SCONE: Gait stability analysis

[suparna dey]

Hello everyone,

I would like to do a forward dynamic simulation and finally get the torso angle and angular velocities to construct a phase portrait for assessing gait stability.
However, I would like to do it given some pre-computed ankle-torque. Basically, my aim is to asses how the stability of gait is affected with the ankle-torque generated by an external predictor. Therefore, my question is, is there a way to achieve this using SCONE? For e.g., could I input the predicted ankle-torque and some other variables, e.g., the hip flexion angle as some constraint while performing the optimization in SCONE?

Thank you very much in advance.

Thanks and Regards

[Thomas Geijtenbeek]

Adding a fixed ankle-torque pattern is definitely possible. The easiest way is to add a FeedForwardController to the GaitStateController that is used in gait tutorial. If you set the control points of the FeedForwardController to fixed values, then the they won't be part of the optimization. The torque pattern will automatically be in phase with the gait cycle.

Setting a constraint to the hip flexion angles directly is not possible, because the joint kinematics are produced by the simulation and cannot be controlled directly. However, you can add an optimization objective that penalizes the hip flexion when it is outside a specified range or trajectory.

[suparna dey]

Dear Thomas,

Thank you very much for the answers. I have three follow-up questions.

1. Just to confirm, to ensure, that the provided torque pattern be automatically in phase with the gait cycle, I need to set the initial states in sync with the start of the gait cycle corresponding to my torque pattern, am I correct (my gait cycle is defined between two consecutive heel contacts of the same leg)?

2. Excuse me if this question sounds lame. To add an optimization objective that penalizes the hip flexion when it is outside a specified range or trajectory, I would like to ensure that this range varies based on the states of the gait cycle and not a constant throughout the cycle. Therefore, can I provide this as a list or array structure?

3. It would be very helpful if you could suggest any other ways using SCONE to assess gait stability, especially fitting to my use case, where I have an ankle-torque predictor that predicts the ankle torque. I would like to validate whether the predictions from this external predictor results in a stable gait. For e.g., COP inside the base of support, or any other methods?

Thank you very much for your time and kind help.

Thanks and Regards

[Thomas Geijtenbeek]

1. Best would be to split up your torque pattern into separate patterns for stance and swing, so that they remain better synchronized with the gait cycle. They get activated automatically at the start of any phase.

2. You probably need to script the measure yourself in this case, using a ScriptMeasure.

3. Just so I understand correctly: is your ankle torque predictor intended for applying actual torques via a motorized device like an exoskeleton? If not, than you probably wouldn't want to add the ankle torques at all. You might instead try to optimize the gait controller so that it produces torque patterns that match your predictor, but this really depends on what you are trying to accomplish.

[suparna dey]

Dear Thomas,

Thank you for the answers. For the answers to questions 1 and 2, I will follow as you suggested. Just one follow-up with adding the feed-forward controller: how do I specify that the control points correspond to the values of the ankle -joint torque pattern in the sagittal plane?

For question 3, yes, exactly, the ankle torque predictor is intended for applying actual torques via a motorized device e.g., an exoskeleton. Therefore, I would like to validate the stability of gait w.r.t the predicted ankle-torques.

Thanks and Regards

[Thomas Geijtenbeek]

If your ankle torque is meant to represent an external device, then the best approach would be to implement the control signal in a separate ScriptController. This controller then runs separate from the GaitStateController, which you can regard as the 'human' control loop.

It is a little bit more work, since you have to do the step detection yourself, but it has the advantage that it then is a direct representation of the controller that will end up in your device, without any dependencies. You can check out the Script Examples to get started. For heel-strike detection, you can use the contact_force() function of the heel body.

[suparna dey]

Dear Thomas,

Thank you very much for the suggestions. I will try them.

Thanks and Regards
Sharmita Dey