Computação de caminhos enganosos considerando características topográficas do terreno e redes neurais profundas

Abstract

There’s a variety of real life applications for algorithms with deceptive characteristics, and it’s an area in Artificial Inteligence with a lot still left to be explored, specially in the context of pathfinding algorithms. Most studies in the literature about deceptive pathfinding are applied in two-dimensional spaces, and there’s little to no studies about the application of machine learning in this context. This work aims to investigate deceptive pathfinding algorithms and apply them in terrains with topographic characteristics, analysing the impact of explicitly considering these characteristics. This work also seeks to explore the application of deep neural networks as a way to optimize the computation of these paths. Afterwards, a statistical analysis of the results is made to determine the impact that these techniques have in the final path and in their processing time.