Lidar-mmW Radar Fusion for Safer UGV Autonomous Navigation with Collision Avoidance

Unmanned ground vehicles (UGVs) find their applications in wide variety of scenarios including search and rescue operations, warehouse automation to logistics and self-driving cars. In addition, to meet the strict conditions for safer navigation and collision avoidance, multitude of sensors of different modality are fused to get a detailed perception of environment. In this work, we describes a method based on lidar-mmW radar fusion towards safer autonomous navigation. By combining the strength of these two sensor modalities, we present a method that leverages radar's narrow and near-field perception with lidar's wide and far-field perception. This produces a broader field of view (FOV) of the UGV's surroundings which would not have been possible using individually. Additionally, we also implement an end-to-end navigational stack incorporating sensing, perception, planning, and navigation with the goal of safer and collision free navigation. Finally, backed by experiment, we successfully demonstrate that the proposed method can detection object of smaller size placed in close proximity of UGV, and avoid collision while lidar alone would not be able to detect and collide with the object. The detail implementation with source code of work is available at https://github.com/didriknr/Master-thesis-Didrik-Robsrud and a video summarizing the results of this work can be seen at https://youtu.be/M2RLsMUXg3k.