The activity discussed intends to increase the capability of Unmanned Aerial Vehicles by incorporating information from a vision sensor into the usual data stream for sensing and control. A particular interest is keeping the implementation simple and the amount of information minimal. It is shown that using only the velocity direction in the camera frame leads to significant improvement in the accuracy of the navigation state estimator. With this improvement, it is then possible to implement on-line parameter estimation techniques to make the controller robust against (a) large modelling errors in the dynamics used to derive the nominal autopilots, and (b) loss of GPS information during flight. As the eventual goal is cooperative activity, the effort also looked at using minimal vision information to follow a lead vehicle, and implementations of cooperative observations of ground targets.
Research in Image-Based Cooperation for Autonomous Conventional Aerial Vehicles
2008
22 pages
Report
No indication
English
British Library Conference Proceedings | 1995
|IMAGE SEGMENTATION BASED EMERGENCY LANDING FOR AUTONOMOUS AND AUTOMATED UNMANNED AERIAL VEHICLES
British Library Conference Proceedings | 2022
|