Accuracy Evaluation of SLAM Algorithms in RGB-D Sensors

Abstract

Since visualization is a powerful technique to reinforce human cognition, real time robots can be used to study its environment to make a map or a model. Mapping is a fundamental task in navigation through unknown environments. Due to its increased application in robotics, surveillance, tracking system used in security and many other areas researchers have propelled to continuously experiment in more efficient and competitive techniques. In this paper we present an experimental evaluation for mapping system that robustly generates 2D maps using a RGB-D sensor. The evaluation was done for our mobile system which comprises a Kinect. This work mainly presents research on earlier related techniques for mapping evaluation and a comparison of SLAM algorithms. The paper concludes with the analysis of mapping accuracy with the ground truth. Experiment indicate that Gmapping gives an accurate solution for 2D mapping among many ROS compatible algorithms for computationally limited robots.