Project - Reinforcement Learning

I have a network 30 nodes 5 of them (nodes 1-5) have data to be sent to the receiver which is one node (node 10). Other nodes are relay nodes Each node has 5 different options for Bandwidths to be used and different options of frequency ( for now we can assume just one frequency option 1.5e11 Hz for simplicity and later you have to add the frequency option ) So the model should select the optimal path to the receiver and the optimal Bandwidth and frequency selection to minimize the energy consumption. I already did the mathematical model and I know the behaviour of the model and I need the RL model to have the same behavior. when the maximum cluster size is too small (1mm *1mm), the sensor nodes select a higher BW option ( e.g. 100 GHz) that would consume less energy compared to a smaller BW option. Still, it can only be used for short distances, up to D_max=11 mm. When the D_max increases, selecting a higher BW will not be optimal because the achieved channel rate is less than the required data rate. Therefore, lower BW should be used at longer distances, with the price of consuming higher energy due to longer pulse duration TP. When the Dmax increased, the model used a larger number of nano-relays to reduce the distance between every pair of communicating nano-nodes. The smaller distances will enable the nano-nodes to operate at higher BW and, hence, at the shortest possible pulse duration TP_w which will minimise the overall energy consumption in transmitting/receiving bits. Results show that there is no BW less than 25 GHz is selected, which utilised lower channel bandwidth and hence longer TP_w. So, I need an RL model to get the same behaviour and the most important point I need to maintain is that the relay nodes need to combine the received traffic and resend all the received traffic as one message to one destination. Finally, when the model is done, I need some graphs to display the results like ( total network energy consumption, which node selects which BW at which distance, reward graphs, and learning graphs). we can discuss the graphs later. I need to understand each step of the full code and I asked for regular meetings to flow the progress.