Autonomous Navigation Support from Real-Time Visual Mapping

Important results from the project

Focussing on vision-based aerial localisation without satellite-based navigation, the goal was to gain and transfer knowledge and build a basis for development of vision-based navigation systems, mainly through collaboration between KTH and Spacemetric. Overall, the project achieved its goals: · Innovative methods were developed and presented at three international conferences; · Several articles were published; · The acquired knowledge has provided a basis for further development of vision-based navigation; and · The work gained interest at national and European level.

Expected long term effects

The project generated three scientific publications presented at international conferences, and two master theses. An innovative method for low-altitude navigation over forested landscapes was developed, exploiting vegetation 3D structures (patent pending). Spacemetric gained knowledge providing a basis for future development of vision-based navigation. A follow-on project continues with funding from FMV. A parallel Vinnova project reported on use cases and non-technical challenges of future autonomous aerial navigation based on interviews with Swedish entities.

Approach and implementation

The project was carried out in close collaboration between KTH and Spacemetric. Daniel Sabel, a Spacemetric employee, undertook PhD studies at the Division of Robotics, Perception and Learning. Two project-specific master theses were carried out under the supervision of the project. I-CONIC Vision, an SME, was engaged for real-time generation of 3D data from UAV images. Data collection campaigns were carried out by Spacemetric with the project’s own UAV. The project also engaged with other industrial and governmental entities in Sweden and in Europe, such as NFFP Cluster Leaders.