Autonomous Mobile Robot

Sin, Wye Yuen (2020) Autonomous Mobile Robot. Final Year Project (Bachelor), Tunku Abdul Rahman University College.

Text SinWyeYuen_Full Text.pdf Restricted to Registered users only Download (2MB)

Abstract

This project is to do research on autonomous mobile robot that is capable to localize and navigate autonomously in an indoor situation and create a new cyber-physical autonomous system for the indoor autonomous mobile robot. Most of the factory mobile robots nowadays uses marker-based localisation method, for example the mobile robots created by Chinese company named Alibaba which use in cargo transportation. This method is not highly efficient as the robot scan and follows the path of QR code on the floor with information in it. This does not guarantee the shortest distance from the starting point to the final destination. It only focusses on reaching the destination which is also important for autonomous mobile robot. The main objective of this project is to develop an autonomous mobile robot that can reach its destination with the shortest path and also the shortest time. To achieve the objective, researches on localisation, navigation, path planning, and object recognition is carried out and the best method among all the researches will be implemented into the project. For the experiment, there are 5 experiments that has been carried out to evaluate the performance of the proposed method which is the method that combined with vision-based localisation, A* path planning, and also developed obstacles avoidance algorithm. There are 2 types of experiment that has been carried out, one is real life and another one is simulation. The first experiment is carried out in simulation to determine the Manhattan heuristic method and Euclidean heuristic method, which is more suitable for this project. It turns out, Euclidean heuristic method is more time efficient, but there will be an angular movement which will cause the TurtleBot to collide with wall, hence experiment 2 is carried out in simulation to eliminate this problem. In experiment 2, a checking function is added into the system to command the TurtleBot to move away from the wall when the Lidar sensor sensed that it is too close to the wall. The third experiment is carried out in simulation to evaluate the performance of developed obstacles avoidance algorithm by comparing with Bug 2 algorithm. By utilising the camera attached on the ceiling, as the TurtleBot move closer to the obstacle, the Lidar sensor will sense the obstacle and return signal to the camera to re-capture and re-plan the route map. The fourth experiment is carried out in simulation to evaluate the performance of proposed method by comparing with SLAM method. And lastly, the fifth experiment is carried out to evaluate the performance of real-life proposed method by comparing with both simulation and SLAM method. In conclusion, A* algorithm with Euclidean heuristic shows the shortest time to reach the destination by implementing another checking function. Faster R-CNN able to detect both TurtleBot and exit. Vision based localisation method has been successfully achieved and obstacles avoidance function also successfully achieved by utilising camera and also Lidar sensor.