The LAPS Group
Our assignment is to find a drone flight path between two points. Normal pathfinding algorithms are extremely slow with very large datasets, and have many issues when used for this purpose. The biggest issue is that the paths they find require post-processing to be flyable by a drone. Without post processing, it will make 90-degree turns, which is not possible on a drone.
To combat this issue, we are using machine learning algorithms to try to not only find a path quickly, but avoid any need for post-processing. Post-processing has the potential of making the generated path sub-optimal, so we want to avoid this step altogether. The hope is that these algorithms also run significantly faster than the traditional algorithms, such that they can be used in real-time on low-spec hardware.
For convenience, we are making a web service where users can calculate paths between points easily. The service caters to developers and end users alike. End-users can compare algorihtms to each other in order to find which algorithm works best for their map. Developers will find this feature useful to continously improve their algorithm.
The Norwegian Mapping Authority provides high resolution heightmaps free to the general public. Some of these laser-scanned heightmaps are sampled every 25cm, and can cover very large areas. Trying to use a traditional pathfinding algorithms on height maps of this resolution is out of the question for the most part.