Ambiguity resolving in structured light 2D range finder for SLAM operation for home robot applications

Abstract

Since the unveil of its first prototype of robot vacuum cleaner in 2003, Samsung Electronics, Korea (SEK) has been developing a commercial version with an onboard global localization system. Global localization can be either problematic or trivial depending on the geometry of the operation field, and affordable sensors on or off the robot. It is generally accepted that home environments are one of the most harsh to artifacts due to the unstructured and unpredictable factors within them. Price limits and average users' reluctance to put off-robot localization sensors at or modify home interior to ease the localization make the situation still more difficult. In this paper, we introduce one of the on-board localization sensor systems developed at Samsung Advanced Institute of Technology for possible installation on SEK's commercial home robots. Application results to home SLAMs (simultaneous localization and mappings) are also reported. The system with an ordinary B/W camera and low power LD was relatively cheap compared to a TOF (time of flight) based laser range finder but showed its efficacy in that its laterally dense property made it possible to find a small corners such as arm rests or door-edges, often existing in home environments, within its reliable sensing range (four meters typically). Undesirable external effects caused by e.g. sunlights or other light sources are rejected using the Bayes' rule.