Slippage of the real-robot during DSP

[Saeed-Mansouri]

Hi Dr. Caron,

We are using $ rosrun mc_rtc_ticker mc_rtc_ticker to connect to the robot. Assuming the condition that we don't use sensor data (joint-encoder, FT sensor, IMU sensor) as the feedback signal. As shown in the attached figure, in the RViZ environment we can see that:

1. joint angles of the real-robot are always in the home configuration
2. control-robot tracks the desired walking trajectory
3. real-robot doesn't move during SSP
4. real-robot slips on the ground, as soon as DSP starts
5. during SSP, one of the legs of the real-robot is coinciding with the stance foot of the control-robot

Would you please explain why we should expect that the real-robot slips on the ground during DSP?

[Saeed-Mansouri]

It has been solved! :)

[stephane-caron]

OK great. Can you share the cause of this behavior and how you solved it? It may help other developers who encounter the same problem in the future (avoiding the DenverCoder9 effect).

[Saeed-Mansouri]

I have found that in void FloatingBaseObserver::estimatePosition(const mc_rbdyn::Robot & realRobot), the anchor frame of the control-robot is used to estimate the floating-base position. Based on this method, if the origin of the real anchor frame doesn't coincide with the control anchor frame, the real-robot slips on the ground in the RViZ environment.