InMoov knee actuation

This my latest refined knee model which is running with the service DIY Servo.

I love that service by the way!! Well let say, I just love MyRobotLab!

It's incredible the evolution it has under gone since I started this project.

Before everyone starts to break it after the Manticore, I am asking Santa to implement velocity in that service to see what is smoothness of movement on payload, and if it can still control with enough power.


Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.
juerg's picture

hi Gael, great - fast mover -

hi Gael, great - fast mover - will it be able to stem up the body from e.g. a resting sitting position?

I still have my doubts about a walking InMoov but if we get there I will start to reprint one.

hairygael's picture

Hello Juerg, It's difficult

Hello Juerg,

It's difficult to say if it will rise from sitting to standing without calculations. But the strength of the mecanism is impressive even running at 12V. At 18V I cannot manually stop it. I must be very carefull with the beast, not to lose a finger.

You are correct walking is a dream goal!

One without dreams achieves nothing, so lets keep dreaming!

In the mean time lets try to make it keep its own balance, that will be already great.

GroG's picture

Great to see all the progress

Great to see all the progress ...

Baby steps - perhaps shortly something like this will be in order ... :D

moz4r's picture

Hehe white walkers are

Hehe white walkers are comming !

We can not wait to break everything so much excinting things to do and learn !

About velocity there is not a lot of days until christmass, and there is something blocking :

But maybe soething can be pushed, not perfect angular speed control but this just should avoid inmoov from kicking you.

Bretzel_59's picture


This is awesome Gael !

To estimate torque, Do you know the total gear ratio ?

moz4r's picture

Idea, wainting perfect

Idea, wainting perfect Velocity implementation, something related to velocity can be pushed, it is MAX velocity. Goal is to tweak the PID based on frequencie. On the paper , the robot will didn't take care off the load ( or no load ) and will never go at full speed anymore.

This will consumme power , only if needed.

That is paper, need to test ...

Bretzel_59's picture

This is nearly what i've done

This is nearly what i've done under labview with my brushless servo test.

I can change the output range of the PID, the effect is limiting the actuator speed, but torque too.

I think in the future, you want to do ramping of the setpoint depending on the time.

moz4r's picture

Yeh dude think to ramping of

Yeh dude think to ramping of the setpoint depending on the time ( and the load )  is the goal.

I like what kwatters said few time ago, "here we like to build our own bicycles" or something similar.

juerg's picture

one of the shortcomings of

one of the shortcomings of current servos is it's recognition of being unable to reach a goal e.g. move an arm when it runs into an obstacle and is unable to perform a move in the requested speed/time the move parameters request.

It might even be applied to existing standard servos? Providing the move parameters it should be possible to calculate the total time needed for the move? If this time, or maybe rather  110..120% of this time has elapsed, reset the goal to the current position therefore stopping the robot from trying to reach an impossible location and also avoid burning the servo? An error message might be nice to see that a certain requested position was modified by the time limit evaluated for the move.

moz4r's picture

yeh Juerg you re right ,

yeh Juerg you re right , velocity will be dynamic and if anything prevents the targetPos to be reached at full power, a warning / autodetach / targetpos change ... will be usefull.

off course read potentiometer value off conventional servo can be usefull too in future releases

For now with DIY we can read diyServo.getCurrentPos() -> realtime current angle position

And getCurrentVelocity ( to know how the motor is reaching the pos )

Todo is cook those things

moz4r's picture

Something is cooked and

Something is cooked and pushed and it is not the capon

We need more magic ( + time, discussions, contributions and coffee ) to get it worky perfect, but seem ok with my hardware. The biggest problem is accuracy, I cannot control diy servo degrees to 1 degree. And target is emulated is most cases ( if servo is not moving anymore with few degrees error correction, we consider targetPos is reached )