Development of Control System of Autonomous Cleaning Robot for Large Indoor Area

Abstract

This thesis presents the development and simulation-based validation of a control system for an autonomous cleaning robot designed for large indoor areas. The selected test environment is the main hall and corridor area of the GMIT first floor, which includes long corridors, curved boundaries, irregular wall geometry, and fixed internal obstacles representing a realistic and navigationally complex indoor space.

The system is structured as a two-stage simulation workflow implemented in CoppeliaSim EDU and MATLAB R2025b. In the first stage, a Pioneer P3DX differential-drive robot equipped with a 360-degree LiDAR sensor assembly performs boundary-following exploration, generating an occupancy grid map that is exported for external processing. In the second stage, MATLAB processes the raw map through noise removal, free-space extraction, obstacle inflation, and edge polyline extraction, then generates a complete cleaning mission combining boundary edge-cleaning paths, interior zig-zag coverage paths, safe transition routes, and a return-to-start path. The mission is executed in CoppeliaSim using a pure pursuit path tracking controller, with cleaning coverage tracked continuously throughout execution.

The system achieved a final cleaning coverage of 99.4 percent, exceeding the predefined success threshold of 95 percent. This result was obtained under ideal simulation conditions in which the map and execution environment are identical and localization error is absent. These conditions represent an upper bound on expected performance, and physical deployment would introduce additional factors including sensor noise, odometry drift, and environmental changes not captured in simulation. A proposed hardware deployment architecture for transferring the developed system to a physical Roborock-based platform is included as a foundation for future implementation and validation.