Instructions

Intention�The Goals slides allow you to showcase your work, ask questions, and provide support to your Workshop colleagues.

Find your slide�Scroll through the slides and find your project title. Every project (whether for a team or individual) should have one or two slides only.

Edit your slide�Use the template slide as a guide for what information to include. Add text, photos and/or short videos (through YouTube).

Present your slide on Monday�Everyone will have 5 minutes (strict) to present their goals on Monday afternoon. After that we’ll all ask questions, share ideas, and give feedback.

​

​

Ian Stavness, Mohammad Shabani

Computer Science, Univ. of Saskatchewan

Modular Set Point Controller

Workshop Goals

• Build components: Controller, Spindle, GolgiOrgan, Delay (thanks to Chris!)
• Build StretchController
• Build generic SetPointController
• in vitro stretch test
• Arm model test (equilibrium-point)
• Bonus: Coordinate Stiffness Solver

​

​

​

Motivation

There is growing interest in modeling neurophysiological structures in OpenSim. Many “reflex” controllers built ad hoc.

Modularity key for adoption and adaptation.

Claudio Pizzolato and Luca Tagliapietra

​

Real-time kinematic estimates using inertial measurement units

Workshop Goals

• Implement the current IMU-based IK in real-time
• Moving to OpenSim 4.0
• Explore alternative solutions based on IMU raw data

Motivation

​

• Inertial Measurement Units (IMUs) are cheap and wearable.
• IMU can be used to measure human motion and drive musculoskeletal models
• Real-time IMU-based inverse kinematics (IK) has the potential to improve current rehabilitation practice by providing objective, accurate, and instant kinematics measurements to the clinician
• IK based solution enables the use of further modelling tools (e.g. muscle analysis, EMG-driven models, etc)

Current Status (OpenSim 3.3)

• Prototype algorithm for IMU-based IK using onboard orientations estimates. Drifting problems with some sensors
• Working software architecture for marker-based real-time IK

Michael Vignos, Scott Brandon, and Colin Smith

University of Wisconsin-Madison

​

Plug-In Development to Expand Functionality of OpenSim Contact Algorithm

Workshop Goals

• Write main program to perform scaling of contact geometries
• Transition scaling program into a plug-in
• Modify Simbody contact formulation to store contact pressures
• Export contact pressures to file suitable for visualization in PostView
• Transition this capability to a plug-in

Motivation

• Overall goal: Implement knee model with 6 DOF TF and 6 DOF PF joints into OpenSim
• Ability to scale contact geometries and export contact pressures on triangulated mesh critical for functionality of model

Andrew LaPre

​

Passive force element modeling

​

Workshop Goals

• Develop code that correlates passive generalized forces from ID with generalized motions from IK, given a user specified function form, and add a function based bushing force to the associated model.
• Identify whether to develop new class, or build on existing class
• Write code that first stores data
• Optimize function coefficients
• Update model
• This will be helpful for many different applications where modeling human-device interaction is wanted

Motivation

• The group’s focus is using predictive simulations to find new prosthesis mechanisms, optimized for limb loading.
• Predictive modeling will allow us to explore an enormous design space with rapid virtual prototyping.
• Device assist is only as good as the connection to the person.
• Accurate modeling of the physical interaction is needed.

William Thompson, NASA Glenn, Digital Astronaut Project

​

Sensitivity Analysis of Biomechanical Models of Resistance Exercises

​

Motivation

DAP uses modeling

• to predict and assess spaceflight health and performance risks
• to enhance countermeasure development.

Our research goal is to use modeling

• to provide quantitative answers to questions regarding astronaut health and performance on NASA missions
• to evaluate countermeasure efficacy on a localized stimulus basis

Current modeling efforts focus on

• Mitigating the risk that, given the small size of candidate exercise devices on Exploration missions to deep space, they may not be sufficient to provide the localized loading stimulus required to maintain musculoskeletal performance in microgravity.
• How does the choice of loading configuration (bar vs. harness) affect a countermeasure’s efficacy at providing localized stimulus at the same load?
• How sensitive are our models to inputs with significant uncertainty?
• Muscle parameters
• Body mass distribution
• Motion capture variance

​

William Thompson, NASA Glenn, Digital Astronaut Project

​

Sensitivity Analysis of Biomechanical Models of Resistance Exercises

​

Workshop Goals

​

• Use/expand the Probabilitistic Plugin (Davidson, Myers) and the MATLAB API for performing sensitivity analyses regarding the outcomes generated by ID, SO, Joint reaction analysis.
• Automate these analyses through effective use of the API and scripting.
• Use the Plugin and Matlab API to generate useful reports regarding model sensitivity.

These goals are important because a model that is shown to be robust to variable input parameters is more credible and becomes more useful for making mission operations decisions regarding devices used, load configuration, choice of stance or cadence and overall exercise prescription for crew.

Load Configuration Analysis

How sensitive is this result to kinematic errors and muscle parameter uncertainty?

Antoine Falisse and Gil Serrancoli

​

Dynamic optimization via direct collocation using implicit formulations and automatic differentiation

​

Workshop Goals

• Integrate automatic differentiation in OpenSim/Simbody
• Solve a simple optimal control problem combining direct collocation, implicit formulations and automatic differentiation in C++

Motivation

• Direct collocation is more efficient computationally than shooting methods
• Implicit formulations of the dynamics improve the numerical condition of the NLP
• Automatic differentiation increases the computational efficiency and the accuracy

Antoine Falisse and Gil Serrancoli

​

Dynamic optimization via direct collocation using implicit formulations and automatic differentiation

​

*Implicit formulations

Dimitar Stanev and Konstantinos Moustakas

​

Task oriented muscle control for inverse and forward simulations

​

Workshop Goals

• Present a prototype of the framework that can be implemented in OpenSim
• Work towards implementing inverse and forward simulation cases that can study simple movements following the task oriented framework

Motivation

Working in task space, provide the means to study the neuromuscular coordination in a different space (not only joint space). The pros of this transformation are:

• it is more intuitive to design new controllers (e.g. gait)
• the motion can be partitioned into individual tasks, thus can be studied separately
• can analyze the impact of the neuromusculoskeletal parameters on the performance of the individual tasks
• opens new opportunities to implement different algorithms to study the neuromuscular complex and to try to think a little bit out of the box

​

Sanjeev Datta

​

Integration of OpenSim with RT-Kinematic Data Captured from SHARP Sensor Suit Platform

Workshop Goals

​

• Create a framework for transmitting data from sensor suit data payload into OpenSim
• Visualize joint angle data within OpenSim (post-processed and then RT)

Motivation

• Lower-body motion capture suit developed to collect and transmit RT inertial data from COTS IMU’s
• Sagittal plane joint angles for hip, knee, and ankle
• Utilize OpenSim modeling to improve suit algorithm and aid in user performance analysis and injury prediction

Alexander Priamikov

​

OpenEyeSim – a biomechanical model

for studying the development of oculomotor control

​

Workshop Goals

​

• To implement a muscle band dynamics by constructing a component to account for a torque (couple) generated by the muscle due to its differential in length over the width of the muscle.

Motivation

For the aims of our studies on functional dependence between the development of visual cortex and the development of eye movement skills, we have created a biomechanical model of an oculomotor control. It models six extraocular muscles (EOMs) per eye as well as the elasticity of eye-surrounding tissues and muscle pulley dynamics. Virtual cameras placed in the eyes sense a 3D simulated computer graphics environment. This enables the simulation of both open-loop and visually guided closed-loop control of different kinds of eye movements.

The simplified representation of the oculomotor system allows for realistic eye movements but is not able to properly fit more detailed data available from static models of the oculomotor system. To achieve better agreement with this data we should consider each muscle as a band and model a band dynamics. As initial step we show, how different is the load on different sides of a band.

​

Mohammadhossein Saadatzi

​

Predictive simulation of exoskeleton-assisted bipedal walking

​

​

Workshop Goals

Predictive simulation of bipedal walking with

• Passive ankle exoskeleton of Collins et al. [1]
• Soft exosuit proposed by Lee et al. [2]�to optimize metabolic cost savings by varying exoskeleton parameters

Motivation

Use predictive simulation to inform design and control of lower-extremity exoskeletons for augmentation, rehabilitation or assistance.

• Narrowing down human experiment scenarios
• Predicting adaptation of human walking to augmenting torques/forces

[1] Collins et al. (2015). Reducing the energy cost of human walking using an unpowered exoskeleton. Nature, 522, 212–215.

[2] Lee et al. (2016). Controlling negative and positive power at the ankle with a soft exosuit. IEEE International Conference on Robotics and Automation (ICRA), 3509-3515.

Russell Johnson

​

Minimally Constrained Criteria for Generating Optimal Control Simulations of Human Walking

​

Workshop Goals

• Implement new objective functions within Matlab (including COM trajectory and joint loads) in simple models to be used in more complex models in the future

​

Motivation

• Develop a hybrid approach between tracking and predictive optimization solutions of gait
• Generate simulations of gait using high-level tracking tasks in Matlab with direct collocation
• Allow direct collocation to find realistic solutions to novel tasks

​

Brad Humphreys

​

Extending OpenSim Predictive Simulation Capabilities with Implicit Formulation of Dynamics

​

Workshop Goals

• Perform a forward implicit simulation
• Tug-of-war Model
• A. van den Bogert’s implicit muscle model
• Utilize ID for multibody dynamics
• Establishes foundation framework for further development of implicit dynamic components

Motivation

• Direct Collocation optimization provides an efficient means to perform task based optimization and predictive modeling
• Implicit system dynamics formulations
• Can be used to perform forward dynamics on stiff systems
• Preferred for use in optimization problems e.g. Direct Collocation.
• Both utilize and produce same mathematical quantities e.g. Jacobians (and better formed).

​

​

​

Ju Zhang

​

Patient-Specific OpenSim Model Generation Using MAP Client

​

Workshop Goals

• Learn about changes to Python API in 4.0
• Understand model scaling API
• Prepare experiment for validating MAP-Client customised OpenSim models.

Motivation

• Making patient-specific OpenSim models is time consuming
• MAP Client automatically customises OpenSim models (Gait2392) from marker and imaging data using statistical models
• Bodies, joints, muscle insertion/via points
• Not all bodies are currently scaled
• Requires validation
• Compatibility with OpenSim 4.0?

​

​

​