Landing Zone Analysis for Autonomous Medevac

Three Trees Flat Field

This basic scenario shows the functionality of the system.  The imaginary casualty is located near a long flat field with few obstacles.  We performed a single pass over the field at 100 meters to survey the area and find a landing spot.

Click on any of the images for a more detailed picture.

Step 1: Survey a Point Cloud

The target region is surveyed by flying low over the target region.  Here, we did a single pass over an open field at an altitude of about 100 meters.  Each measurement of the laser range finder returns a "point" that is registered in a global coordinate frame using the inertial navigation system.  The survey results in a collection of points called a "point cloud," which represents the shape of the terrain.

In the picture below, the point cloud is color-coded for altitude.  Bands in the color are the same as reading a topographic map.



Step 2: Rough Terrain Analysis

 After a point cloud is acquired, the system divides the map up into a grid of three-meter sections.  Each section is categorized for its fitness as a potential landing site.  The algorithm considers slope, smoothness, and distance from other unsuitable areas.  For example, a tree will have points that vary widely between branches, and will show up as very rough.  Flat ground will be both smooth and level. In the picture below, areas that may be suitable for landing are colored on a gradient between yellow and green, with green being better.  The other colored squares have been deemed unsuitable for landing and will not be considered in further processing.    

Step 3: Fine Terrain Analysis

 Potential landing sites found by the rough analysis are then passed on for further inspection.  The fine terrain analysis fits a three-dimensional model of the helicopter at each landing site in many different rotation orientations.  The score of a certain position and orientation is a function of how close the body of the helicopter comes to obstacles, how good the contact of the landing gear is with the terrain, and the angle of the vehicle when sitting on the ground at that position. In the illustration below, the parallel lines show the position of the landing skids on the ground for the best-ranked orientation at that position.  Again, the yellow->green gradient shows the score, with green being best.  Red skid marks show locations that have been found unsuitable for landing. A good landing zone will ideally show a cluster of many good landing spots.  Also notice, that the algorithm will reject any landing spots that are within a certain radius of the casualty, as denoted by the red circle surrounding the casualty icon.   

Step 4: Landing Site Selection

After the candidate landing sites have been closely examined and ranked, the system then decides where to actually land.  An ideal location will not only have a cluster of appropriate landing sites surrounding it, but it will also be close and accessible to the casualty.  The final decision is a trade off between proximity to the casualty, and the suitability of the landing location.  The decision is shown by the white concentric circles. Notice in the figure below, that the system has chosen to land slightly farther away from the casualty than it could have, in order to reduce the risk associated with landing closer to the obstacles.