Loading...
Mobile Robot Navigation in Dynamic Environments Using an Improved RRT* Approach
Mohammed, Hussein Ali
Mohammed, Hussein Ali
Date
2018-11
Author
Advisor
Type
Thesis
Degree
Citations
Altmetric:
Description
A Master of Science thesis in Mechatronics Engineering by Hussein Ali Mohammed entitled, “Mobile Robot Navigation in Dynamic Environments Using an Improved RRT* Approach”, submitted in November 2018. Thesis advisor is Dr. Lotfi Romdhane and thesis co-advisor is Dr. Mohammad Jaradat. Soft and hard copy available.
Abstract
A very prominent area in the field of Mechatronics is robot navigation and path planning. This area deals with the problem of autonomously calculating the least cost path in a provided environment, whether it is static or dynamic, and navigating the robot platform though this environment. Over the last decade, since the year 2001, there have been major breakthroughs in this field after LaValle introduced his revolutionary algorithm, the Rapidly-Exploring Random Tree (RRT) approach. Later, in the year 2011, Karaman introduced his novel modification to RRT which he called the Rapidly- Exploring Random Tree Star (RRT*). The main advantage of RRT* is its effectiveness and robustness in finding the path to the target and its probabilistic completeness property which guarantees the best theoretical path if given enough run time. However, RRT* still suffers from long processing times to provide paths with satisfactory quality in terms of cost and smoothness. Having said that, in this thesis we propose an improved version of the RRT* algorithm which will address the issue of long processing times and sub-par path quality. This new method is called Rapidly-Exploring Random Tree Star Normal (RRT*N). The presented method can handle static and dynamic obstacles in 2D and 3D environments. This improved method uses a Gaussian probability distribution to generate new nodes which have a higher probability of being generated along the vector pointing from the starting point to the goal point, which results in a tree centered on the line joining the robot to the target. It is shown that this method can be three times faster in finding the path to the target in static scenarios, and upto 20 times faster in dynamic environments compared to RRT*. Furthermore its rate of achieving satisfactory paths is consistently more than 95% while maintaining similar path quality. For instance in 250 trials of the presented static scenario, RRT*N had an average processing time and average path length of 38.46 seconds and 240.99 units, respectively. Meanwhile, RRT* resulted in 121.58 seconds and 259.11 units. This work is based on an extensive literature review to validate this work’s novelty and the simulation and experimental results presented show the robustness of the proposed RRT*N method.