Autonomous Obstacle Avoidance and Control Using Voxel Segmentation of 3D Lidar Data

The purpose of this work was to determine if a single 3D lidar sensor could provide enough data to conduct obstacle detection and avoidance for a small ground-based autonomous vehicle in an indoor environment. This work was based on previous Naval Postgraduate School work with simultaneous localization and mapping using a 2D lidar sensor and a 3D time of flight camera. A voxel-based point cloud filtering method was used to interpret data and classify objects as large, small, or negative. The data was then used as an input to a control algorithm using a potential field control model to navigate around the identified obstacles. The classification and control algorithm was proven successful through four separate experiments, and a definition for a small object was developed. Areas for future study were identified to include the development of a localization method using a single 3D lidar sensor, the implementation of the obstacle avoidance algorithm on an autonomous platform with six degrees of freedom, and the development of a path planning algorithm based on an initial point cloud.