Parameter Optimization Framework for NS-3 network optimization
This software assumes the following prerequisites are available on the host system:
First, the ns3 module qos-app must be added to the ns3 src/ folder and then ns3 must be rebuilt. Note:- The ns3 model under test must implement at minimum the client and server ns3 application models for this to work
-
MutationChance (%): The percentage chance of a mutation occuring per trait
-
MutationRate (%): The percentage variation possible from the original trait due to a mutation. This only affects numerical traits.
-
MaxElite : The number of Elites allowed for breeding future generations.
-
MaxPopulation : The Maximum number of number of potential parents
-
MaxGeneration : The maximum number of generations which will be run
-
StepSize : The minimum numerical difference between child traits when optimizing using gradient descent method
Results are stored in different locations depending on their aggregation level:
- Raw_Results : Stores each result for individual runs of the model
- Av_Results : Stores each run result which consists of averaged result from each device across the run
- Optimal_Results : Stores optimal result of each generation which consists of lowest cost result from each device across the run
The following publication has been released detailing this software. Please cite this paper if publishing any work using this tool:
A. Hasandka, J. Zhang, S. M. S. Alam, A. R. Florita and B. Hodge, "Simulation-based Parameter Optimization Framework
for Large-Scale Hybrid Smart Grid Communications Systems Design," 2018 IEEE International Conference
on Communications, Control, and Computing Technologies for Smart Grids (SmartGridComm), Aalborg, 2018, pp. 1-7.
doi: 10.1109/SmartGridComm.2018.8587472