safety_chicken.avi
- shows multiple navigating agents using Dynamics Safety Search in
conjunction with ERRT to navigate while avoiding one another.
Only a single collision between a pair of robots occurs during the
test, despite aggressive speed settings and highly conflicting
goals (the robots repeatedly swap target positions).
safety_teleop.avi
- shows a single agent being teleperated with a joystick
controlling the target velocity. At first, DSS is used as a
post-process to the calculated target acceleration, making it
difficult to hit obstacles even when the operator deliberatly
tries to hit an obstacle. In the second part of the video, DSS is
switched off, and the operator attempts to perform the same
manuevers while manually avoiding obstacles.
ERRT Navigation Videos
The following short videos show the earliest version of the ERRT
motion planner running on real robots in 2001. The robot is a two
wheel differential drive soccer robot, and reaches a maximum speed of
1.7m/s
The following short videos show ERRT and PRM variants choosing actions
for a QRIO humanoid robot as it builds and updates an environment
model determined entirely from stereo data.
ERRT Gap
- shows a QRIO navigating between two obstacles using ERRT.
ERRT Slalom
- shows a QRIO navigating between several obstacles in an offset
pattern using ERRT.
DPRM Uneven Gap
- shows a QRIO navigating a gap with uneven obstacle heights.
Note that it uses the heightmap to avoid the low obstacle less
than the taller one.
DPRM Stair (1st run)
- shows navigation and ascention of a step. The mode transition
to the climbing module can be seen when QRIO stops looks straight
down to position itself with respect to the stair more accurately.
DPRM Stair (2nd run)
- a second run of the step navigation task, this time with an
additional obstacle. QRIO has some difficulty after positioning
itself near the side of the step, but eventually corrects its
position and succeeds.