Target Tracking, Approach, and Camera Handoff For Automated Instrument Placement Max Bajracharya, Antonio Diaz-Calderon, Matthew Robinson, Mark Powell Jet Propulsion Laboratory 4800 Oak Grove Dr. Pasadena, CA 9 1 109, USA firstname.lastname@jpl.nasa.gov Abstract-This paper describes the target designation, track- ing, approach, and camera handoff technologies required to achieve accurate, single-command autonomous instrument placement for a planetary rover. It focuses on robust track- ing integrated with obstacle avoidance during the approach phase, and image-based camera handoff to allow vision-based instrument placement. It also provides initial results from a complete system combining these technologies with rover base placement to maximize arm manipulability and image- based instrument placement. 3 APPROACH 4 HANDOFF 5 ROVER BASE AND INSTRUMENT PLACEMENT 6 INITIAL RESULTS AND VALIDATION 7 CONCLUSION 8 ACKNOWLEDGEMENTS In order to increase science return or decrease surface time, future Mars rover missions will need to be able to desig- nate, safely approach, and accurately place an instrument on a target in a single sol (Mars day). This paper focuses on the approach, tracking, and handoff technologies required to achieve this capability and describes a complete end-to-end demonstration of the integrated system. Because the communication window to a Mars rover is lim- ited during the course of a day, autonomous operation of the rover is essential to increasing its productivity. Currently on the Mars Exploration Rover (MER) mission, the process of designating a target, approaching it, and placing an instru- ment on it takes a minimum of 3 communication cycles (3 sols) (one to approach it, one to refine the rover placement, and one to place the instrument). This paper describes the technology to allow a scientist to designate a target (generally 0-7803-8870-4/05/$20.00/~ zws IEEE IEEEAC paper # 1631 a location on a rock) in a panorama with one command and then autonomously approach the target avoiding any hazards, refine the rover placement to maximize the manipulability of the arm, and place the instrument on the target. All experiments described were carried out in JPL's Mars Yard (an outdoor terrain simulating the rock densities at the Viking 1 and 2 landing sites on Mars) on the Rocky 8 re- search rover. The Rocky 8 rover is a MER-like rover, with a six-wheel rocker-bogie suspension, a fixed padtilt mast with panorama and navigation stereo camera pairs, and front and rear body-mounted hazard stereo camera pairs. However, un- like the MER rovers, it has full six-wheel steering allowing it to execute crab maneuvers. The rover and it's environment can be seen in Figure 1. Figure 1. The Rocky 8 rover in the JPL Mars Yard The experiment starts with the rover taking a panorama of the terrain with its mast cameras pointing out to ten meters on the ground. After taking the panorama, the images are down- linked to Maestro, the ground software system. Maestro is then used to designate a target in one of the images and gen- erate a command to approach and place an instrument on that target. After receiving the command, the rover calculates an initial goal position such that the target is in the workspace of the arm, and uses the Morphin navigation algorithm (similar to the Gestlat algorithm being used on MER) to avoid ob- stacles and achieve the goal location. During approach, the rover also tracks the target in its mast cameras. When the