PDA

View Full Version : Human Emulation



RobotServicesGroup
11-17-2007, 06:32 PM
We have started a new project at RobotServicesGroup.com that we are calling Human Emulation. We want to determine how closely robot motions mimic human motions. We have standardized the testing of robot motions for comparisons against humans, and other robots. In our first test, we examine the forward walking motion of a Robonova-1, without any gyros on a vinyl tile floor. Future tests could include; adding gyros, testing other robots, testing on different floor surfaces, and testing motions other than just walking.

Please check out the results at http://www.robotservicesgroup.com/Emulation1.html

asbrandsson
11-18-2007, 07:26 PM
Hello,

I think that the biggest problem with a robonova-1 on its own is that it is not really that interesting. You need to add Gyros, Accelemeters, double knees, grippers, etc. Then it is something that can be made to jump, climb strairs, climb rope, etc. It can then become an object of scientific study. In my obinion.

Asbrandsson

Droid Works
11-19-2007, 07:47 AM
If you are trying mimic a human the the best bot to use is a KHR-1HV with 2 gyros and 1 accelerometer. Without hip rotation a Robonova is not capable of a human walk. A robonova has to do a pendulum walk due to no hip rotation. But a 1 HV has hip rotation so it can step turn like a human.

RobotServicesGroup
11-19-2007, 07:36 PM
Thanks for the feedback. One of the main points of our testing is to identify how well each robot mimics human motion using only the base model, with no upgrades, gyros or sensors. It wasn't too long ago that cars didn't come with air conditioning or power locks as standard, but if enough people demand these features, like gyros on robots, eventually the manufacturers will do it.

Using the base model can also be helpful to new users trying to decide which robot to buy. If they are just looking for something less expensive to get started with, then perhaps the walking skills of the Robonova will do just fine. If they are willing to spend the extra money for hip servos, then perhaps the KHR-1HV is best.

We want to test each robot, and let the users decide if the price difference is worth it.

asbrandsson
11-23-2007, 08:46 PM
Hello,

I think that the problem is that they kits are supposed to basically be a starting point from which you can develop an understanding of robotics and not really an end product. It is not an out of the box robotic companion.

I think that it is there to get you thinking about the most basic way to build a robot that is an android.

Asbrandsson

Alex
11-26-2007, 02:44 PM
One of the main points of our testing is to identify how well each robot mimics human motion using only the base model, with no upgrades, gyros or sensors. Using the base model can also be helpful to new users trying to decide which robot to buy.
I think this would be very helpful to anyone new to robotics, looking for some info to help determine which robots right out of the box with no mods will have the most realistic walking gait. BTW, the newer RoboPhilo has hip rotation servos and is less than $500. No clue when it'll be out though.


It wasn't too long ago that cars didn't come with air conditioning or power locks as standard, but if enough people demand these features, like gyros on robots, eventually the manufacturers will do it.You bring up a really good point. Once the technology is widespread enough, and the general public understands what a Gyro is/does (i.e. AC will cool you off), the demand for a robot to come pre-equipped with them will be much higher. With the small customer base that the humanoid robotics market has, the chances that the manufacturers receive as many demands as it takes to make decisions like this is pretty low. Cars have been around since the early 1900's, AC didn't come out for cars till mid 1900's (guessing here, but you get the point), but it wasn't until the late 80's, early 90's that most all cars came pre-equipped with AC.

RobotServicesGroup
11-26-2007, 08:49 PM
Robots-Dreams.com has posted a video showing a foot race between some I-sobots. http://www.robots-dreams.com/2007/11/i-sobot-worlds.html It appears that they covered approximately 3 body lengths in about 30 seconds using about 37 steps. These are just rough estimates, but it would appear to be better than the results we got on the Robonova. Of course, the I-Sobots are using a continuous walking cycle, and they come with a gyro while the Robonova has neither of these.

We can't wait to test other robots and compare them against our initial Robonova test. Right now the user voting for our next test has "test a different robot" in the lead with more than 50% of the vote. "Add a gyro to the Robonova" is in second place with over 25% of the vote.

Be sure to visit http://www.robotservicesgroup.com/Emulation3.html (http://www.robotservicesgroup.com/Emulation3.html)and cast your vote (near the bottom of the page) to see what we will test next!

Alex
11-27-2007, 11:56 AM
Cool, I just voted. I love the great work you guys are doing!

Hilarious video. I wish I understood what they were saying though, especially when the one I-Sobot veered off the path, haha!

From the video, it looks like you're pretty much right on with your estimate. Does the I-Sobot come with a Gyro though? I didn't think it did. I thought the nice walk was caused by well written routines.

RobotServicesGroup
12-07-2007, 09:00 PM
To continue with our Human Emulation project, we have performed a new test. This time we focused on turning 180 degrees to the right or left. Our test subject is the same Robonova-1 we used for our forward walking test.

How long do you think it takes for a Robonova-1 to turn around?

To find out, please check the results at www.RobotServicesGroup.com/Emulation4.html (http://www.RobotServicesGroup.com/Emulation4.html).

electrocute
12-11-2007, 03:12 PM
Double Knee and gyros are a must to be added in there so as to smoothen out the movements of the bot.

Alex
12-12-2007, 11:46 AM
The stock Robonova-1 just isn't really all that great at turning due to all of the reasons that you mentioned (no gyros, no hip/waist servos), but I know that you're intent is to test out unmodded RN's. To perform more human-like, Robonova's will need modification, there's just no way around it. Still though, excellent work on collecting all of the data!

RobotServicesGroup
02-18-2008, 10:06 PM
We have continued the work on our Human Emulation project by testing an I-Sobot.
We subjected the I-Sobot to all of the same tests we did on the Robonova-1 to see which robot does a better job of mimicing Human Motions.

Which one do you think will do better?
The $300 I-Sobot or the $1,000 Robonova-1?
The results may surprise you.

You can find all the details at http://www.robotservicesgroup.com/Emulation1.html (http://www.robotservicesgroup.com/Emulation1.html)

Don't forget, you can take either one of these robots for a test drive and check out the walking and turning motions for yourself.
Sign up today for your FREE Robot Test Drive at http://www.robotservicesgroup.com/TestDrive.html

tom_chang79
02-19-2008, 02:39 AM
I enjoyed your article... I have the i-sobot myself and it's so amazing. This little guy is so small but yet so sophisticated. I can't wait for someone to hack into this little thing to be able to control it through programming. One of the first things I want to do is change the stride on this guy. The default walking gait uses very minimal stride, probably because it uses the least amount of power, but I want to change the stride to have a higher clearance so it can possibly walk on carpet.

The article goes to great lengths on relative comparison and it's very surprising to see the little guy beat out the bigger opponent proportionally. Sure, the little i-sobot will probably get his little butt kicked in a robo fight against a Robonova, but the sheer potential in the mechanical component makes this guy very mod-worthy...

Please keep up the great articles/reviews/comparisons on your site, I am a big fan of it...

:)

tom_chang79
02-19-2008, 09:41 AM
Well, I submitted a vote at the end of the article, and I voted to test other robots. How about a KHR-1HV versus a Bioloid Humanoid? I think that would be a good 1 to 1 comparison since they have nearly an equal articulation.

Robin Hewitt
03-06-2008, 01:55 PM
One curious human movement I discovered in my quest for life simulation is that hands move in straight lines unless there is an obstacle in the way. The path of the elbow is complicated but almost irrelevant, becoming curious only if the hand can move directly but it can't.

I'm still bogged down on the feet, I can ramble on about feet all day, it's such a curious device and has to be right if you want any hope of a credible gait. The achilles tendon pulls on the ball of the heel which connects via a sheet of tendons to the balls of the toes. Just in front of the heel the tendons are supported by a ligament looping around them from above. This means that the front of the foot comes down at nearly twice the angular rate that the heel moves at making the heel pronounced, surprised your shoes don't drop off. When the weight rolls on to the foot it is quite easy to sprain the ankle, I don't think it resists twisting until after it starts to flex... etc etc drones on for hours

RobotServicesGroup
03-13-2008, 09:45 PM
Robots-Dreams.com has a video of a Manoi AT01 which moves at blazing speeds. The robot is able to cover a distance equal to 3 body lengths in roughly 5 seconds, taking about 20 steps, with very little variance from side-to-side. This blows away the performance we saw from the Robonova-1 and the I-Sobot we tested. In fact, the robot is so fast, it is almost equivalent to Human speed! Check out this video:

http://www.robots-dreams.com/2008/03/dr-giy-out-to-s.html

Unfortunately, it took the expertise of Dr. GIY and some upgraded servos to get this performance. The bright side is that near Human performance is possible! We only hope that this kind of fluid movement will come as standard equipment on robots of the future so that those of us who are not as talented as Dr. GIY can enjoy similar results.

Alex
03-14-2008, 10:43 AM
That is almost hard to believe it's so freakin fast!! Thanks for givin us the heads up:)


http://www.youtube.com/watch?v=pYDHrigqReM

RobotServicesGroup
10-26-2008, 01:32 PM
We continue our Human Emulation testing by subjecting our KHR-1 to the same forward walking and 180 degree turn tests that we have already performed on the Robonova and I-Sobot. Which robot do you think will be the winner? The cheap robot with continuous motion? The more expensive robot? Since the KHR-1 is the first robot ever produced, more than 4 years old, how can it possibly compete with the newer robots?

Find out all the details on our summary page here. (http://www.robotservicesgroup.com/Emulation1.html)

The individual results for the KHR-1, including videos of each walking test, are available here. (http://www.robotservicesgroup.com/Emulation5.html)

And analysis of the data for all three robots is here. (http://www.robotservicesgroup.com/Emulation3.html)

http://robosavvy.com/Builders/RobotServicesGroup/heightthreebots_402trim240.jpg

RobotServicesGroup
11-22-2008, 12:01 PM
There is a new robot in Japan called Melissa (see this thread (http://robosavvy.com/forum/viewtopic.php?p=18365&sid=a2d39c7a29e1cfe29d194adb91e3c799)). When we watched the videos it looks like Melissa has some great human emulation skills.

The forward walking motion appears to take 11 steps and 4 seconds to move 1 body length. Our standard test is 3 body lengths, so it is possible that Melissa could take 33 steps and 12 seconds under our test. While the number of steps is just average, 12 seconds would be the fastest time of any robot we have tested.

Surprisingly, Melissa was able to turn 360 degrees using 10 steps in 5 seconds. Our Human Emulation test for turning is only 180 degrees, so Melissa should be able to turn with 5 steps in 2.5 seconds, which is nearly identical to our Human score! The truly amazing thing is that Melissa performs this turn with only a single servo in the torso for rotation, rather then two servos at the waist.

This robot has great potential. Check out the forward walking video here (http://www.rakuten.ne.jp/gold/grass-road/melissa/movie/melissa-S.F.wmv)and the turning video here (http://www.rakuten.ne.jp/gold/grass-road/melissa/movie/melissa-S.STU.wmv). You can also check all of our Human Emulation tests here. (http://www.robotservicesgroup.com/Emulation1.html)

LinuxGuy
11-24-2008, 12:56 AM
Double Knee and gyros are a must to be added in there so as to smoothen out the movements of the bot.
I've even designed, and assembled the brackets for, a 6DOF SES biped leg. I am not sure I included hip rotation, but will have to check that. I'd also want waist rotation for an even more Human like walk. The leg I designed has three double joints. :)

8-Dale