Page 242 - The-5th-MCAIT2021-eProceeding

2. Methodology

2.1 Duct Environment

A real air duct simulation is used as a reference to develop the duct simulation environment. We develop
an area of 8 meters (width) x 9.5 meters (length). The hallway size varies from 0.5 to 1-meter consist of the
main hallway, connection, and blower channel. There are four junctions involves in this simulation setup. The
starting point was set in the third row of the channel. Six reference points are defined in x, y, z coordinates as
[-3.9,5.8,0,1.1,5.8,0,-3.9,3.2,0,1.2,3.2,0,-4,-1.7,0,1.1,-1.7,0].

2.2 Hardware requirement

Turtlebot3 Waffle is used as the main robot platform to navigate inside this tunnel due to the suitability of
its height (15cm) and width (30cm). The localization of the robot utilizes the laser range finder, which ranges
up to a 5-meters radius. Turtlebot3 Waffle is embedded with 2 wheels motor that helps to balance its
movement. Turtlebot3 Waffle camera is used for visualization during navigation.

2.3 Mapping

This project implements a known environment where the mapping of the environment is conducted before
the path planning algorithm is initiated. The mapping process is performed using the teleoperation function.
The map is further improve using the Image Editor tool to clean up the unnecessary and unconnected lines.
This map has been tested during the 2D navigation process to ensure the robot can move in all duct areas.

2.4 RRT Algorithm

RRT is a path planning algorithm (Zhang et .al 2018) that samples space using randomly generated points.
Random points are used to extend edges in a tree-like structure, which consists of nodes and edges. One
possible mode of RRT-based exploration is to make robots follow the above-mentioned tree structure as the
tree structure grows. In this project, the RRT algorithm is used to explore the path planning to navigate to all 6
pre-set reference points. The RRT algorithm is chosen because it is biased towards unexplored regions,
encouraging the tree to detect frontier points.
In RRT, the exploration mechanism is achieved in three modules: the frontier detector, filter, and robot
task allocator, as shown in Fig. 1. The frontier detector is responsible for detecting frontier points and passing
them to the filter module. The filter module clusters the frontier points and stores them. The filter module also
deletes invalid and old frontier points. The task allocator module receives the clustered frontier points from
the filter module and assigns them to a robot for exploration.

Fig. 1. RRT structure

E- Proceedings of The 5th International Multi-Conference on Artificial Intelligence Technology (MCAIT 2021) [227]
Artificial Intelligence in the 4th Industrial Revolution