Little Dog learning of tractive and compressive terrain characteristics
2011
In recent years research into legged locomotion across extreme terrains has increased. Much of this work was
done under the DARPA Learning Legged Locomotion program that utilized a standard Little Dog robot platform
and prepared terrain test boards with known geometric data. While path planing using geometric information
is necessary, acquiring and utilizing tractive and compressive terrain characteristics is equally important. This
paper describes methods and results for learning tractive and compressive terrain characteristics with the Little
Dog robot. The estimation of terrain traction and compressive/support capabilities using the mechanisms and
movements of the robot rather than dedicated instruments is the goal of this research. The resulting characteristics
may differ from those of standard tests, however they will be directly usable to the locomotion controllers
given that they are obtained in the physical context of the actual robot and its actual movements. This paper
elaborates on the methods used and presents results. Future work will develop better suited probabilistic models
and interwave these methods with other purposeful actions of the robot to lessen the need for direct terrain
probing actions.
Keywords:
- Correction
- Source
- Cite
- Save
- Machine Reading By IdeaReader
11
References
0
Citations
NaN
KQI