Trabalho de Conclusão de Curso | Graduação
URI permanente desta comunidadehttps://repositorio.insper.edu.br/handle/11224/3244
Navegar
2 resultados
Resultados da Pesquisa
Trabalho de Conclusão de Curso Mapeamento de Carrinho em Supermercado(2025) Campelo, Ananda Julia Galvão; Shinohara, Kevin Nagayuki; Machado, Luca Cazzolato; Sousa, Thiago Gonçalves Guadagnoli deSupermarkets and large retailers strive to understand customer flow to optimize store layouts and increase sales. While technologies like SLAM and smartphone data crowdsourcing have been explored for indoor mapping, the specific context of supermarkets and the use of high-precision technologies such as Ultra-Wideband (UWB) remain largely underexplored. This project proposes the creation of supermarket maps using the trajectories of shopping carts tracked by UWB, a technology that offers greater spatial accuracy than methods based on Wi-Fi or Bluetooth. The proposal also includes generating heatmaps that visualize the frequency of cart movement. The ongoing methodology addresses outlier treatment through data cleaning, trajectory rasterization, and the concept of density matrices. The generated maps are expected to be accurate representations of the supermarket’s floor plan, and the density heatmaps should precisely translate customer flow within each sector. The proposed solution serves as a basis for an Indoor Positioning System (IPS), consumer behavior analysis, and shopping route optimization.Trabalho de Conclusão de Curso Intelligent Minigame Selection for the Game Arena of Dreams(2025) Machado, Luca Cazzolato; Almeida, Pedro Luiz Fracassi de; Colpas, Pedro Henrique RizoThis project addresses the challenge of content repetition in Arena of Dreams, a partyroyale mobile game developed by Fanatee, where excessive repetition of minigames can undermine player engagement and retention. The project aimed to replace the game’s purely random minigame selection system with an intelligent algorithm that reduces the player’s perception of repetitiveness by spacing out similar experiences. Through a methodology involving online and in-person user surveys, computer vision and data analysis, the team quantified how players perceive similarity between minigames. Multiple distance matrices were generated from different perspectives (user perception, visual features, semantic descriptions) and combined using Multi-View Multidimensional Scaling (MVMDS) to create n-dimensional embeddings representing each minigame. The selection algorithm then uses these embeddings to calculate the optimal minigame choice based on players’ recent match history, selecting minigames that are furthest from what players have recently experienced. Validation results demonstrate that this approach successfully reduces the perception of repetition by understanding and quantifying similarity, ultimately creating a more enjoyable and engaging gaming experience for players.