RoboSavvy Designing and Fabricating a 1m tall Humanoid

RoboSavvy distributes and manufactures robots. This forum is dedicated to robots and other bits designed or manufactured by RoboSavvy and robot software developed by RoboSavvy.
52 postsPage 1 of 41, 2, 3, 4
52 postsPage 1 of 41, 2, 3, 4

RoboSavvy Designing and Fabricating a 1m tall Humanoid

Post by PedroR » Thu Oct 25, 2012 11:14 am

Post by PedroR
Thu Oct 25, 2012 11:14 am

Hi all

RoboSavvy is currently developing a 1m tall Humanoid for a UK University as part of a consulting/research project.

The Humanoid will be based on Robotis servos (MX series) and will include advanced features such as onboard i7 PC, 24+ degrees of freedom, advanced Vision systems and Physics simulator.

The whole project has been divided into several steps, which we'll be blogging about over the coming weeks.

We're currently on Step1/2 which is CAD Design and BOM assessment.

The next steps will Involve the actual Fabrication of the frames (check our progress cutting metal with our in house CNC) and Development of the Robot Software and Physics Simulator.

A Body shell is also planned, incorporating clever, hybrid fall management solutions combining software and hardware features.

Over the coming weeks we will be sharing with the community our progress and our findings and progresses addressing the challenges of the added size, weight, power consumption and fabrication.

Regards
Pedro.
Hi all

RoboSavvy is currently developing a 1m tall Humanoid for a UK University as part of a consulting/research project.

The Humanoid will be based on Robotis servos (MX series) and will include advanced features such as onboard i7 PC, 24+ degrees of freedom, advanced Vision systems and Physics simulator.

The whole project has been divided into several steps, which we'll be blogging about over the coming weeks.

We're currently on Step1/2 which is CAD Design and BOM assessment.

The next steps will Involve the actual Fabrication of the frames (check our progress cutting metal with our in house CNC) and Development of the Robot Software and Physics Simulator.

A Body shell is also planned, incorporating clever, hybrid fall management solutions combining software and hardware features.

Over the coming weeks we will be sharing with the community our progress and our findings and progresses addressing the challenges of the added size, weight, power consumption and fabrication.

Regards
Pedro.
PedroR offline
Savvy Roboteer
Savvy Roboteer
Posts: 1199
Joined: Mon Jun 16, 2008 11:07 pm

Post by TiagoR » Wed Oct 31, 2012 10:28 am

Post by TiagoR
Wed Oct 31, 2012 10:28 am

Hello all

Since the robot we are developing is based on DARwIn-OP the first approach was to keep the design of the structure and increase the size of the brackets, resulting in the following parts.

Bracket from Hip to Knee:
Image
Hip2Knee

Bracket from Knee to Ankle:
Image
Knee2Ankle

Just by looking at the first bracket it is possible to see that with the weight that it will support it will probably bend and compromise the legs.

Because of that we decided to add a reinforcement:
Image
LegReinforcement

The resulting leg looks currently like this:

Image
LeftLeg

Regards
Tiago
Hello all

Since the robot we are developing is based on DARwIn-OP the first approach was to keep the design of the structure and increase the size of the brackets, resulting in the following parts.

Bracket from Hip to Knee:
Image
Hip2Knee

Bracket from Knee to Ankle:
Image
Knee2Ankle

Just by looking at the first bracket it is possible to see that with the weight that it will support it will probably bend and compromise the legs.

Because of that we decided to add a reinforcement:
Image
LegReinforcement

The resulting leg looks currently like this:

Image
LeftLeg

Regards
Tiago
TiagoR offline
Robot Builder
Robot Builder
Posts: 19
Joined: Mon Oct 29, 2012 1:12 pm

Post by Novice » Thu Nov 01, 2012 10:48 am

Post by Novice
Thu Nov 01, 2012 10:48 am

Hi Tiago.
Be carefull,the lenght of the leg is too long.You have to use the high power servos for that, otherwise the ones may be burnt.The Darwin-OP is good disigned and counted robot, I think you could change the basic dimension of one plus or minus 15-20% maximum.If more- you have to refind new datas for everything.
Please note, even though you did the size and the weight of the robot small different agaist of the its firm size, it is the reason why the robot cant do the standard movements from the firm labruary, be ready to design all the movements and write your own new basic labruary.
Best regards
Oleg
Hi Tiago.
Be carefull,the lenght of the leg is too long.You have to use the high power servos for that, otherwise the ones may be burnt.The Darwin-OP is good disigned and counted robot, I think you could change the basic dimension of one plus or minus 15-20% maximum.If more- you have to refind new datas for everything.
Please note, even though you did the size and the weight of the robot small different agaist of the its firm size, it is the reason why the robot cant do the standard movements from the firm labruary, be ready to design all the movements and write your own new basic labruary.
Best regards
Oleg
Novice offline
Savvy Roboteer
Savvy Roboteer
Posts: 26
Joined: Wed Dec 30, 2009 1:21 pm

Post by limor » Thu Nov 01, 2012 11:50 am

Post by limor
Thu Nov 01, 2012 11:50 am

Hi Oleg

Thank you for your suggestions.
Indeed the servos are being replaced in most cases by the more powerful MX-64 and MX-106.
In some cases such as the knees it is obvious that 106 needs to be deployed while in other places such as the 3 axis of the hip, it is not so clear if MX28 or MX64 will be needed.

The CAD model produces all the intertial data of the robot and it will be inserted into a simulator (V-Rep most probably or Ezphysics or Gazebo). In the simulator the plan is to run the Gait generator of Darwin-OP, adjusted for the model.

The movements of the robot determine what power the servos need to have.
1) Darwin-OP Gait generator
2) Getting up sequence from laying on front and back - very tricky
3) manipulation of objects using hands

A simulator will never be able to actually represent reality but if the simulator shows that the range of torques for the given motions is within say 70% of the specs of the servos, then it is likely that the real robot will be able to perform the motion in real life
Hi Oleg

Thank you for your suggestions.
Indeed the servos are being replaced in most cases by the more powerful MX-64 and MX-106.
In some cases such as the knees it is obvious that 106 needs to be deployed while in other places such as the 3 axis of the hip, it is not so clear if MX28 or MX64 will be needed.

The CAD model produces all the intertial data of the robot and it will be inserted into a simulator (V-Rep most probably or Ezphysics or Gazebo). In the simulator the plan is to run the Gait generator of Darwin-OP, adjusted for the model.

The movements of the robot determine what power the servos need to have.
1) Darwin-OP Gait generator
2) Getting up sequence from laying on front and back - very tricky
3) manipulation of objects using hands

A simulator will never be able to actually represent reality but if the simulator shows that the range of torques for the given motions is within say 70% of the specs of the servos, then it is likely that the real robot will be able to perform the motion in real life
limor offline
Savvy Roboteer
Savvy Roboteer
User avatar
Posts: 1827
Joined: Mon Oct 11, 2004 1:00 am
Location: London, UK

Post by TiagoR » Fri Nov 02, 2012 5:46 pm

Post by TiagoR
Fri Nov 02, 2012 5:46 pm

Hello all

Today I bring some details on the torso construction.

First a perspective of the front and back of it.

Front:
Image

Back:
Image

As you can see in the previous images the torso is divided in essentially 5 parts: Hips, batteries, electronics, motherboard and Arms & Neck.

The hips part consist in the remaining degrees of freedom of the hip that are not on the legs picture, and right above them is the batteries room.
Image

Next is an empty room where some of the electronics needed, but not yet built will be placed. On top of that is the room where the Motherboard will fit.
Image

Motherboard:
Image

Finally is the room where the Neck and shoulder servos will fit.
Image

Best Regards
Hello all

Today I bring some details on the torso construction.

First a perspective of the front and back of it.

Front:
Image

Back:
Image

As you can see in the previous images the torso is divided in essentially 5 parts: Hips, batteries, electronics, motherboard and Arms & Neck.

The hips part consist in the remaining degrees of freedom of the hip that are not on the legs picture, and right above them is the batteries room.
Image

Next is an empty room where some of the electronics needed, but not yet built will be placed. On top of that is the room where the Motherboard will fit.
Image

Motherboard:
Image

Finally is the room where the Neck and shoulder servos will fit.
Image

Best Regards
TiagoR offline
Robot Builder
Robot Builder
Posts: 19
Joined: Mon Oct 29, 2012 1:12 pm

Post by SK » Sat Nov 03, 2012 1:46 pm

Post by SK
Sat Nov 03, 2012 1:46 pm

This robot seems to be very similar to the 95cm Nimbro-OP (http://www.nimbro.net/OP/). For that one, both hardware and software sources are available and it is made by Sven Behnke's group who has 5+ years of experience with humanoids of that size. I know that the Nimbro-OP doesn't feature a i7 CPU and more DOF by default, but wouldn't it make sense to use it as a starting point?
This robot seems to be very similar to the 95cm Nimbro-OP (http://www.nimbro.net/OP/). For that one, both hardware and software sources are available and it is made by Sven Behnke's group who has 5+ years of experience with humanoids of that size. I know that the Nimbro-OP doesn't feature a i7 CPU and more DOF by default, but wouldn't it make sense to use it as a starting point?
SK offline
Savvy Roboteer
Savvy Roboteer
User avatar
Posts: 105
Joined: Mon Jun 26, 2006 1:00 am
Location: near Frankfurt/Main, Germany

Post by TiagoR » Tue Nov 06, 2012 11:19 pm

Post by TiagoR
Tue Nov 06, 2012 11:19 pm

Hello SK

First of all I am sorry for the late reply.

We are using Nimbro-OP not exactly as a starting point but mostly as a reference.

There are some reasons for that, such as we want a slightlly taller robot, and we still don't know if we will be able to build the carbon fiber parts or not, among other reasons.

Best Regards
TiagoR
Hello SK

First of all I am sorry for the late reply.

We are using Nimbro-OP not exactly as a starting point but mostly as a reference.

There are some reasons for that, such as we want a slightlly taller robot, and we still don't know if we will be able to build the carbon fiber parts or not, among other reasons.

Best Regards
TiagoR
TiagoR offline
Robot Builder
Robot Builder
Posts: 19
Joined: Mon Oct 29, 2012 1:12 pm

Post by TiagoR » Tue Nov 06, 2012 11:21 pm

Post by TiagoR
Tue Nov 06, 2012 11:21 pm

Hi all

Today I bring some updates on the arms design.

Image

Image

Image

The mechanisms of the hand are still in study so this is not the final version.

Best regards
TiagoR
Hi all

Today I bring some updates on the arms design.

Image

Image

Image

The mechanisms of the hand are still in study so this is not the final version.

Best regards
TiagoR
TiagoR offline
Robot Builder
Robot Builder
Posts: 19
Joined: Mon Oct 29, 2012 1:12 pm

Post by xevel » Wed Nov 07, 2012 7:40 am

Post by xevel
Wed Nov 07, 2012 7:40 am

All of this looks really interesting, thanks for sharing!

I see your arm and hand seems to have a lot of DOFs! Are you planning to have them under-actuated?
All of this looks really interesting, thanks for sharing!

I see your arm and hand seems to have a lot of DOFs! Are you planning to have them under-actuated?
xevel offline
Savvy Roboteer
Savvy Roboteer
Posts: 73
Joined: Sun Mar 27, 2011 6:37 pm

Post by TiagoR » Wed Nov 07, 2012 10:34 am

Post by TiagoR
Wed Nov 07, 2012 10:34 am

Hello xevel

Thank you for your interest.

We are planning on having 5 degrees of freedom in the hands:
-> Wrist
-> Thumb
-> Index finger
-> Middle finger
-> Ring finger and Pinky (actuated by the same servo)

So the fingers are under-actuated in the sense that either you move the finger or you don't, you can't close the finger arround itself like humans do for example, you can only close it fist-like.

The rest of the arm has for now 3 degrees of freedom (2 on shoulder, 1 on elbow), comparing to a real human it is lacking and it is restricting when you are working on a plane, for example on top of a table.

We will try to explore the possibility of adding a servo before the elbow which would increase significantly the working area on a plane.

I hope this answers your question.

Best regards
TiagoR
Hello xevel

Thank you for your interest.

We are planning on having 5 degrees of freedom in the hands:
-> Wrist
-> Thumb
-> Index finger
-> Middle finger
-> Ring finger and Pinky (actuated by the same servo)

So the fingers are under-actuated in the sense that either you move the finger or you don't, you can't close the finger arround itself like humans do for example, you can only close it fist-like.

The rest of the arm has for now 3 degrees of freedom (2 on shoulder, 1 on elbow), comparing to a real human it is lacking and it is restricting when you are working on a plane, for example on top of a table.

We will try to explore the possibility of adding a servo before the elbow which would increase significantly the working area on a plane.

I hope this answers your question.

Best regards
TiagoR
TiagoR offline
Robot Builder
Robot Builder
Posts: 19
Joined: Mon Oct 29, 2012 1:12 pm

Post by TiagoR » Thu Nov 08, 2012 7:33 pm

Post by TiagoR
Thu Nov 08, 2012 7:33 pm

Hi all

Today I am showing the head of the robot.

Image

The head is equiped with a stereo camera and an Asus Xtion as it's "eyes".
There will be LEDs to show emotions and it will have 2 degrees of freedom on the neck.

Best regards
Hi all

Today I am showing the head of the robot.

Image

The head is equiped with a stereo camera and an Asus Xtion as it's "eyes".
There will be LEDs to show emotions and it will have 2 degrees of freedom on the neck.

Best regards
TiagoR offline
Robot Builder
Robot Builder
Posts: 19
Joined: Mon Oct 29, 2012 1:12 pm

Post by mantrid » Fri Nov 09, 2012 2:06 pm

Post by mantrid
Fri Nov 09, 2012 2:06 pm

Can I ask what may seem like a disconnected question for my own interest sake?

How complex would the ankle joint and foot be for ambulatory motion in terms of software (processing power) and control and feedback signals?
Can I ask what may seem like a disconnected question for my own interest sake?

How complex would the ankle joint and foot be for ambulatory motion in terms of software (processing power) and control and feedback signals?
mantrid offline
Savvy Roboteer
Savvy Roboteer
Posts: 53
Joined: Mon Oct 22, 2012 2:56 pm

1m Humanoid - three questions

Post by firemonkey » Mon Nov 12, 2012 8:31 am

Post by firemonkey
Mon Nov 12, 2012 8:31 am

Excellent project! Looking forward to seeing this progress.

3 questions immediately occur to me, the first two are related to "would this be capable or safe of tactile human interaction":

- are you planning any sensors around the body to detect "foreign body" proximity?

- are these actuators, or some control algorithm, you are planning capable of detecting "externally applied torque"? i.e. capable of detecting object collision, and thus capable of varying their "motion prerogative" (in a control loop) when a human is holding and moving them?

- what's the estimated cost?
Excellent project! Looking forward to seeing this progress.

3 questions immediately occur to me, the first two are related to "would this be capable or safe of tactile human interaction":

- are you planning any sensors around the body to detect "foreign body" proximity?

- are these actuators, or some control algorithm, you are planning capable of detecting "externally applied torque"? i.e. capable of detecting object collision, and thus capable of varying their "motion prerogative" (in a control loop) when a human is holding and moving them?

- what's the estimated cost?
firemonkey offline
Newbie
Newbie
Posts: 2
Joined: Mon Nov 12, 2012 8:26 am

Post by TiagoR » Mon Nov 12, 2012 7:03 pm

Post by TiagoR
Mon Nov 12, 2012 7:03 pm

Hello all

First answering to mantrid, it all depends on how you are building the ankle: how many degrees of freedom , type and dynamic of the actuators, and sample time.

The more actuators you have the more feedback signals you'll need, therefore bigger processing power.

Same with sample time, smaller sample time, closer to reality, bigger need of processing power.

Unfortunatly i can't give you a much better answer than this.
Hello all

First answering to mantrid, it all depends on how you are building the ankle: how many degrees of freedom , type and dynamic of the actuators, and sample time.

The more actuators you have the more feedback signals you'll need, therefore bigger processing power.

Same with sample time, smaller sample time, closer to reality, bigger need of processing power.

Unfortunatly i can't give you a much better answer than this.
Last edited by TiagoR on Tue Nov 13, 2012 12:19 am, edited 1 time in total.
TiagoR offline
Robot Builder
Robot Builder
Posts: 19
Joined: Mon Oct 29, 2012 1:12 pm

Post by TiagoR » Mon Nov 12, 2012 7:07 pm

Post by TiagoR
Mon Nov 12, 2012 7:07 pm

Today i bring you an image of how the robot currently looks like.

Image

It is still missing the outer shell, but the structure itself probably won't change much from what exists now.

Best regards.
Today i bring you an image of how the robot currently looks like.

Image

It is still missing the outer shell, but the structure itself probably won't change much from what exists now.

Best regards.
TiagoR offline
Robot Builder
Robot Builder
Posts: 19
Joined: Mon Oct 29, 2012 1:12 pm

Next
Next
52 postsPage 1 of 41, 2, 3, 4
52 postsPage 1 of 41, 2, 3, 4