Autonomous Navigation of Indoor Service Robot with Object Recognition using Deep Learning

• Published in: International journal for research in applied science and engineering technology Vol. 10; no. 6; pp. 2339 - 2353
• Main Authors: S, Abhijith, S, Dr. Sreeja
• Format: Journal Article
• Language: English
• Published: 30.06.2022
• ISSN: 2321-9653; 2321-9653
• DOI: 10.22214/ijraset.2022.44316

Indoor Service robots are becoming popular day by day. The task of service robot is to perform some predefined tasks originally done by human in houses, schools, offices etc. This paper proposes an autonomous indoor service robot system integrated with multi-sensors such as Lidar and Kinect for the smooth functioning of indoor service robots. Firstly, a map of the environment is created using Simultaneous Localization and Mapping (SLAM), a commonly used technique for creating map and positioning the robot at the same. In this paper different Lidar-based SLAM are compared and the best SLAM is chosen for creating 3D map for this system. For autonomous navigation, path planning plays a key role for fulfilling the task. A* algorithm is incorporated with this system to plan a global path. But global path planning algorithm fails to complete its task if any new or dynamic objects come to its path of navigation. To overcome such situations, a local path planning algorithm called Dynamic Window Approach (DWA) is added to this system. Another inevitable element in autonomous navigation is localization of robots in the map. Adaptive Monte Carlo Localization (AMCL) is the technique applied in the proposed system to localize the robot for the effective navigation of indoor service robot. To enhance the performance of service robot, YOLOv5 model is integrated with this system for real time object detection. The entire system is simulated in an open-source framework called Robot Operating System (ROS).