Autonomous mapping and path planning for mobile robots in unknown environments using karto slam / MST Ishrat Jahan

Faculty of Electrical & Electronics Engineering Technology

Thesis (Master of Science) -- Universiti Malaysia Pahang - 2026

Include bibliographical reference

Path planning is a very essential aspect in the field of autonomous robotics in order to enable the mobile robots to navigate safely and accurately in the complex and dynamic environment. The path-planning algorithm is needed to take into consideration various working parameters, including minimizing search nodes, button cost of path, and time complexity. Conventional grid-based (also known as grid planners) such as the A* algorithm has performed well because of the decision-making process which is based on heuristics. Nevertheless, when applied in the real-world, disorganized, and unfamiliar unknown environments, the use of traditional procedures is insufficient because of the constraints of flexibility and reliability of odometry. This reasearch addresses the challenge of mapping and mobility planning in uncertain unknown environments through a Karto SLAM method based on high mapping accuracy and adaptability in challenging environments. The objectives of the study are: (i) exploration and optimization of estimation and path planning of mobile robots in unfamiliar environments, (ii) comparison of the Karto SLAM with other popular techniques of SLAM, including Gmapping, Hector SLAM, Cartographer, and Generic SLAM, and (iii) real-time navigation can be improved by the use of the combination of new advanced planning algorithms, e.g., Dijkstra, A*, and TEB. Simulation environment was built with the help of Turtlebot3 on ROS2 platform with Gazebo dynamic model and RViz visualization. Such an arrangement allowed building precise two-dimensional (2D) and three-dimensional (3D) maps, which represent a solid base of future real-world robotics. The simulation-based experimentation results obtained prove that Karto SLAM is always better than other SLAM methods, the mapping accuracy of Karto SLAM improves by 12% relative to Gmapping and Hector SLAM, especially in a restricted odometry environment and cluttered environment. Karto SLAM reduced the odometry reliability, as evidenced by the drift percentage in the course of navigation, by 15% of that when other approaches were used. Furthermore, although A* is more efficient in finding the path, Dijkstra algorithm gave 25% cutoff on the length of the traveled path giving a smooth path with better optimization particularly when used in conjunction with global planning strategies. The TEB (Timed Elastic Band) algorithm helped in real-time local navigation to make the robot more responsive to dynamic obstacles and thereby enhance safety and operational efficiency with a 20% enhancement in real-time obstacle avoidance. The overall amount of time of the robot to reach the target position was also cut by 18% with Dijkstra compared to A* and the percent change in the actual length of the path compared to the hypothetical minimum was kept to a minimum of less than 5% using Karto SLAM and Dijkstra. The study demonstrates the importance of extensive parameter optimization and realistic validation in the real-life scenarios. In general, the given model provides a flexible structure of robotic navigation that does not rely on a high cost of the path but enhances the environmental awareness and the trajectory planning, which is a major advancement in the field of autonomous mobile robots.