Abstract
The robots that can perform cleaning in the staircase region are gaining interest and are lined up to release in commercial robot space. Even though several precedents in the literature reported the development of staircase cleaning, their primary focus was on autonomous staircase ascending, which curtails their performance by increasing the power consumption and constraints to achieve full area coverage in a multistory building. The main objective of this research article is to develop a novel autonomous descending staircase-cleaning robot named sTetro-D. The developed robot uses a Deep Convolution Neural Network (DCNN) to autonomously detect a descending staircase, approach it, and perform maximum area coverage. This article presents the technical details of the developed robot and its DCNN-based autonomous descending staircase coverage ability. Also, the developed system was validated in terms of accuracy in detecting the descending staircase, reaching the stairs, and performing maximum area coverage through conducting experimental trials in two real-world scenarios. In all considered scenarios, the developed robotic platform exhibits significantly superior performance in detecting the descending staircase with an average accuracy of 85%, successfully approaching the descending staircase, and achieving 98% area coverage.