Sensor Simulation and Environment Mapping using UWB Radar for Industrial Drone Inspection

Abstract

The use of unmanned aerial drones for inspection of industrial environments have great potential in lowering both cost and risk. Development of such systems are going on at full speed, but many hurdles remain, especially that of creating robust sensors for object detection. The main obstacle is to find a sensor for object detection that is small, cheap, accurate and can fit on a small flying drone. Radar is proposed for this purpose, but little testing has been performed on smaller flying drones to either confirm or deny this prospect. New tools will need to be developed to increase the development process. In this thesis, a simulation framework has been implemented that allows for real-time testing of radar sensors within a 3D environment. The system is based on the software packages of ROS and Gazebo, using the ray-tracing technique to simulate the effects of electromagnetic waves on the environment. A radar sensor model was developed in Gazebo based on a 3D scanning lidar. The radar performed well through designed simulations, providing similar output to that of the radar sensor it simulates. In addition, a mapping technique called Occupancy Grid Mapping was implemented to complement the radar sensor, expanding its application area. The mapping technique showed good results, mapping clustered environments with low uncertainty.