USING HIGHTHROUGHPUT COMPUTING FOR DYNAMIC SIMULATION OF BIPEDAL - - PowerPoint PPT Presentation

▶
using high throughput computing for dynamic simulation of
SMART_READER_LITE
LIVE PREVIEW

USING HIGHTHROUGHPUT COMPUTING FOR DYNAMIC SIMULATION OF BIPEDAL - - PowerPoint PPT Presentation

Jan 21, 2023 24 likes •325 views

USING HIGHTHROUGHPUT COMPUTING FOR DYNAMIC SIMULATION OF BIPEDAL WALKING Mohammadhossein Saadatzi, Anne K. Silverman, Ozkan Celik Outline Simulation framework for combined human bipedal walking and wearable robots Introduction

slide-1
SLIDE 1

USING HIGH–THROUGHPUT COMPUTING FOR DYNAMIC SIMULATION OF BIPEDAL WALKING

Mohammadhossein Saadatzi, Anne K. Silverman, Ozkan Celik

slide-2
SLIDE 2

1  Simulation framework for combined human bipedal

walking and wearable robots

 Introduction  Problem definition  Using HTC to solve the problem

 Applications of the prepared framework

 A passive hip exoskeleton  Walking energetics at different speeds

Outline

slide-3
SLIDE 3

2

Introduction

 Previously, high-throughput computing (HTC) was used

for stochastic simulation and analysis of human movements [1,2].

 By considering parametric uncertainty in musculoskeletal model,

HTC was used to investigate cartilage loading in the knee during movement.

 Simulations of movement are traditionally performed

using generic models with assumed parameters and geometries.

 Population variability and parametric uncertainty become

critically important to consider when interpreting model predictions [1,2].

[1] Smith CR, et al. J Knee Surg, 29 (2), 99-106, 2016. [2] Smith CR, et al. J Biomech Eng, 138 (2), 021017, 2016.

slide-4
SLIDE 4

3

Introduction

 Simulations and dynamic optimization for investigation

human biomechanics and human-robot interaction.

 We have developed a

framework for combined simulation of human bipedal walking and exoskeletons [1].

 Simulation of human bipedal walking through muscle

actuation presents an optimization problem with a substantial number of parameters [2,3].

[1] Saadatzi M, et al. Proceedings of 41st ASB, 451–452, 2017. [2] Miller RH. J Biomech, 47 (6), 373–1381, 2014. [3] Umberger BR, J. Royal Soc. Interface, 7 (50), 1329-1340, 2010.

slide-5
SLIDE 5

4

Introduction

 Calculation of the cost function for this optimization

problem, necessitates significant computational power [1].

 Because of the redundant nature

  • f the human musculoskeletal system,

the problem is highly prone to getting stuck in local minima.

 Researchers have chosen to solve the

problem several times [2-5].

 Miller [2], single step, 4 times, 2.5 days using HPC.  Umberger [3], single step, 10 times.  Ong et al. [4], 10 seconds, 20 times.

 They used the best solution to seed another round of 20 optimizations.  This process was repeated until the change was less than 5%.

[1] Ackermann M, van den Bogert AJ, J Biomech, 43 (6), 1055–1060, 2010. [2] Miller RH. J Biomech, 47 (6), 373–1381, 2014. [3] Umberger BR, J. Royal Soc. Interface, 7 (50), 1329-1340, 2010. [4] Ong CF, et al. IEEE Trans Biomed Eng, 63 (5), 894–903, 2016. [5] Ong CF, et al. Proceedings of ISB-TGCS, 19–20, 2017.

slide-6
SLIDE 6

5

Introduction

We use HTC as a tool to perform computations related to predictive simulation of human bipedal walking through muscle actuation.

slide-7
SLIDE 7

6

Problem definition

[1] Farris RJ, et al. IEEE Trans Neural Syst Rehabil Eng, 19(6), 652-659, 2011. [2] Delp SL, et al. IEEE Trans Biomed Eng, 54 (11), 1940-1950, 2007.

Exoskeleton structure and control parameters

Human-exoskeleton kinematic and dynamic data

Optimization Algorithm Objective function Forward dynamic simulation Actuation profile Human-exoskeleton combined model

slide-8
SLIDE 8

7

Problem definition:

Human-exoskeleton model

The musculoskeletal model has 9 sagittal degrees of freedom and 18 muscle groups (24 muscles):

 DOFs: pelvis tilt, two planar pelvis translations,

hip and knee flexion/extension, ankle plantar/dorsiflexion

 Muscle groups: soleus, vasti,

gastrocnemius, tibialis anterior, bicep femoris short head, hamstring, rectus femoris, gluteus maximus, and iliopsoas.

slide-9
SLIDE 9

8

Human-exoskeleton combined model

Problem definition

[1] Farris RJ, et al. IEEE Trans Neural Syst Rehabil Eng, 19(6), 652-659, 2011. [2] Delp SL, et al. IEEE Trans Biomed Eng, 54 (11), 1940-1950, 2007.

Exoskeleton structure and control parameters

Human-exoskeleton kinematic and dynamic data

Optimization Algorithm Objective function Actuation profile Forward dynamic simulation

slide-10
SLIDE 10

9

Problem definition:

Forward dynamic simulation

Two controller states are considered, corresponding to stance and swing phases.

Swing

Heel strike

Stance

Toe off Heel strike

slide-11
SLIDE 11

10

Human-exoskeleton combined model

Problem definition

[1] Farris RJ, et al. IEEE Trans Neural Syst Rehabil Eng, 19(6), 652-659, 2011. [2] Delp SL, et al. IEEE Trans Biomed Eng, 54 (11), 1940-1950, 2007.

Human-exoskeleton kinematic and dynamic data

Optimization Algorithm Objective function Forward dynamic simulation Actuation profile Exoskeleton structure and control parameters

slide-12
SLIDE 12

11

Problem definition:

Actuation/exoskeleton parameters

Actuation profile of each muscle is defined as two piecewise linear functions

 One with 6 parameters for stance phase,

and

 One with 4 parameters for swing phase.

The values of signals at discretized points constitute the optimization parameters.

Exoskeleton structure and control parameters

slide-13
SLIDE 13

12

Human-exoskeleton combined model

Problem definition

[1] Farris RJ, et al. IEEE Trans Neural Syst Rehabil Eng, 19(6), 652-659, 2011. [2] Delp SL, et al. IEEE Trans Biomed Eng, 54 (11), 1940-1950, 2007.

Exoskeleton structure and control parameters Optimization Algorithm Forward dynamic simulation Actuation profile Objective function

Human-exoskeleton kinematic and dynamic data

slide-14
SLIDE 14

13

Problem definition:

Objective function

 A weighted combination of several cost

functions constitutes our optimization goal:

 Duration of walking  Target velocity error  Excessive alternating motion of the upper body  Metabolic expenditure of muscles

Objective function

slide-15
SLIDE 15

14

Human-exoskeleton combined model

Problem definition

[1] Farris RJ, et al. IEEE Trans Neural Syst Rehabil Eng, 19(6), 652-659, 2011. [2] Delp SL, et al. IEEE Trans Biomed Eng, 54 (11), 1940-1950, 2007.

Exoskeleton structure and control parameters

Human-exoskeleton kinematic and dynamic data

Objective function Forward dynamic simulation Actuation profile Optimization Algorithm

slide-16
SLIDE 16

15

Problem definition:

Optimization Algorithm

 Covariance Matrix Adaptation (CMA) [1].

 Previously has successfully been used for

simulation of bipedal walking [2-4].

 Available with OpenSim API.

 We use the computed muscle control

(CMC) method to generate favorable initial optimization parameters,

 which reduces convergence time and

probability of getting stuck in local minima.

[1] Hansen N, et al. Evol Comput, 11(1), 1–18, 2003. [2] Wang JM, et al. ACM Trans Graph, 31(4), 2012. [3] Dorn TW, et al. PLOS ONE, 10 (4), 1-16, 2015. [4] Ong CF, et al. Proceedings of ISB-TGCS, 19–20, 2017.

Optimization Algorithm

slide-17
SLIDE 17

16  HTC enables the use of many computers and cores

simultaneously

 Ideally, we would like to combine the results from several

computers to reduce

 necessary time for the convergence of our optimization  probability of getting stuck in local minima

Methods

[1] Sfiligoi I, et al. IEEE-CSIE, 428–432, 2009. Pordes, R et al. J Phys Conf Ser 78 (1), 012057, 2007. [2]

HTC

slide-18
SLIDE 18

17

Methods:

High-level algorithm

 We run several instances of our optimization, and combine

their results periodically to generate a more efficient

  • ptimization algorithm.

 We seed a round of optimization on remote computers with different

sets of initial optimization parameters.

 We compare and combine their results to generate initial

  • ptimization parameters for subsequent optimizations.

Initialization Compare & combination Optimization Optimization Optimization Convergence Detection Submit server Remote computers

slide-19
SLIDE 19

18  Directed Acyclic Graph Manager (DAGMan): a meta-

scheduler for the execution of programs (computations)

 DAG file is used by DAGMan when submitting programs.

Methods:

High-level algorithm

PS.dag # Predictive Simulation DAG Job PSitr PSitr.sub SCRIPT PRE PSitr initPop.sh SCRIPT POST PSitr nextItrPrep.sh RETRY PSitr 10 InitPop.sh Prescript Postscript Job

J1 J2 J3 Jn J4

PSitr: PSitr.sub Retry NextItrPrep.sh

slide-20
SLIDE 20

19

Methods:

High-level algorithm

 Prescript bash “initPop.sh”

generates the initial population.

 Postscript bash

“nextItrPrep.sh” generates the population for the next iterations.

 We used the best results to

seed the following iterations of

  • ptimizations.

 Job “PSitr”, runs the sub-

jobs, which consist of the

  • ptimizations of each round.

InitPop.sh Prescript Postscript Job

J1 J2 J3 Jn J4

PSitr: PSitr.sub Retry NextItrPrep.sh

slide-21
SLIDE 21

20

Methods:

Individual jobs

PSitr.sub # Jobs submit file

executable = myExe.sh arguments = $(Process) transfer_input_files = bipedWalkingOptFiles.tar.gz, pop/bestParametersBinary.$(Process) transfer_output_files = bestSoFarLog_$(Process).txt, bestParametersBinary_$(Process) periodic_remove = (CurrentTime - QDate) > 4200 queue 100

slide-22
SLIDE 22

21

Results & discussion

 We tested different intervals of running individual jobs, and explored fixed

and adaptive initial step-size of the CMA algorithm

 OSG 8-core computers (3.89) and AWS 36-core instance (3.86). Doesn’t walk the entire duration Walks the entire duration, when gets smaller becomes more realistic OSG 8-core computers, and AWS 36-core computer

slide-23
SLIDE 23

22

Conclusion & future work

 Our results suggest that HTC is a promising tool for

solving computationally expensive optimization problems, which arise during forward dynamic simulation of human movement

 Using conventional algorithms for combining the results of

remote computers

 In [1], a HTC based distributed genetic algorithm is presented for

solving building energy optimization problem.

 In [2], a pattern search optimization method is used for combination

  • f HTC resources to maximize capacity of electrical transmission

lines.

[1] Yang C, et al. Energ Buildings, Anderson EJ, Linderoth J, IEEE Trans. Smart Grid, 8 (3), 1427–1435, 2017. 76, 92–101, 2014. [2]

slide-24
SLIDE 24

23  Simulation framework for combined human bipedal

walking and wearable robots

 Introduction  Problem definition  Using HTC to solve the problem

 Applications of the prepared framework

 A passive hip exoskeleton  Walking energetics at different speeds

Outline

slide-25
SLIDE 25

24

 Iliopsoas performs negative work in the first half of the gait cycle.

 Could potentially be replaced by a torsional spring.

 Energy stored in the spring during the first half of the stride can be used during

the rest of the stride to fulfill the positive work normally done by Iliopsoas.

[1] S. H. Collins et al., Nature, vol. 522, no. 7555, pp. 212–215, Jun. 2015.

10 20 30 40 50 60 70 80 90 100

Hip (Deg)

  • 40
  • 20

20

Joint Kinematics Stride Phase (%)

10 20 30 40 50 60 70 80 90 100

Muscle Force (BW)

1 2

ILPSO

Estimated from Experimental Data Simulated

Negative work Positive work

Example: A Passive Hip Exoskeleton

slide-26
SLIDE 26

25

 Based on our framework, an optimal spring is predicted to decrease the metabolic

cost of walking by 10.3%.

Stiffness of 0.99 (Nm/deg)

Equilibrium angle of 10.04 (deg) flexion  The hip exoskeleton replaces majority of the work done by the Iliopsoas muscle.

[1] Uchida TK ,et al. PLOS ONE 11(9), 1-16, 2016.

Example: A Passive Hip Exoskeleton

Normal With Exo Average metabolic power (W/kg) 1 2 3 4 5 6 7 ILPSO GMAX Other Muscles

slide-27
SLIDE 27

26

 Our simulation framework captured a realistic parabolic trend

across speed

 Among the seven simulated speeds, the minimum cost of

transport took place at speed 1.25 m/s (within the range reported in experimental studies).

New simulations: Walking energetics at

different speeds

Speed (m/s)

0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4

Cost of Transport (J/kg.m)

2.5 3 3.5 4 4.5 5 5.5 6 6.5

Ralston 1958 (experimental) Martin 1992 (experimental) Browning 2006 (experimental) Simulation results

  • 10. Martin PE, et al. J Appl Physiol 73: 200-206,1992.
  • 11. Browning RC, et al. J Appl Physiol 100: 390-398, 2006.
  • 12. Ralston HJ, Int Z Angew Physiol 17: 277-283,1958.
slide-28
SLIDE 28

27

V = 1.25 m/s T = 0.52 s Distance = 0.65 m V = 0.5 m/s T = 0.74 s Distance = 0.38 m V = 2 m/s T = 0.37 s Distance = 0.73 m

slide-29
SLIDE 29

28

Thank You

slide-30
SLIDE 30

USING HIGH–THROUGHPUT COMPUTING FOR DYNAMIC SIMULATION OF BIPEDAL WALKING

Mohammadhossein Saadatzi,

Mechanical Engineering Department, Colorado School of Mines

Problem:

  • Simulation of human bipedal walking through muscle

actuation presents an optimization problem with a substantial number of parameters.

  • Because of the redundant nature of the human

musculoskeletal system, the problem is highly prone to getting stuck in local minima. Solution: We use HTC for robust and efficient computations of predictive simulation of human bipedal walking through muscle actuation.