KHATIB Oussama
<p class="MsoNormal" >KHATIB,Oussama, Robotics, June 15, 2018
This is a noon seminar in Gates 104 in CSD at Stanford.
OK = Oussama Khatib is a prof. in the computer science dept.
well known roboticist.
Romeo and Juliet... his early cooperative bots.
shows video of work in the 80s.
using artificial potential field for avoidance of obstacles.
work in 1981.... connecting planning with trajectory.
guys blowing on ping pong paddle to move the ball to and fro.
we still have fixed trajectories...
ship sunk in 1664.
was discovered in 1993.... near Toulon.
Ocean One discovered it... in waters 100 meters deep.
how to enable human robot collab eg for underwater exploration.
shows several underwater vehicles.
Ocean One... place hands forward... anthropoid robot.
want autonomy for hands (or semi-autonomous)
use force feedback grippers
weighs 250 kg... mission is archeology... drag it to the pool...
want to take it to the Red Sea.
or underwater pipeline inspection and maintenance.
Mining eg.... new frontiers.
say to 3000 meters... in enviromnments where it's hard for humans to work...
he visited a gold mine in Peru (in hard hat)
and to Titicaca Lake... v deep.
Disaster environments.
high altitude stations...
similar to ISRG intuitive surgical.
or to Hansen Medical.
Domestic Robots.
Hal 3 exoskeleton in Japan...
Cyberdyne exoskeleton.
Urban farming...
or confined spaces... hire midgets (small people) to place rivets.
challenge in robots.... get control ME/ algos interface with the hardware guys
b/c sometimes just adding more sensors really helps.
move issues up or down the stack,
(many control problems can be solved just by having more or better sensors.)
Oussama worked with Asimo for about ten years....
but how general is it? how much programming, eg
to twist off the lid of a camping bottle...
but cannot do that in unstructured environments.
inverse kinematics... (but these were open loops)
but must be able to dynamically adjust compliance;
old days: structured enviros...
needed rigid robots
nowadays: want flexible... adaptive;
torque control that is human friendly...
torque control is a basic capability.
force feedback...
fine motion synthesis...
humans better at prediction...
separate posture control and hand control...
screwing a bottle cap...
injured football player....
human kinematics.... shows skeleton passing a football.
(massive interplay among all the muscle groups and bones of the
entire body)
MRI... subject specific with
246 muscle actuators.
shows a golfer driving a ball (skeleton) ... uses entire body.
humans robot competition with swords like Star Wars.
this was a 3 week project!!!
SAI multi-contact simulation.
automate sono for pregnancy fetal sonogram.
contact at many points eg window washer.
must deal with posture, tasks behavior, multi contacts,
internal constraints, local obstacles, balance.
humans do this all simultaneously and dynamically...
we implement this in
constrainted-consistent whole body control algo.
must maintain all constraints...
posture, consistent with joints, torque limits.
constraints map to Jacobian...
slippery surface that robot is standing on with feet separated...
must be inside the friction cone.
shows a simulated robot... rock climbing...
SupraPed.... use hiking poles...
plan must take into account the lower levels...
Stfd Robotic Center.... huge new bldg!!!
Ocean One.
connect Jacobian to the task.
in A to Pat Langley.
shipped to Andre Malraux (sailing vessel)
we tested at one meter...
then went to 15 meters.
from DRASSM by Frederic Osada and Teddy Seguin.
arms were stuck under rocks.
sediment from the bottom...
we got a vase from the wreck in 1664.
shows video of Ocena One....
The Lune Shipwreck...
Caught between the cannons...
rescued by pushups.
has a lot of sensors...
robot can go to 200 meters... redesign for 1000 meters.
this was april 30 2016.... big press event.
Panda robot from Germany...
Eva from Kuka was similar...
<b>Franka</b>.... is a new robotics company...
Panda is the new robot.
great food today at this seminar: sushi... (Oussama's favorite);
my seatmate is Carl Fattal (CEO of start-up)
</p>