Real-time and robust odometry estimation using depth camera for indoor micro aerial vehicle

Real-time, robust and precise estimation of a robot´s ego-motion is a crucial requirement for higher level tasks like autonomous navigation. In this paper, a real-time and robust odometry estimation system for indoor micro aerial vehicle (MAV) is developed by only using the point cloud generated from the depth camera. First, local surface normal features are used to select points with most constraints. Then, an improved iterative closest point method is used to calculate the relative transformation, which is robust against sensor noise and outliers. To further improve the robustness of the estimation, this paper constructs a local point cloud map and compares current point cloud to the local map. Besides, ground plane is also used to simplify the 6DOF estimation problem as a 3DOF estimation problem, which not only reduces the drift but also improve the estimation speed. To validate the performance of the proposed method, we compared our method to several visual odometry methods using different kind of real dataset. The experiment results show that depth only odometry can achieve similar estimation results as state of the art visual odometry methods.