Dani and Gaby

Team Members

• Dani Mendoza
• Gaby Uribe

Project Video

Presentation Video

Project Files

For this project we used unloaded and loaded walking data from Subject 09 trail 04 from Dembia et al. published work 1. 

Custom Static Optimization MATLAB code

Loaded IK results MOT file Subject 09 trial 04

Loaded subject 09 opensim scaled model

Loaded GRF XML file subject 09 trial 04

Loaded GRF MOT file subject 09 trial 04 

Unloaded GRF MOT file subject 09 trial 04

Unloaded GRF XML file subject 09 trial 04

Unloaded Subject 9 OpenSim Model Scaled in AddBiomechanics

Unloaded IK Results MOT File

Background

In the field of biomechanics, understanding the intricate mechanisms involved in human locomotion has long been a subject of interest. One fundamental aspect of this research is investigating how muscle activations vary between different walking conditions. In particular, this project looks to explore the disparities in muscle activation patterns during unloaded and loaded walking. This investigation aims to shed light on the adaptations that occur in the neuromuscular system when individuals carry external loads while walking over a flat surface. To achieve this, our project employs static optimization code, a powerful computational tool that utilizes mathematical optimization algorithms to estimate muscle forces and activations based on experimental motion data. By comparing muscle activations between unloaded and loaded walking using this approach, we hope to gain valuable insights into the motor control strategies employed by the human body in response to varying loads. Such knowledge may have implications in areas such as rehabilitation, sport performance, and ergonomics, potentially leading to the development of improved training techniques and interventions to enhance human locomotion under different loading conditions.

Research Question(s)

How do differences in ankle, hip, and knee kinematics cause changes in muscle activation patterns during loaded vs unloaded level walking? 

Methods

1. Scaled OpenSim models in AddBiomechanics. 
   • The model we used from Dembia et al. [1] is a 3-dimensional lower-limb model with 39 degrees of freedom (8 of them are locked). The loaded version of the model has an additional 38.73 kg "backpack" body that is welded to the torso. 
   • In Dembia et al. [1], the researchers manually scaled the model. AddBiomechanics is a new software that has the ability to replace labor-intensive manual scaling and marker registration. To see if the marker errors would be reduced compared to manual scaling, in this project, we looked to scale the unloaded and loaded models utilizing AddBiomechanics. Scaling the unloaded model in AddBiomechanics resulted in reduced marker errors. However, we were unable to scale the loaded model in AddBiomechanics because the software continuously looked to reduce the load of the backpack that we wanted to keep constant on the loaded model. Therefore, we used the manually scaled loaded model utilized in Dembia et al. [1].  

1. Carried out Inverse Dynamics using custom static optimization code above.
   •  The first link under project files above leads to the MATLAB code used for this project. The beginning sections of the code ask for the motion capture files output from Inverse Kinematics. For the unloaded model, the inverse kinematics output came from AddBiomechanics. For the loaded model, the inverse kinematics output came from the Inverse Kinematics tool in OpenSim. After inputting the inverse kinematic files into the MATLAB code, the code is able to conduct inverse dynamics for each model. 
2. Used static optimization to find muscle activations.
   • Part 5 of the MATLAB code attached above conducts static optimization to find the optimal muscle activations necessary in order to minimize activation squared, while still achieving the kinematic and dynamic constraints. 

1. Validated results with joint kinematics and muscle activations from Dembia et al. paper 1. 
   • As shown in the results section below, to validate our results from the static optimization code, we compared our results of joint kinematics and muscle activations to those published from Dembia et al.'s work using the same data set [1]. 

Results

It is important to note that we stitched the data in order to match the gait progression of the paper. Heel strike occurs at 0, the vertical lines show toe-off, and the remainder of the graphs show swing phase.

Joint Angles from Inverse Kinematics. Takeaway #1: Loaded walking results in greater hip, knee, and ankle flexion. 

Joint Moments from Inverse Dynamics. Takeaway #2: Generally, loaded walking results in greater moments of extension at the hip, knee, and ankle. 

Muscle Activations from Static Optimization

Glut Med Activation from Static Optimization

Takeaway #3: Changes in kinematics lead to increases in loaded walking muscle activations compared to unloaded walking. 

Limitations

One of the major limitations of our study that likely led to some discrepancies between our results and the results of the paper is the fact that our static optimization assumes a rigid tendon. The Dembia paper utilizes Computed Muscle Control (CMC) analysis, which involves an elastic tendon. In walking trials this tendon difference can definitely have an impact. For example, in walking, if the muscle-tendon complex is shortening, then our model's muscles would need to shorten at a faster velocity than a model with an elastic tendon, which would stretch. This quicker shortening would cause the possible output force of the muscle to decrease. Therefore, the muscles in our model would need to be activated more than the Dembia model in order to achieve the same forces and walking results.

Another limitation of our analysis is that, because of the data we had access to, we could not use the "matched" trials of loaded walking. Instead, we used the "free" trials for both loaded unloaded walking, meaning the walking speed was different between the loading conditions. This makes it a bit more difficult to compare the results. For the subject/trial we used, the subject walked at 1.16 m/s for the unloaded trial and walked at 1.08 m/s for the loaded trial. Since the subject walked (understandably) slower when wearing the backpack, then the differences in joint moments and muscle activations from the unloaded trial may have been less drastic than if the unloaded speed had been matched.

Future Work

For the loaded walking trials conducted in Dembia et al. [1], the subject carried a backpack (8 kg) and 3 weight vests containing lead (30 kg). The backpack worn by the subjects did not have a hip belt [1]. In the OpenSim model, the total load was modeled as a hollow cylindrical channel, with uniform density, welded to the torso [1].

In the future, we would aim to explore the effects of representing the load as a solid cylinder welded to the torso. Furthermore, in the future we seek to examine the influence of redistributing a portion of the load by welding it to the shoulders, as well as attaching additional weight to the back and front of the body. Simulating these different load configurations, would allow us to closely replicate the strain exerted on the body by backpack straps and weight vests. This endeavor promises to provide valuable insights into the resulting joint kinematics and muscle activation patterns, contributing to a more accurate understanding of the biomechanical response to external loads during flat surface walking. 

In addition to exploring the effects of load modeling on joint kinematics and muscle activations, it would be interesting to investigate the influence of load magnitude on gait biomechanics. By systematically varying the weight of the load and assessing its impact on joint kinematics and muscle activations, we can gain a comprehensive understanding of how load magnitude interacts with load placement. Furthermore, valuable insights are to be gained through the investigation of the temporal aspects of load carriage on gait biomechanics, examining how load duration and intermittent loading patterns may affect the body's response during walking. 

Lastly, only flat surface walking was investigated throughout this project. Future research may focus on how loaded vs unloaded walking gait mechanics change while walking up or down an incline.  

Acknowledgments

Big thank you to the ME485 teaching team: Scott Delp, Reed Gurchiek, Jon Stingel, Nicos Haralabidis, Nick Bianco, and Carmichael Ong. Without them, this project would not have been possible. 

References

1 Dembia, C. L., Silder, A., Uchida, T. K., Hicks, J. L., & Delp, S. L. (n.d.). Simulating ideal assistive devices to reduce the metabolic cost of walking with heavy loads. PLOS ONE. https://journals.plos.org/plosone/article?id=10.1371%2Fjournal.pone.0180320