Corridor Segmentation for Automatic Robot Navigation in Indoor Environment using Edge Devices

Abstract Corridor segmentation is an essential component of the Indoor Navigation System (INS) for robotic activity. It is required for navigation from one place to another in the indoor environment for different applications such as autonomous corridor cleaning, robot-based customer service in hotels, malls, and hospitals. Global Positioning System (GPS) does not work in indoor as it is applicable only in the outdoor environment for navigation and existing beacon-based systems require infrastructure support for mapping the floor plan. The robots with edge devices are resource-constrained. Hence there is a need for memory, computationally, and communicationally efficient solutions. The applications of a robot for these specific task encourages us to solve the navigation problem by visual perception. Image processing based solution cuts the communication cost however it is challenging to stack the right set of operations in proper order for the segmentation. To address such a challenge, this paper builds a right recipe that detects and segments the corridor by real-time image processing and tackles the side obstacles using sensor input. The proposed region-based method using Gray conversion, Canny edge detection, Morphological operations, can process frames at the speed of 6 fps. Its effectiveness is verified by testing on varied real-world illumination scene videos and reported significant precision. With this hardware cum software prototype, an average precision score of 0.82 has been achieved.