The MANTIS Project aims at exploring the combintation of simulation, data augmentation and Machine learning for indoor localization. The quality of the simulated data for AI training has a strong impact on the performance and thus on the final localization indication from the AI. The project will focus on the introduction of more realistic elements in the modeling in INET framework for OmNET++, like walls and obstacles, to find the most suitable combination for proper dataset creation. We will test and identify the most suitable ML models and elaborate a data augmentation strategy along the appropriate simulator parameters, using reinforcement learning.
Last updated:19.03.2022
