A lander can autonomously determine the areas within its robotic arm’s workspace that have the least risk for digging hazards.
This software evaluates the terrain within reach of a lander’s robotic arm for dig hazards using a stereo pair of cameras that are part of the lander’s sensor system. A relative level of risk is calculated for a set of dig sectors. There are two versions of this software; one is designed to run onboard a lander as part of the flight software, and the other runs on a PC under Linux as a ground tool that produces the same results generated on the lander, given stereo images acquired by the lander and downlinked to Earth.
Onboard dig hazard assessment is accomplished by executing a workspace panorama command sequence. This sequence acquires a set of stereo pairs of images of the terrain the arm can reach, generates a set of candidate dig sectors, and assesses the dig hazard of each candidate dig sector.
The 3D perimeter points of candidate dig sectors are generated using configurable parameters. A 3D reconstruction of the terrain in front of the lander is generated using a set of stereo images acquired from the mast cameras. The 3D reconstruction is used to evaluate the dig “goodness” of each candidate dig sector based on a set of eight metrics. The eight metrics are:
1. The maximum change in elevation in each sector,
2. The elevation standard deviation in each sector,
3. The forward tilt of each sector with respect to the payload frame,
4. The side tilt of each sector with respect to the payload frame,
5. The maximum size of missing data regions in each sector,
6. The percentage of a sector that has missing data,
7. The roughness of each sector, and
8. Monochrome intensity standard deviation of each sector.
Each of the eight metrics forms a goodness image layer where the goodness value of each sector ranges from 0 to 1. Goodness values of 0 and 1 correspond to high and low risk, respectively. For each dig sector, the eight goodness values are merged by selecting the lowest one. Including the merged goodness image layer, there are nine goodness image layers for each stereo pair of mast images.
There are three modes of operation for the ground tool version of the software:
1. View image, dig sector, and “digability” data products generated onboard the lander.
2. Given a set of raw images from a stereo pair of mast cameras, generate image, dig sector, and dig hazard products identical to what would be generated onboard the lander and view them.
3. Given a set of image products downlinked from the lander, generate dig sector and dig hazard products identical to what would be generated onboard the lander and view them. The ground tool can be used to view the 3D reconstruction of the terrain. The mouse buttons can be used to rotate the 3D model of the terrain and zoom in and out. Drop-down menus enable the user to display the dig sectors, one of the eight goodness image layers, and the merged goodness map layer. When viewing a goodness map layer, the dig sectors within the 3D reconstruction are color-coded. Green sectors are safe for digging. The colors between green and red correspond to the increasing level of risk.
This work was done by Arturo L. Rankin and Ashitey Trebi-Ollennu of Caltech for NASA’s Jet Propulsion Laboratory.
This Brief includes a Technical Support Package (TSP).
Dig Hazard Assessment Using a Stereo Pair of Cameras (reference NPO-48448) is currently available for download from the TSP library.
Please Login at the top of the page to download.