Article Series on DARwIn Biped: Ball Joints, Dynamic Walking

Dedicated to the new Dynamic Anthropomorphic Robot with Intelligence - Open Platform By Robotis and related projects
2 postsPage 1 of 1
2 postsPage 1 of 1

Article Series on DARwIn Biped: Ball Joints, Dynamic Walking

Post by NullARC » Thu Dec 07, 2006 3:14 am

Post by NullARC
Thu Dec 07, 2006 3:14 am

I came across a very interesting article in the December 2006 issue of SERVO magazine that some of you might be interested in. On page 40 there is the first of what appears to be at least three articles on the DARwIn (Dynamic Anthropomorphic Robot with Intelligence) humanoid robot.

DARwIn is a "humanoid robot capable of walking and performing human-like motions." He was developed at the Robotics & Mechanisms Laboratory (RoMeLa) at Virginia Tech U. He was designed for competition in RoboCup competitions. "...DARwIn will be able to implement human-like dynamic gaits while navigating obstacles and traverse uneven terrain while implementing complex behaviors such as playing soccer."

The first article discusses the implementation, application and benefits of imitating the ball and socket joint. Also very interesting, is the discussion on Dynamic vs. Static Walking. The article talks about how most biped robots (including most of ours :D ) implement 'Static Walking'. This is described as always keeping the robots 'Center of Mass' (COM) "over the area formed by connecting all of the outermost parts points of the robot (usually the foot / feet) in contact with the ground. The robot stands on one foot and moves the other while keeping the (COM) over the grounded foot by constantly adjusting it's posture. When both feet are on the ground, the robot can shift it's COM over the forward foot. The robot repeats this process for each step." This is why our robots look so "mechanical" when they move. Conversely, "Humans walk in a DYNAMIC fashion, a state of constant falling, where our COM is not always over our foot. When we pick up our foot and walk forward, we actually fall forward and catch ourselvs as we step... Dynamic walking is generally faster and much more efficient mode of walking." Not to mention much more 'human like'. Which is what we're all trying to achieve. Well, most of us! :lol:

Those were the most intriguing aspects of the initial article. Other topics coverd are: Using the "CM-2 board, C++, and LabView" software. "Zero point movement." And "Moment of inertia."

Subsequent articles will cover "describing the design and fabrication process for DARwIn's harware, highlight salient (?) features, and briefly discuss some of the research issues related to creating the walking gait and control algorithm for the robot."

I thought some of you might be interested in this series of articles as it discusses the progression of biped robot movement.

Hope you all find this as fassinating as I did!
I came across a very interesting article in the December 2006 issue of SERVO magazine that some of you might be interested in. On page 40 there is the first of what appears to be at least three articles on the DARwIn (Dynamic Anthropomorphic Robot with Intelligence) humanoid robot.

DARwIn is a "humanoid robot capable of walking and performing human-like motions." He was developed at the Robotics & Mechanisms Laboratory (RoMeLa) at Virginia Tech U. He was designed for competition in RoboCup competitions. "...DARwIn will be able to implement human-like dynamic gaits while navigating obstacles and traverse uneven terrain while implementing complex behaviors such as playing soccer."

The first article discusses the implementation, application and benefits of imitating the ball and socket joint. Also very interesting, is the discussion on Dynamic vs. Static Walking. The article talks about how most biped robots (including most of ours :D ) implement 'Static Walking'. This is described as always keeping the robots 'Center of Mass' (COM) "over the area formed by connecting all of the outermost parts points of the robot (usually the foot / feet) in contact with the ground. The robot stands on one foot and moves the other while keeping the (COM) over the grounded foot by constantly adjusting it's posture. When both feet are on the ground, the robot can shift it's COM over the forward foot. The robot repeats this process for each step." This is why our robots look so "mechanical" when they move. Conversely, "Humans walk in a DYNAMIC fashion, a state of constant falling, where our COM is not always over our foot. When we pick up our foot and walk forward, we actually fall forward and catch ourselvs as we step... Dynamic walking is generally faster and much more efficient mode of walking." Not to mention much more 'human like'. Which is what we're all trying to achieve. Well, most of us! :lol:

Those were the most intriguing aspects of the initial article. Other topics coverd are: Using the "CM-2 board, C++, and LabView" software. "Zero point movement." And "Moment of inertia."

Subsequent articles will cover "describing the design and fabrication process for DARwIn's harware, highlight salient (?) features, and briefly discuss some of the research issues related to creating the walking gait and control algorithm for the robot."

I thought some of you might be interested in this series of articles as it discusses the progression of biped robot movement.

Hope you all find this as fassinating as I did!
"She'll make point five past lightspeed. She may not look like much, but she's got it where it counts, kid. I've made a lot of special modifications myself."
NullARC offline
Savvy Roboteer
Savvy Roboteer
User avatar
Posts: 57
Joined: Sun Nov 05, 2006 1:00 am
Location: Akron, Ohio USA

Post by limor » Mon Dec 11, 2006 12:57 am

Post by limor
Mon Dec 11, 2006 12:57 am

Current RC-servo driven robots use position-control servos. These servos were invented in order to keep the RC car front wheels stable and the RC airplane ailerons stable. No one at the RC-servo companies has realized (yet) that in order to achieve motions that look like animals doesn't require a focus on holding torque. Instead, the direction should be to allow the controller to update torque at high speed (in relation to sensory inpu including external torque (current consumption), shaft position, gyro, and any others).

The articulated RC-servo driven robot walk is currently a transition between one position-holding to another.

I think that this year we will see several implementation of torque driven humanoids on this forum. This is because Robotis lends this kind of development environement, Hitec's HMI protocol does aswell (Kondo may surprize us too), OpenServo may allow old KHR humanoids and others to do the same.

Here are a couple of animal-walk preventing-factors with current RC-servo and gait generation programs/controllers paradigm:

1) if an obstacle or uneven terrain is encountered but the robot is transitioning between position A and position B, it does not change its course. (currently only gyros can be plugged into our humanoids to provide this type of low-level real-time corrective position updates. but gyros is only one of several sensor that are needed for the complete job)

2) high-torque position holding is not the way animals move. Animals continously flex and relax their muscles in reaction to the terrain and their own inertia in order to avoid falling, avoid hurting themselves, avoid tearing their tendons etc. and as you mentioned, walking is in essence a perpetual fall. So continous torque control of the servos will allow us to mimic this flexing and relaxing of the servos rather than have the "classic" rigid robotic walk.
Current RC-servo driven robots use position-control servos. These servos were invented in order to keep the RC car front wheels stable and the RC airplane ailerons stable. No one at the RC-servo companies has realized (yet) that in order to achieve motions that look like animals doesn't require a focus on holding torque. Instead, the direction should be to allow the controller to update torque at high speed (in relation to sensory inpu including external torque (current consumption), shaft position, gyro, and any others).

The articulated RC-servo driven robot walk is currently a transition between one position-holding to another.

I think that this year we will see several implementation of torque driven humanoids on this forum. This is because Robotis lends this kind of development environement, Hitec's HMI protocol does aswell (Kondo may surprize us too), OpenServo may allow old KHR humanoids and others to do the same.

Here are a couple of animal-walk preventing-factors with current RC-servo and gait generation programs/controllers paradigm:

1) if an obstacle or uneven terrain is encountered but the robot is transitioning between position A and position B, it does not change its course. (currently only gyros can be plugged into our humanoids to provide this type of low-level real-time corrective position updates. but gyros is only one of several sensor that are needed for the complete job)

2) high-torque position holding is not the way animals move. Animals continously flex and relax their muscles in reaction to the terrain and their own inertia in order to avoid falling, avoid hurting themselves, avoid tearing their tendons etc. and as you mentioned, walking is in essence a perpetual fall. So continous torque control of the servos will allow us to mimic this flexing and relaxing of the servos rather than have the "classic" rigid robotic walk.
limor offline
Savvy Roboteer
Savvy Roboteer
User avatar
Posts: 1845
Joined: Mon Oct 11, 2004 1:00 am
Location: London, UK


2 postsPage 1 of 1
2 postsPage 1 of 1