The Mobility and Robotic Systems Section develops control technology for many types of surface mobility, and implements it on real robotic vehicles -- from fast off-road vehicles to slower, lower-powered, off-planet Mars Rovers. Wheeled systems include the Mars Exploration Rovers (MERs) and Mars Pathfinder rover, Sojourner; their progenitors from our technology research, like the Rocky- and FIDO-series rovers; and terrestrial HMMV and Unmanned Ground Vehicle (UGV) platforms instrumented for autonomous driving. Other mobility research platforms include tracked, legged, and hybrid systems that enable even greater mobility in sloped or slippery terrains.
Several technologies contribute to robust, reactive control systems that provide greater capability while driving in unknown terrain. Active traction control helps reduce longitudinal slip by dynamically adjusting the drive rate of each wheel based on each motor's response to the local terrain. As shown in Figure 1, Visual Odometry eliminates the uncertainty in position and attitude arising from vehicle slip or inertial-sensor drift by autonomously tracking features in images taken while driving. And Visual Tracking techniques provide a robust way to reach a goal even in the presence of large position uncertainty.
 |
 |
Fig. 2: Graphic showing MER navigation technique of evaluating terrain traversability along discrete arcs in the imaged terrain. ( Click here for full movie.) |
|
A variety of predictive technologies keep vehicles safe by avoiding potentially dangerous areas. Passive Stereo Vision or active LIDAR systems provide a geometric model of the nearby terrain in the form of a cloud of 3D points. Figure 2 shows how a traversability analysis is performed over the geometric model, using robust estimation techniques to find step obstacles (e.g., rocks or ditches), high-tilt areas, and rough terrain. Additional terrain analyses performed on multispectral image texture or color has been demonstrated on research vehicles and provides knowledge about the type of terrain in addition to its geometry.
While the capabilities listed above all run on computer systems embedded on individual robots, we also provide systems for terrain visualization, drive planning, and drive simulation to aid humans in the operation of robotic mobility systems. High-resolution
visualization of terrain is critical for planning effective drives toward specific goals. High-fidelity rover simulations provide a useful tool for "test driving" a given set of commands across a newly-visualized terrain.
We also provide technologies that greatly accelerate the development of specialized mobility systems. New designs emerge from extensive field and laboratory experience with rough terrain or complex man-made environments. High-fidelity robot and environment simulations can enable early development of control algorithms while vehicles are still being designed or built. Unified software architecture for robots provides a framework for the development of new capabilities, and implements many component technologies like stereo-vision processing and terrain assessment. These components may then be readily incorporated into new vehicles or licensed for use with collaborators outside JPL.
