Bearings-only localisation of targets from low-speed UAVs

We perform a simulation study to compare the performance of several of single-camera mapping and localisation algorithms for a search and rescue application using low-speed, low-altitude UAVs. We investigate the effects of a range of conditions including target location, nadir angle of the camera and the trajectory of the UAV on three bearings-only SLAM algorithms: Delayed Initialisation (DI), Inverse Depth Point (IDP), and Anchored Homogeneous Point (AHP). Our results show that DI is robust but there can be significant delays before a landmark is initialised. IDP can produce landmark estimates of similar quality, but without the delays. However, this performance can only be achieved through the use of log parameterisations of depth and second order filters. AHP does not yield consistent estimates under any circumstance and is not appropriate for our application.