacation and connecting via ssh and not the local network).

Jetson ethernet is like any other computer, and if there is a network service or program on it compatible with some other computer, then it shouldn’t matter that it is a Jetson. Wired ethernet is the preferred way of talking between local computers in most cases. Does your application use networks to talk? There are multiple protocols and such, but the ethernet side should not be an issue.

1 Like

Thank you for you clarification. This was very helpful; you can close the ticket.

I do not need to utilize port forwarding. My application is a robotics application that requires a monitoring computer physically disconnected from the robot. Additionally, I am trying to avoid connecting to other networks (besides the one through my own router) as I expect my implementation to work in any facility with minimal configurations. It is particularly best in my application not to use the network already existing in the facility. And yes, I have seen implementations similar to mine that use networks to talk, so that is the path I will take next. Thank you!