How can I generate synthetic data from network simulation tools (eg: NetSim) to train ML algorithms running in O-RAN RAN Intelligent Controller (RIC)? What kind of data should we capture from the simulator?
I can see some of the measurement values in NetSim are shown in https://tetcos.com/5g.html. But I want to know if all information in NetSim is available periodically or is it event triggered?
And is it also possible to application-level data like DL throughput, UL throughput, etc?
I would think that RL is the right approach. For example, you can have gNB(s) transmit power(s) as an “action” that changes the internal “state” of the network (desired signal and interference powers) and the “reward” can be the throughput or delay.
I appreciate all your answers. I was wondering if it is possible to use NetSim to generate synthetic data of (i) Radio resource allocation (for scheduler optimization) and (ii) Handovers (to tune handover margin, time to trigger etc)