Feature-relative real-time obstacle avoidance and mapping

Abstract

A substantial challenge in robotics is integration of complex software systems for real- time performance. This thesis integrates the robust and generic mapping framework Atlas, a feature-based local Simultaneous Localization and Mapping (SLAM) module, and obstacle avoidance using information from mapped features. The resulting system performs autonomous feature-relative Real-time Obstacle Avoidance and Mapping (ROAM) with laser or sonar range sensors, and results are shown for wide-beam sonar. This system will allow high-speed feature-relative obstacle and avoidance and navigation on mobile robots with wide-beam sonar and/or laser sensors.