PurposeThe purpose of this paper is to present a novel approach to the design of an autonomous Unmanned Aerial Vehicle (UAV) to aid with the internal inspection and classification of tall or large structures. Focusing mainly on the challenge of robustly determining the position and velocity of the UAV, in three dimensional space, using on‐board Simultaneous Localisation and Mapping (SLAM). Although capable of autonomous flight, the UAV is primarily intended for semi‐autonomous operation, where the operator instructs the UAV where to go. However, if communications with the ground station are lost, it can backtrack along its path until communications are re‐established.Design/methodology/approachA UAV has been designed and built using primarily commercial‐off‐the‐shelf components. Software has been developed to allow the UAV to operate autonomously, using solely the on‐board computer and sensors. It is currently undergoing extensive flight tests to determine the performance and limitations of the system as a whole.FindingsInitial test flights have proven the presented approach and resulting real‐time SLAM algorithms to function robustly in a range of large internals. The paper also briefly discusses the approach used by similar projects and the challenges faced.Originality/valueThe proposed novel algorithms allow for on‐board, real‐time, three‐dimensional SLAM in unknown and unstructured environments on a computationally constrained UAV.