ResearchBib Share Your Research, Maximize Your Social Impacts
Sign for Notice Everyday Sign up >> Login

Automatic mapping and localization in large-scale cyclic using K-nearest neighbours

Journal: International Journal of Advanced Technology and Engineering Exploration (IJATEE) (Vol.9, No. 97)

Publication Date:

Authors : ; ;

Page : 1802-1811

Keywords : Localization; K-Nearest neighbourhood; Robot path planning; Large-scale cycle.;

Source : Downloadexternal Find it from : Google Scholarexternal

Abstract

Simultaneous localization and mapping (SLAM) is a process or technique used by autonomous mobile robots to identify the location and regenerate a map for the surrounding environments where the robot moves. One of the preconditions for autonomous mobile robots is the ability to learn a regular environment model. Closed loops are considered one of the most effective issues in SLAM research areas. One of the main challenging problems in generating the environment map of closed loop is the data association problem, where loops in the surrounding will activate data association problem. The difficulty of a particular surrounding with closed loop is based on the value of uncertainty in local mapping and the productivity of the local map representation. Uncertainty management is a core challenge in SLAM. False matching due to unclear structure in the environment represent one of the most substantial difficulties to suitably closing large loops. In this paper, combination of scanning laser, distance meter, compass and k-nearest neighbourhood (KNN) were discussed to construct an absolute localization system. The KNN and distance equation as similarity measurement with specified threshold are used to solve the uncertainty problem and to specify the node that is closest to robot location. The results showed that the cosine method had the least error value with execution time (0.000405 s) while the Chebyshev method had the least execution time (0.000360 s) and error value of (2). The results indicate the cosine method with KNN has the minimum error and less execution time.

Last modified: 2023-01-05 19:21:00