CTW Localization-Competition

DL+Fingerprinting model for localizing a robot with Channel responses from antenna.

Hybrid model:

1. FingerPrint_Model: contains FP model with data Outputs: csv file with the estimation from the FP model
2. DL Model: Contrains the trained DL models and the Hybrid (a regression model to merge the FP with the DL models)

Two methods were chosen to merge the models:

• Simple divergence trained with a single parameter
• XGB

XGB giving better results. The data for the later training was ensured with new tets data and 10 fold cross-validation.

This images shows the two trainings, after, we apply the regression model to the result of the training of both models: