Robotic Visual-Based Navigation Structures Using Lucas-Kanade and Horn-Schunck Algorithms of Optical Flow

This paper aims to present vision-based navigation structures for a wheeled mobile robot using optical flow techniques. The two algorithms of the differential approach are examined and investigated for visual motion in unknown static and dynamic indoor environments. Horn-Schunck (HS) and Lucas-Kanade (LK) algorithms of the optical flow (OF) technique are employed to extract information about the environment surrounding the controlled robot by an installed color camera on the robot platform. Obstacles and objects are identified and detected based on image processing and video acquisition steps for the different tasks of mobile robots: navigation of one robot with static obstacle avoidance, navigation with dynamic obstacle avoidance, and multi-robot navigation with a static obstacle. The proposed control structures are based on motion estimation and decision mechanisms that use the necessary measured variables calculated by optical flow algorithms to carry out the appropriate steering actions to guide autonomously the robot in its workspace. The efficiency of the proposed control structures is tested in 2D and 3D environments using the Virtual Reality Modeling Language (VRML) Toolbox of Matlab. The obtained simulation results are discussed and investigated, and they will be compared to demonstrate the autonomous navigation of mobile robots without any collision with obstacles for these visual-based navigation systems.