ITEE Innovation Expo 2008 : Project Details

ROBUST BEHAVIOURS FOR RATSLAM

Many high level Simultaneous Localization And Mapping (SLAM) algorithms, including RatSLAM, have difficulty navigating through obstacle rich environments. This is especially true when sensor readings are intermittent for particular objects, such as table legs or office chairs. Robust Behaviors for RatSLAM provides a potential solution to the problem. The solution has been written to operate within the RatSLAM system while running on a live robot. A user interface was developed to give insight on how the algorithms maintain navigational stability within a local environment. To sustain an accurate pose of the robot the system employs a scan matching technique, whereby offsets are applied to the current sensor readings to most accurately calculate the change in pose. Intermittent object visibility is countered with an obstacle retention algorithm which determines which and how many points to store in the robots local visibility memory.

The system was tested on a Pioneer 3 robot with a HYOKU laser range finder and left to roam autonomously in random goal selection mode for 2 hours. Over the course of the demonstration the algorithm only had difficulty on one obstacle which was a table leg that the range finder was unable to detect from particular approach vectors. However, the rest of the demonstration was incident free and showed that the Robust Behaviors algorithm is an effective and viable solution to the obstacle rich local environment problem encountered by many SLAM systems.