Ramwell, APretorius, ABotha, Natasha2026-01-052026-01-0520252261-236Xhttps://doi.org/10.1051/matecconf/202541704011http://hdl.handle.net/10204/14546Unmanned aerial vehicles are a stimulating and practically promising solution for automating tasks in inaccessible environments. An appealing minimal sensor suite for localisation (the first challenge in automation) is the combination of monocular camera and inertial measurement unit. Although visual localisation is a well-developed field, fiducial marker-less methods remain computationally expensive. In this study, a simple, lightweight visual–inertial state estimator is evaluated in physical experiments, achieving a mean positional and angular error of within 15 cm and 5°, respectively, in non-ideal conditions (periodic image loss and slow update rates). This study’s contribution is thus the development, description, and benchmarking of a stripped-down, practical localisation solution that is simple to understand and interface with. After some refinement, the system is expected to serve as a skills development tool and foundation for higher-level navigators.FulltextenUnmanned aerial vehiclesMonocular camerasVisual localisationArticleN/A