This afternoon/evening we took Ben out to the small road behind the USPS parking lots and managed to fix the few remaining issues with drivers and we successfully demonstrated trajectory following with Ben, following (steering control only due to the confined space) straight and curved trajectories previously driven by Tully and recorded/played back via Sheep. Attached is the proof - the results were very encouraging the performance in terms of tracking is really quite good considering that there is still a reasonably significant amount of noise in the pose estimate which affects both the recording of the trajectory and the following (we are still occasionally bermuda triangulated to somewhere in manhattan momentarily). What is also impressive is that the controller that we tested is unchanged from the version that gives near equivalent performance in simulation - no gains were modified.
Alex
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment