Accuracy Evaluation of LiDAR-SLAM Based 2-Dimensional Modelling for Indoor Environment: A Case Study

Abstract

The rapid development of sensor technologies has led to smaller sensor sizes and lower costs. Today, the easy-of-use purchasing of sensors such as cameras, Light Detection and Ranging (LiDAR), Radio Detection and Ranging (RADAR), Inertial Measurement Units (IMUs), and Global Navigation Satellite System (GNSS) receivers have led to significant developments in many applications such as robotics and unmanned vehicles. Sensor data is transformed into information or products thanks to the methods. Simultaneous Localization and Mapping (SLAM) is one of the critical methods in which the vehicle's location is determined, and the environment is modelled. This method can realize applications using detection sensors such as cameras, LiDAR, or RADAR. This study aimed to model an indoor area with a twodimensional (2D) LiDAR sensor placed on an Unmanned Ground Vehicle (UGV) and to analyse the accuracy of the produced model. Normal Distribution Transform (NDT)- Particle Swarm Optimization (PSO) algorithm was used to generate the 2D model from the collected LiDAR data. The NDT-PSO algorithm was executed on the Robot Operating System (ROS) installed on the Jetson Nano Developer Kit, and a real-time 2D model of the working area was processed. The reference lengths of the 75 facades in the 232 m2 indoor space were measured using a total station and calculated with CAD software. Percent error values were evaluated by comparing the reference and model lengths of the facades.