Autonomous Detection of Stair Dimensions for the Motion Planning of Stair Climbing Robots

The need for service robots to help humans in daily tasks is increasing globally. Whether in an indoor or outdoor environment, service robots need to plan their motion to climb stairs autonomously. So, in this paper, we propose a new approach using a depth camera to determine the features of stairs which are required as an input to autonomous motion planning for stair climbing robots. The proposed approach uses a pinhole camera model to compute the point cloud of the perceived environment. After that, determining the normal vector of the points in the point cloud helps to organize the data into two clusters: horizontal and vertical. For each of the two clusters, the Euclidean distance algorithm is used to generate a class for each step of the stairs. Finally, the Maximum Likelihood Estimation Sample Consensus (MLESAC) algorithm is used to fit a plane on each class. Thus, the dimensions of the stairs are determined accurately by subtracting two subsequent plane equations from each other.