The Performance Impact when Optimizing Mapping Algorithms for an FPGA-based Mobile Robot

• Tipo: Artigo em Livro de Atas de Conferência Nacional
• Data: 2010

Autores

Manuel Luís C. Reis (Autor) FEUP; João M. P. Cardoso (Autor) FEUP; João P. C. Ferreira (Autor) FEUP

Ata de Conferência Nacional

Título: Actas das VI Jornadas sobre Sistemas Reconfiguráveis - REC'2010

Páginas: 103-110

VI Jornadas sobre Sistemas Reconfiguráveis (REC'2010)

Aveiro, 4 a 5 de Fevereiro, 2010

Classificação Científica

FOS: Ciências da engenharia e tecnologias > Engenharia electrotécnica, electrónica e informática

CORDIS: Ciências Tecnológicas > Engenharia > Engenharia electrónica ; Ciências Tecnológicas > Engenharia > Engenharia de computadores

Outras Informações

Abstract (EN): FPGA-based solutions are being used to meet performance requirements in embedded systems. The mobile robotics field is an appropriate domain to evaluate whether FPGA-based systems are capable of managing complex computing tasks. This paper presents an autonomous mobile robot prototype that includes an FPGA board as the main central processing component. In this work we evaluate the prototype using mobile robotics mapping algorithms (i.e., algorithms able to build a map of the environment). These algorithms rely mainly on probabilities and are computationally intensive. The implementation used in this paper is based on an environment occupation grid. An uncertainty model of the sensor used to measure the distance of the robot to the obstacles in the environment is used to update the probabilities. We start the experiments by analysing the performance impact of using hardware modules, such as FPU and cache memory. Then, we consider modifications in the updating algorithm to reduce the overall execution time. The overall improvements allowed for an execution time of the updating task of 9.83 ms for an implementation based on Bayes’s algorithm and 10.86ms for Dempster-Shafer’s algorithm, both at the maximum distance considered.

Idioma: Inglês

Tipo (Avaliação Docente): Científica

Nº de páginas: 8