Autonomous Rover Traverse and Precise Arm Placement on Remotely Designated Targets
- Saturday, 01 September 2012
This software controls a rover platform to traverse rocky terrain autonomously, plan paths, and avoid obstacles using its stereo hazard and navigation cameras. It does so while continuously tracking a target of interest selected from 10–20 m away. The rover drives and tracks the target until it reaches the vicinity of the target. The rover then positions itself to approach the target, deploys its robotic arm, and places the end effector instrument on the designated target to within 2–3-cm accuracy of the originally selected target.
This software features continuous navigation in a fairly rocky field in an outdoor environment and the ability to enable the rover to avoid large rocks and traverse over smaller ones. Using pointand- click mouse commands, a scientist designates targets in the initial imagery acquired from the rover’s mast cameras. The navigation software uses stereo imaging, traversability analysis, path planning, trajectory generation, and trajectory execution. It also includes visual target tracking of a designated target selected from 10 m away while continuously navigating the rocky terrain.
Improvements in this design include steering while driving, which uses continuous curvature paths. There are also several improvements to the traversability analyzer, including improved data fusion of traversability maps that result from pose estimation uncertainties, dealing with boundary effects to enable tighter maneuvers, and handling a wider range of obstacles.
This work advances what has been previously developed and integrated on the Mars Exploration Rovers by using algorithms that are capable of traversing more rock-dense terrains, enabling tight, thread-the-needle maneuvers. These algorithms were integrated on the newly refurbished Athena Mars research rover, and were fielded in the JPL Mars Yard. Forty-three runs were conducted with targets at distances ranging from 5 to 15 m, and a success rate of 93% was achieved for placement of the instrument within 2–3 cm of the target.
This work was done by Issa A. Nesnas and Mihail N. Pivtoraiko of Caltech, Alonzo Kelly of Carnegie Mellon University, and Michael Fleder of MIT for NASA’s Jet Propulsion Laboratory.