Co-design of architectures and algorithms for mobile robot localization and model-based detection of obstacles

Abstract

This thesis proposes SoPC (System on a Programmable Chip) architectures for efficient embedding of vison-based localization and obstacle detection tasks in a navigational pipeline on autonomous mobile robots. The obtained results are equivalent or better in comparison to state-ofthe- art. For localization, an efficient hardware architecture that supports EKF-SLAM's local map management with seven-dimensional landmarks in real time is developed. For obstacle detection a novel method of object recognition is proposed - detection by identification framework based on single detection window scale. This framework allows adequate algorithmic precision and execution speeds on embedded hardware platforms.

Ova teza bavi se dizajnom SoPC (engl. System on a Programmable Chip) arhitektura i algoritama za efikasnu implementaciju zadataka lokalizacije i detekcije prepreka baziranih na viziji u kontekstu autonomne robotske navigacije. Za lokalizaciju, razvijena je efikasna računarska arhitektura za EKF-SLAM algoritam, koja podržava skladištenje i obradu sedmodimenzionalnih orijentira lokalne mape u realnom vremenu. Za detekciju prepreka je predložena nova metoda prepoznavanja objekata u slici putem prozora detekcije fiksne dimenzije, koja omogućava veću brzinu izvršavanja algoritma detekcije na namenskim računarskim platformama.