Algorithm Implementation: Safe Landing Zone Identification

Algorithm Implementation:Safe Landing Zone

Identification

Presented by

Noah Kuntz

Problem Under Investigation

UAV flying in unknown terrain Typically helicopter

Map terrain Vision LIDAR

Identify landing sites Hazard free Terrain is suitable Large enough to fit UAV

Papers Reviewed

“Towards Vision-Based Safe Landing for an Autonomous Helicopter”Pedro J. Garcia-Pardo, Gaurav S. Sukhatme and James F. MontgomeryRobotics and Automated Systems 2001

“Vision Guided Landing of an Autonomous Helicopter in Hazardous Terrain”Andrew Johnson, James Montgomery and Larry MatthiesInternational Conference on Robotics and Automation 2005

“The JPL Autonomous Helicopter Testbed: A Platform for Planetary Exploration Technology Research and Development”James F. Montgomery, Andrew E. Johnson, Stergios I. Roumeliotis, and Larry H. MatthiesJournal of Field Robotics 2006

“Lidar-based Hazard Avoidance for Safe Landing on Mars”Andrew Johnson, Allan Klumpp, James Collier and Aron WolfAIAA Journal of Guidance, Control and Dynamics 2002

Algorithm Selection – Option 1

Source: “Towards Vision-Based Safe Landing for an Autonomous Helicopter”

Strengths: Using vision requires low weight camera Processing power required is not high

Weaknesses: Assumes flat underlying terrain, which severely limits practical

usage. Assumes high contrast between obstacles and underlying

terrain, risks failure to detect some objects. Could pick to land on an obstacle if it is large enough.

Algorithm Selection – Option 2

Source: “Vision Guided Landing of an Autonomous Helicopter in Hazardous Terrain,” “The JPL Autonomous Helicopter Testbed: A Platform for Planetary Exploration Technology Research and Development”

Strengths: Using vision requires low weight camera Considering slope as well as roughness of underlying terrain

produces a robust cost map in most terrain conditions

Weaknesses: Requires extensive vision processing

Algorithm Selection – Option 3

Source: “Lidar-based Hazard Avoidance for Safe Landing on Mars”

Strengths: Using LIDAR produces accurate terrain maps at wide

angles and moderate processing power Considering slope as well as roughness of underlying

terrain produces a robust cost map in most terrain conditions

Weaknesses: Most LIDAR systems are heavy

Algorithm Choice – Option 3

Reasoning: Most robust in terms of accurately detecting obstacles Other than the sensor, identical to the 2nd best choice,

option 2

Mitigation of Weaknesses: Helicopter must be capable of lifting sufficient weight

Overview of Algorithm Implementation

Digital Elevation Map SICK data is interpreted and flattened Pose and position at each SICK scan point is

recorded from UAV autopilot Elevation map is generated by correlating the

scanned depths with the position and pose at which they were recorded

Safe Landing Zone Elevation map is analyzed in units the size of the lander

footprint, incremented at a fraction of the footprint size Planes are fit to each unit using least mean squares Slope of each plane is calculated from the center of the region Roughness is calculated as the difference between the original

map and the fitted map Roughness and slope maps are normalized and added to

produce the cost map Cost map is blurred to prevent landing on a good zone adjacent

to a highly unsafe zone Landing zone is found in resulting image as the minimum cost

point

Overview of Algorithm Implementation

Hypothetical terrain was generated with a graphics program

Generated Terrain

Fit planes to the original data

Landing Zone Algorithm Step 1

Original: Fitted:

Find the slopes of the fitted planes Find the roughness based on the difference between the

original maps and the fitted planes

Landing Zone Algorithm Step 2,3

Normalized Slope Cost: Normalized Roughness Cost:

The cost map is produced by the adding the normalized roughness and slope cost, then blurred with a 3x3 Gaussian

Landing Zone Algorithm Step 4

Total Cost: Blurred Cost:

The landing zone is chosen as the minimum cost point

Landing Zone Algorithm Step 5

Map With Landing Zone: Image With Landing Zone:

More Examples

More Examples