Host: The Japan Society of Mechanical Engineers
Name : [in Japanese]
Date : May 10, 2017 - May 13, 2017
The paper presents a failure detection of the Gyro-integrated Iterative Closest Point Simultaneous Localization and Mapping (G-ICP SLAM) for useful 3D mapping application. The proposed failure detection use gyroscope input and localization results of SLAM process. It might be effective for other ICP based SLAM algorithms. In SLAM based mapping system, it generates the map by adding time series of observation data sequentially. Some wrong projections affect subsequent calculation and the mapping error detection is important function for practical use. We first define the error pattern of mapping result and propose evaluation functions to detect de- fined mapping errors. The proposed method is implemented to G-ICP SLAM and evaluated with hand-held sensor observation data. The evaluation result show that the system could correctly detect skewed part of mapping process and recover automatically.
