[Feature Request]: Dynamic role switcher

[techycpa]

Platform

Cross-Platform

Description

We have seen many questions about what device role to use with most suggestions to select CLIENT. If each node could automatically change its role type based on changing conditions could improve the stability of the network. I’d propose a node type of AUTO which is set as default and dynamically adjusts.

Example, a node normally located on the edge of the mesh travels into a high density coverage area. Setting to CLIENT MUTE may make sense as it won’t need to retransmit messages. Another example might be a node set as ROUTER goes offline. Another well placed node that is CLIENT or ROUTER LATE might get promoted to ROUTER to stabilize the mesh.

[grleblanc]

+1, I really like this idea.

[fifieldt]

It's a nice idea, but probably pretty hard to implement. Especially to the point it'd be a reasonable default, which I assume is the aim.

Consider eg that a city-wide mesh is just one kind of use case. Also consider that someone going hiking up the nearby mountain doesn't necessarily want their personal device to drain battery faster than before. Consider also devices that are a little "deaf" compared to their "shout".

We already do some scaling automatically based on number of nodes seen, and the air utilization. Retransmissions are already attempted or not attempted based on what other transmissions are seen.

I think to make this idea reality, the next step would be to test out some approaches in the simulator: https://github.com/meshtastic/Meshtasticator

If there are significant benefits in there, then we might have a winner.

[rellimmot]

+1 I would love programmatic role switching for off-grid backup, so if my ROUTER/REPEATER loses grid power input it could automatically switch to ROUTER_LATE when running off solar-battery-backup or perhaps even some programmatic threshold like battery SoC.

[fifieldt]

converted to a discussion so we can keep talking about this idea without worrying about the stale issue bot :)

[Ryu945]

I was thinking about smartly doing routing based on network analysis. The only conclusion I came up with is that the network doesn't have the bandwidth to analyze itself. That is why it is dependent on the user who can look at the network and tell it the correct role.

Long term, V3 needs a new packet type that is rather small and can be spammed through the network for the purpose of seeing how each node connects together. Then it could analyze itself and make smart decisions. Right now, we only have one packet type in V2 and it is far to large for network analysis. It will clog the network to try to do network analysis with a packet that large.