Page 241 - The-5th-MCAIT2021-eProceeding

Multi-points Navigation for Autonomous Robot in Duct
Environment

a
a
Ghassan Jasim AL-Anizy , Khairunnisa’ Ahmad Shahrim ,
a
b
Mehak Raibail , Abdul Hadi Abd Rahman *

a Machine Learning and Vision Lab, Faculty of Information Science and Technology, Universiti Kebangsaan Malaysia
b Center for Artificial Intelligence Technology, Universiti Kebangsaan Malaysia.
* Email: abdulhadi@ukm.edu.my

Abstract
Navigation in a duct environment is challenging due to several factors such as size, shape, and lighting. Various research
has been initiated to discover the best approach to allow autonomous robot navigation using appropriate path planning
algorithm. However, most papers focused on the sewer environment, which is different compared to the air duct
environment. While the lighting is considered ineffective, robot navigation should rely on a laser range finder and multi-
point references. This paper we study the effectiveness of rapidly exploring random tree (RRT) and path planning
algorithms for a mobile robot in a simulated duct environment. The RRT algorithm is chosen due to its biased towards
unexplored regions. The experiment is conducted using a single robot in Robot Operating System (ROS) framework to
evaluate the capability of multi-point robot navigation. It is observed that successful navigations are achieved in repeated
experiments. Future work will concentrate on multi-agent system and optimizing different path planning algorithm.

Keywords: RRT; Path Planning; Duct Environment; Autonomous Robot

1. Introduction
Navigating mobile robots in a duct environment is challenging due to several factors, such as the small size
of ducts (Koledoye et al. 2017), various shapes and dimensions (Chataigner et al. 2020). The ability to
overcome all these constraints allows autonomous mobile robots to perform various tasks such as surveillance
or cleaning tasks. Exploration techniques based on frontiers (Wang et al. 2011) are widely used for robotic
exploration to direct robots to frontier edges. However, there are no general techniques that can ensure
successful autonomous navigation in all kinds of environments, especially in the duct.
Exploration of duct environment for mapping requires high computational due to the existence of various
junctions and long tunnels. Senarathne et al. (2013) highlight the benefit of occupancy grid map used to store
the map as the map grows larger, processing it can consume more and more computational resources. Various
exploration strategy has been investigated by previous researchers using single and multiple robots. Other
widely used navigation components, such as the simultaneous localization and mapping (SLAM) module
(Grisetti et al. 2007), and the route planner module, are used to evaluate the exploration strategy.
Varnell et al. (2016) suggested RRT-based exploration approaches for quick path planning and exploration
in higher dimensional spaces for computing invariant sets of dimensional systems. Although various path
planning algorithms are available, RRT proves that it can provide an effective solution for solving path
planning in various environments. This paper investigates the effectiveness of RRT and path planning
algorithm for multi-point navigation of a mobile robot in a simulated duct environment.

E- Proceedings of The 5th International Multi-Conference on Artificial Intelligence Technology (MCAIT 2021) [226]
Artificial Intelligence in the 4th Industrial Revolution