The document describes a study that used an instrumented ramp and plantar pressure insoles to develop regression models for predicting complete ground reaction forces (GRFs) during inclined and declined walking. Subject-specific linear regression models were created to estimate GRFs from plantar pressure data at ramp angles from 0 to 20 degrees. The results indicate this method shows potential for determining GRFs in various settings where force plates cannot be used.
1. • Knowing the complete set of ground reaction forces (GRFs),
i.e. Anterior/Posterior, Vertical, and Medial/Lateral, during
walking is useful for various biomechanical analyses.
• Previous studies have shown that plantar pressure insoles can
be used to estimate the complete set of GRFs using linear
regression models [1,2]. However, the correlation between
pressure insole data and GRFs is unclear for inclined and
declined walking.
• A regression model for angled surfaces may allow the
estimation of GRFs for the study of hiking shoe design in the
field over mountainous terrain where force plates would not
be available.
Andrew Crechiolo, Jerrod Braman MS, Feng Wei PhD, Roger C. Haut PhD
Orthopaedic Biomechanics Laboratories, Michigan State University, East Lansing,
MI
Estimating Complete Ground Reaction Forces by Plantar Pressure Insoles
during Incline and Decline Walking
Force plate with
steel support rails
Build an instrumented ramp with the following requirements:
• Build and use an instrumented ramp with embedded force
plate and inclinations up to 20°
• Create subject specific linear regression models to predict
GRFs during incline and decline walking
• Subjects performed 10 walking trials for each condition (0º;
inclined and declined walking at 5º, 10º, 15º, and 20º) while
wearing a Merrell Trailglove minimalist shoe outfitted with
Novel Pedar-X plantar pressure measurement insoles
ACKNOWLEDGEMENT
The authors would like to thank Wolverine World Wide, Inc. for partial
support via a gift to the MSU Orthopaedic Biomechanics Laboratories.
Pressure insole and
minimalist shoe
𝐹(𝑝 𝑚𝑎𝑠𝑘) 𝑥,𝑦,𝑧
𝜃 = 𝑀 ∗ 𝑝 𝑚𝑎𝑠𝑘 + 𝐵
• 𝜃 and X,Y,Z denote which model is being used i.e. the ramp
angle and component of the GRF being predicted
(X=Medial/Lateral, Y=Anterior/Posterior, Z=Vertical)
• M and B represent the slope and intercept constants of each
equation
• 𝑝 𝑚𝑎𝑠𝑘is the averaged value from three pressure sensors that
were highly correlated to the respective GRF data recorded
from the force plate, at a given instant
Instrumented Ramp
• The instrumented ramp has shown to be a useful tool for
generating this project’s angled GRF prediction
regressions and various future studies involving inclined
and declined motion
• A linear staged regression model appears to be a viable
method for predicting GRFs under these test conditions
• 6 subjects tested at angles 0, 5, 10, 15, 20
• This method may serve as a viable option for determining
GRFs in various environmental settings
REFERENCES
[1] Fong et al., J of Biomechanics 41, 2008; [2] Rouhani et al., Gait & Posture 32,
[3] Weaver et al., J of Sports Eng and Tech 227:4, 2013
• Embedded AMTI force plate
• 0 -20º incline in 5º
increments
• Raising/lowering mechanism
• Minimize ramp deck
flex
• High traction surface
• Handrails
• Subject and angle specific models were developed using the
plantar pressure sensors that were highly correlated to their
respective GRF values.
Introduction
Objective
Methods
Results
Ramp angles: 0° 5° 10° 15° 20°
Discussion
Regression Model Format
Subject trial
-400
-200
0
200
400
600
800
1000
0 0.2 0.4 0.6 0.8 1
Force(Newtons)
Time(seconds)
Subject 4 - 0° GRFs vs Time
Vertical RMS error: 20.02 N - Medial/Lateral RMS error: 8.46 N - Anterior/Posterior RMS error: 21.37
-400
-200
0
200
400
600
800
1000
0 0.2 0.4 0.6 0.8 1
Force(Newtons)
Time (seconds)
S4 - 5° Incline GRFs vs Time
Vertical RMS error: 21.34 N - Medial/Lateral RMS error: 8.17 N - Anterior/Posterior RMS error: 22.73
-400
-200
0
200
400
600
800
1000
0 0.2 0.4 0.6 0.8 1
Force(Newtonws)
Time (seconds)
S4 - 5° Decline GRFs vs Time
Vertical RMS error: 25.89 N - Medial/Lateral RMS error: 16.32 N - Anterior/Posterior RMS error: 40.87 N
Subject 4 Graphs Key
-400
-200
0
200
400
600
800
1000
0 0.2 0.4 0.6 0.8 1
Force(Newtons)
Time (Seconds)
S4 - 10° Incline GRFs vs Time
Vertical RMS error: 26.39 N - Medial/Lateral RMS error: 7.27 N - Anterior/Posterior RMS error: 22.43 N
-400
-200
0
200
400
600
800
1000
0 0.2 0.4 0.6 0.8 1
Force(Newtons)
Time (seconds)
S4 - 10° Decline GRFs vs Time
Vertical RMS error: 15.17 N - Medial/Lateral RMS error: 11.14 N - Anterior/Posterior RMS error: 30.60
-400
-200
0
200
400
600
800
1000
0 0.2 0.4 0.6 0.8 1
Force(Newtons)
Time (Seconds)
S4 - 15° Incline GRFs vs Time
Vertical RMS error: 24.33 N - Medial/Lateral RMS error: 6.16 N - Anterior/Posterior RMS error: 23.09 N
-400
-200
0
200
400
600
800
1000
0 0.2 0.4 0.6 0.8 1
Force(Newtons)
Time (Seconds)
S4 - 15° Decline vs GRFs
Vertical RMS error: 24.92 N - Medial/Lateral RMS error: 11.76 N - Anterior/Posterior RMS error: 67.8 N
-400
-200
0
200
400
600
800
1000
0 0.2 0.4 0.6 0.8 1
Force(Newtons)
Time (Seconds)
S4 - 20° Incline GRFs vs Time
Vertical RMS error: 35.58 N - Medial/Lateral RMS error: 11.20 N - Anterior/Posterior RMS error: 36.64
-400
-200
0
200
400
600
800
1000
0 0.2 0.4 0.6 0.8 1
Force(Newtons)
Time (Seconds)
S4 - 20° Decline GRFs vs Time
Vertical RMS error: 45.31 N - Medial/Lateral RMS error: 14.61 N - Anterior/Posterior RMS error: 38.29 N
GRF Mean Values of RMS error & RMS error to Peak Value
Vertical Medial/Lateral Anterior/Posterior
RMS error (N) RMS error (N) RMS error (N)
0° 26.95 3.30 10.05 21.36 20.12 12.96
0° Fong '08 38.43 5 11.71 28 27.41 12
0° Cordero '04 27.84 NA 7.30 NA 7.53 NA
5°
Incline 35.28 4.10 9.94 27.63 19.59 8.82
Decline 35.23 3.92 14.01 23.83 30.87 14.80
10°
Incline 40.69 4.77 9.40 24.20 26.15 9.80
Decline 31.64 3.36 14.56 26.11 38.59 13.77
15°
Incline 37.29 4.45 8.92 20.94 29.08 9.36
Decline 33.46 3.63 12.47 19.35 42.30 12.71
20°
Incline 42.22 5.25 10.71 22.63 39.25 10.80
Decline 38.09 4.47 12.52 19.52 40.35 10.06
Denotes results of previous published 0° walking prediction methods
𝑹𝑴𝑺 𝒆𝒓𝒓𝒐𝒓
𝑷𝒆𝒂𝒌 𝑽𝒂𝒍𝒖𝒆
(%)
𝑹𝑴𝑺 𝒆𝒓𝒓𝒐𝒓
𝑷𝒆𝒂𝒌 𝑽𝒂𝒍𝒖𝒆
(%)
𝑹𝑴𝑺 𝒆𝒓𝒓𝒐𝒓
𝑷𝒆𝒂𝒌 𝑽𝒂𝒍𝒖𝒆
(%)