|
|||||||||||
The suspension structure of the CRAB is based on two parallel bogies linked with pivot joints on the middle wheel and on top. It is a passive structure, i.e. no motors are used to modify its shape, and it adapts naturally to the terrain.
Such a structure has great advantages compared to an active structure:
Here is a brief description of the software running on the CRAB. The divers parts are described below:
The E* algorithm is a path planner which is based on a weighted region approach. Thus, the underlying technique is expressed within the continuous domain, and corresponds to a wavefront propagating from the goal, toward the rover. The environment in which the rover is evolving is represented as a grid, constituted of node, or cells.
| The environment is represented within a grid. The rover position is marked with the cross (on the left hand-side) while the goal position is depicted with the circle on the right hand-side. | |
| Two areas are identified, one in white and the other with a greyish color. In this situation, the behavior of the rover is better on the white area and it means a wavefront evolving faster in this area than in the greyish one. | |
| Here the gradient descent of the navigation function is depicted. The resulting trajectory is provided as a vector corresponding to a series of waypoint leading to the goal. The trajectory is called trace |
More information about the E* algorithm can be found there.
| The trace can be followed as described here. The CRAB rover, can be represented as a simple differential robot to which constraints are added. Thus, the following equations can be used: |
The idea is to allow the rover to learn online (while moving) to link the rover-terrain characterics (RTC), called local data, with remote data corresponding to the terrain appearance (e.g. acquired by a camera). Such knowledge would allow the rover to estimate its RTC in a similar situation.
The algorithm has to be able to decide whether a terrain is different from those already met, in its appearance, its characteristics or both. This corresponds to unsupervised learning.
The CRAB measures both local and remote data and computes their corresponding features, Fl and Fr respectively.
Samples associating Fl and Frt can be determined, where Frt corresponds to a delayed Fr, matching Fl.
Based on these samples, the probability distribution function (pdf) is learned, using ProBT.
P(Fl|Frt)
The same pdf is then used to asses the RTC (corresponding to Fl) of areas of which Fr only is known. These correspond to terrain ahead of the rover.
The CRAB I is the first version that was built in 2005. The length of the parallel bogies is equal, affecting unequally the mass repartition on the wheels.
The design of the CRAB I was upgraded in 2006 for this one, which has a better suspension mechanism, providing an equal mass repartition on the wheels.
In 2007, several elements of the CRAB II were modified. Among these, the most important changes lie in the addition of tactile wheels and the change of the steering mechanism.
Wichtiger Hinweis:
Diese Website wird in älteren Versionen von Netscape ohne
graphische Elemente dargestellt. Die Funktionalität der
Website ist aber trotzdem gewährleistet. Wenn Sie diese
Website regelmässig benutzen, empfehlen wir Ihnen, auf
Ihrem Computer einen aktuellen Browser zu installieren. Weitere
Informationen finden Sie auf
folgender
Seite.
Important Note:
The content in this site is accessible to any browser or
Internet device, however, some graphics will display correctly
only in the newer versions of Netscape. To get the most out of
our site we suggest you upgrade to a newer browser.
More
information