Modeling, Evaluation, and Control Optimization of Exosuit with OpenSim

Team Member

Jaehyun Bae

Overview

Motivations

DARPA Warrior Web Program

This project is motivated by DARPA Warrior Web Program.

There has been a lot of research into exoskeletons over the years to alleviate heavy loads that soldiers should burden, but strapping a person into a robotic outfit just isn't practical in a combat zone yet.
DARPA's Warrior Web program aims to build a lightweight suit that improves a soldier's endurance and overall effectiveness, while preventing injuries.
The main goals by developing the warrior web are

1. To prevent and reduce musculoskeletal injuries.
2. To augment positive work done by the muscles and reduce the physical burden

You can see the details on http://www.darpa.mil/Our_Work/DSO/Programs/Warrior_Web.aspx

Harvard Exosuit

In order to develop an under-suit that doesn’t interrupt wearer’s free movement, researchers are trying to make it soft and deformable, but still capable of applying force to body joints. Harvard exosuit is the example of new approach to create under-suit in a soft and deformable manner.

Experimental data proved that it can help loaded walking by reducing metabolic costs.
This suit applies force to lower limb joints by a cable driven by the actuators on the backpack.

Challenge

As it is a new approach to assist human gait with deformable structure, there are many challenges in developing the exosuit. The challenges are

It is difficult to analyze the effectivness of the suit
It is difficult to find the optimal input force for actuators to reduce the metabolic cost
It is difficult to identify the effect of change of design parameters

The reasons for the challenges are

The Exosuit is deformable
We can not predicet how external actuation assists muscles during loaded walking
Experimental data is inconsistent

Goals

Through this project, I tried to resolve the challenges in developing exosuit with OpenSim. Simulation can help developing soft wearable exosuit as it can give an intuition on how exosuit help muscles, and what are the key features that one should take care of when developing the suit.

I hope this project will construct a systematic way of analyzing and designing soft wearable device. The initial goals of this project are

Evaluate the effectiveness of wearing active actuator on metabolic cost reduction during loaded walking.
Explain how Exosuit can help loaded gait
Verify the impact of changes of design parameters.
Find optimal control inputs for exosuit actuators.
Evaluate the effectiveness of Biarticular actuator

Strategy

Experimental data

Two different data were collected from a same subject.

1. One gait cycle of loaded walking (From left toe off to next left toe off)
2. One gait cycle of unloaded walking (From left toe off to next left toe off)

The subject didn't wear a suit and walked freely.
walking speeds are identical in both cases
Mass of the load for loaded walking was 38kg.

Modeling

To simulate the movement of exosuit wearer, the first thing to do is to create a model which can replicate a real subject as possible as we can. Before I created our model with active actuator on it, I had the generic gait model in opensim to go through the basic steps of modeling procedure in OpenSim. By doing so, I could make my model dynamically consistent to the experimental data. And then, I added actuators and metabolic cost probes to the model.

How to model a subject wearing active actuator

The diagram describes what procedure the generic gait model had gone through.

The first three steps are basic modeling procedure in Opensim to make a model dynamically and kinematically consistent to the experimental data. For more information, read

After step 3 is complete, probes for calculating metabolic cost were added to the model. . For more information about how to add the probes, read

Simulation-Based Design to Reduce Metabolic Cost

Finally, I added active actuators to the model. Here, I used PathActuator class to simulate active actuators of exosuit, as they are cable driven actuator. If you are interested in how PathActuator works, read

OpenSIm::PathActuator Class

Sample models

Here are the figures of sample simulation models. I created several different types of models by modifying RRA-adjusted model for both loaded gait and unloaded gait.

Loaded gait model

Three different types of models were created for loaded gait simulation

A model without actuator
A model with path actuators supporting plantar flexion
A model with path actuators supporting hip extension

The path actuator supporting plantar flexion is attached to heel and tibia, and the path actuator supporting hip extension is attached to backpack and femur. Loaded mass was added to torso for simplicity.

Unloaded gait model

Same types of models were created for unloaded gait simulation. The main difference between loaded gait model and unloaded gait model is the mass of torso, and transparancy of backpack.

Optimization methodology

The idea to optimize the control input force for the actuators is to take advantage of CMC tool. The main reason we use CMC in OpenSim is to find a most suitable excitations for muscles to create body movement while minimizing activation. To see how it works, read

How CMC Works

In this project, I make different use of the optimization process in CMC in order to optimize the control input force for active actuators.

CMC procedure is static optimization process, and it minimizes the cost function J which can be represented as

When we add active actuators on OpenSim Model, the activation term in cost function becomes

Where X_muscle is muscle control and X_actuator is actuator control. As X_actuator is part of activation states, it is also adjusted after the optimization process.

Now, if we diminish the influnece of X_actuator on J, and run CMC, the optimizer tries to find X_actuator in a manner of minimizing muscle activation.
We know that minimizing muscle activation correponds to minimizing metabolic cost, so we can come up with the conclusion that the actuator input force resulted from CMC after diminishing the influence of X_actuator is the optimal actuator input for most efficient metabolic reduction.
Muscle force is constructed from the equation F_actuator = F_actuator^max * X_actuator. if we assign large value of maximum force to each actuator, actuator control X_actuator decreases, so that the influence of actuator to J is decreases.

Using this methodology, I could find an optimal input for each actuator, and also found the metabolic cost reduction after running CMC with a model where active actuators are added.

Result & Discussion

Metabolic cost change when active actuators are added to model

I investigated how much metabolic cost is reduced when optimal input force is applied to a model by actuators. I did simulation for both loaded and unloaded walking cases, and I compared the influence of hip actuator and ankle actuator to metabolic cost reduction. I assigned 10,000N to maximum active actuator force( F_actuator^max) for this simulation.

Loaded walking	Unloaded walking

Metabolic cost reduction when active actuators are added to loaded gait model

1. Ankle actuator: 10.35%
2. Hip actuator: 6.62%

Metabolic cost reduction when active actuators are added to unloaded gait model

1. Ankle actuator: 10.62%
2. Hip actuator: 1.04%

Things to notice

1. The metabolic cost is much lower during unloaded walking than loaded walking. Loaded walking costs only 75% metabolic energy compared to loaded walking.
2. Ankle actuator works better to reduce metabolic cost than hip actuator when we can apply optimal input force.
3. Hip actuator is not assistive to unloaded gait.

Therefore, we can say that ankle actuator helps metabolic cost reduction better than hip actuator if we have an optimal actuator which has no maximum force limitation.

Optimal actuator input force

	Loaded walking	Unloaded walking
Ankle actuator
Hip actuator

Optimal input force for Ankle actuator

Actuation is started right after the toe-off of a foot on the opposite side, and the peak force occurs 7.12% of gait cycle before toe off, and ends at the toe-off of a foot on the same side
This actuation scheme is valid for both loaded gait and unloaded gait
The force signal is clear and easy to implement in real world
However, the maximum actuation force is about 2500 N, which is too high to achieve in reality

Optimal input force for Hip actuator

Hip actuator can reduce the metabolic cost with lower maximum force than ankle actuator
However, it is hard to identify how the actuator assists walking
Also, it is difficult to implement the optimal control input for hip actuator in real world

How ankle actuator assists loaded gait

I could explain how the optimal actuation input for ankle actuator could help loaded gait by investigating the change of plantar flexor muscle forces

The gastrocnemius muscle forces are barely changed.
Other plantarflexor muscle forces, including Soleus muscle forces, are significantly decreased.
If we draw the sum of baseline uniarticular forces and active actuator input force together, we can see that the active actuator force follows baseline uniarticular muscle forces.

To sum up, we can say that ankle actuator assists uniarticular muscles during loaded walking.

Best realistic actuation input force for ankle actuator

Optimal input force when actuation force is limited to 400N

From the previous results, we found that the optimal force for hip actuator is small enough to be achieved by real actuator, while the optimal input force for ankle actuator is not realistic. (>2000N). Therefore, I tried different types of input forces for ankle actuators up to 400N, and compare the results to optimal control input case.

1. My initial guess was to saturate the optimal input force that I found earlier at 400N. I generated an new input force which is identical to optimal input force up to 400N, and saturated once optimal input force exceeds 400N. And then, I run CMC after setting the actuator input forces as a control constraints for each case.
2. The second input force I tried is a new result from different CMC procedure. The new CMC result was acquired after assigning 4000N to maximum actuation force and bounding control input between 0 and 0.1.

In other word,

According to the formula F_actuator = F_actuator^max * X_actuator, the new CMC results also has maximum force of 400N. As X_actuatoris bounded between 0 and 0.1 and X_musclehas a range of 0 and 1, the influence of X_actuatorto objective function of CMC procedure is relatively lower than that of X_muscle, so we can use this idea to create optimal input for active actuator when the maximum actuation force is limited.

When we compare the saturated optimal input and the result from new type of CMC procedure, we can find a similarity between the saturated optimal input and a results of new CMC procedure. Now, let's compare the metabolic cost reduction when each control input is applied to ankle actuators.

Metabolic cost reduction

Loaded walking	Unloaded walking

Metabolic cost reduction when active actuators are added to loaded gait model

1. optimal: 10.35% reduction
2. Saturated: 1.84% reduction
3. New CMC: 2.68% reduction

Metabolic cost reduction when active actuators are added to unloaded gait mode

1. optimal: 10.62% reduction
2. Saturated: 3.46% reduction
3. New CMC: 3.82%

The result from new CMC procedure reduces metabolic cost more efficiently.
However, the reduction is not significant, and it is much lower than the optimal case.
The interesting thing is that the realistic actuation input force works better in unloaded walking case than loaded walking case. It makes sense because we requires lower force to assist unloaded walking than to assist loaded walking.

Biarticular actuator

Now that we know both ankle actuator hip actuator can reduce metabolic cost during loaded walking, the natural procedure is to test the actuators which can affect both ankle plantar flexor and hip extensor. In order to reduce the number of actuator, I added biarticular actuator of 1 DOF affecting ankle plantar flexion and hip extension to each leg, and see how much it reduces metabolic cost.

Modeling

The main idea to create biarticular actuator is to let the path actuator go through the axis of ankle joint rotation. I chose the attachment points of ankle actuator and hip actuator as the via points and end points of biarticular actuator line, and also set the origin of ankle joint rotation as one of the via points. By doing so, I create a biarticular actuator as a combination of ankle actuator and hip actuator synchronized in series.

Simulation result

I run CMC on loaded gait model with biarticular actuator. I set the maximum force of biarticular actuator to be 10,000N, in order to see the optimal input force and best metabolic cost reduction.

Optimal input	Metabolic cost reduction

Metabolic cost reduction when biarticular actuators are added to loaded gait is 3.12% from baseline. It is much lower than reduction rate of ankle actuator or hip actuator.
Control input is complex, which makes it hard to realize
Biarticular actuator is not as effective as uni-articular actuators in terms of metabolic cost reduction.

Conclusion

Three types of active actuators are evaluated in terms of metabolic cost reduction and controllability

1. If we can apply sufficient amount of force, it is better to apply force to ankle joint.
2. If not, hip actuator is a good alternative, even though it is hard to control
3. Biarticular actuator doesn’t assist loaded walking very well and the force input is not consistent.

Active actuators offer greater assist for loaded walking than unloaded walking when we can apply sufficient amount of force
Optimal input of ankle actuator is consistent with gait cycle and muscle forces data, while that of hip actuator is not.
The optimal input force for ankle actuator when it’s maximum force is bounded is similar to the general optimal input force saturated at maximum force

Limitations

The experimental data was obtained from a subject without exosuit. Exosuit may change the kinematics of a subject as well as GRF.
The simulation methodology to use CMC as an optimization tool works, but more improvement is needed.
CMC process doesn’t minimize metabolic cost. Instead, it minimizes 2-norm of activation.
The experimental data was only one gait cycle.
More realistic actuator simulation is needed. (E.g. Combination of passive & active actuator).

Source code

You can find the model that I created for the project in SimTK website. You can reproduce the result using the model on the website. Please visit http://www.simtk.org/~~~