Implementation of Levenberg-Marquardt Point to Line Iterative Closest Point and Pose Graph Optimization for 2D Indoor Mapping on Differential Drive Mobile Robot

PT Adhikara Wiyasa Gani (AWG) faces a physical track especially a magnetic tape durability issue on its Automated Guided Vehicle (AGV) due to crossing by heavy-duty vehicles.Free navigation and path planning can be a solution that allows the AGV to dynamically adjust its path without relying on physical trajectories.For free navigation to be realized, the robot needs a map or knowledge of its working area.This research uses Iterative Closest Point (ICP) and Pose Graph Optimization (PGO) for mapping methods with LSLiDAR N10 and DDSM115 motors on three different artificial maps.The robot has a differential drive steering model with dimensions of 35 cm x 30 cm.The mapping results were compared to ground truth maps using Average Distance Nearest Neighbor (ADNN) and Structural Similarity Index Measure (SSIM) metrics.The results show that mapping method can be used for room localization and mapping quite well.The ground truth map is formed on a 10 x 6 squares grid map, with dimensions of 60 cm x 60 cm for each square.Mapping with the combination of ICP, PGO, and wheel odometry produced ADNN and SSIM values of 5,5 cm and 0,601; 8,8 cm and 0,669; and 8,5 cm and 0,629, respectively, for the three maps tested.The largest value of the ADNN metric is 8,8 cm, this value is used as padding in the robot dimensions so that there is a remaining 16,2 cm on the length side of the robot and 21,2 cm on the width side of the robot with respect to a square grid.