Um sistema para mapeamento de grandes regiões usando GraphSLAM

• Autor(a) principal: Mutz, Filipe Wall
• Data de Publicação: 2014
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Institucional da Universidade Federal do Espírito Santo (riUfes)
• Texto Completo: http://repositorio.ufes.br/handle/10/4264
• topic: Localização e mapeamento simultâneos (SLAM); Ciência da Computação; Robótica; Inteligência computacional; Aprendizado do computador; Veículos autônomos; Mapas - Escalas; 004

One of the big challenges of robotics in the last years has been the development of robotic vehicles with high level of autonomy able of navigating for long periods of time without human intervention. There is an high expectation that, in future, autonomous vehicles can be used to provide accessibility to injured people, reduce time and cost of transportation systems, and offer comfort to people that do not want (or cannot) to drive. Besides of that, the use of autonomous vehicles has the potential to highly increase the safety in traffic due to the increased capacity of robots to observe the world, and due to the fact that robots are not subject to dangerous states, such as tiredness, anger, drunkenness, hurry or stress. The focus of this work was the problem of map building in big environments in order to make autonomous vehicles able to localize themselves and navigate. It was studied the well succeeded solutions found in literature and it was developed a Large-scale Environment Mapping System (called LEMS) using pose-based GraphSLAM algorithm. Experiments revealed that the LEMS was able to build highquality maps of different regions with specific sizes and features. During the development of LEMS, it was observed that the odometer data errors are non-Gaussians and subject to biases. Because of that, it was created a tool to automatically find the biases using particle swarm optimization. Experiments revealed that integrating the biases to odometer data, the dead-reckoning was able to reproduce with a good precision the path measured by GPS.