Inverse Dynamics Solver
Posted: Fri Aug 27, 2021 11:07 pm
Sir/Ma'am,
We are trying to use MATLAB-OpenSIM API commands for our code to get the joint moments at each time instant of gait motion. We are using Inverse Dynamics Solver for this. The issues we are facing are,
1. We used states (generalized coordinates and velocity) and generalized acceleration (udot) as the inputs to IDsolver at each time instant without any ground reaction force for Gait2354 model. The results does not match with the same generated using IDTool in OpenSIM GUI (without any grf). The code is attached with this message. We tried applying the same code for Arm26 model and it works.
2. How to apply GRFs as an input to IDSolver? Should we convert the ground reaction forces an torques to 'mobility force' or 'body force' and use them as input to IDSolver? Or we can apply those forces and torques using external point and torque actuators?
Attached files: Gait2354 present result for knee moment, Gait2354 OpenSIM result for knee moment without grf, Comparison to OpenSIM ID results for arm26 model, code.
Thanks and regards,
Sujash Bhattacharya
We are trying to use MATLAB-OpenSIM API commands for our code to get the joint moments at each time instant of gait motion. We are using Inverse Dynamics Solver for this. The issues we are facing are,
1. We used states (generalized coordinates and velocity) and generalized acceleration (udot) as the inputs to IDsolver at each time instant without any ground reaction force for Gait2354 model. The results does not match with the same generated using IDTool in OpenSIM GUI (without any grf). The code is attached with this message. We tried applying the same code for Arm26 model and it works.
2. How to apply GRFs as an input to IDSolver? Should we convert the ground reaction forces an torques to 'mobility force' or 'body force' and use them as input to IDSolver? Or we can apply those forces and torques using external point and torque actuators?
Attached files: Gait2354 present result for knee moment, Gait2354 OpenSIM result for knee moment without grf, Comparison to OpenSIM ID results for arm26 model, code.
Thanks and regards,
Sujash Bhattacharya
Code: Select all
filename = 'Walk 2.xlsx'; % Name of the file containing joint trajectories and GRF data
sheet = 1; % Sheet number of the excel file, is where the trajectory data is.
[NUM,TXT,RAW] = xlsread(filename,sheet); % Read the file and access data
[nRows, nCols] = size(RAW); % Calculate the number of rows and columns in the file. Number of columns is equal to degrees of freedom and one column for time. Number of rows is number of instances.
time = NUM(:,1); % First colmun is time data.
gencood_rad = NUM*2*pi/360; % All the data is in degrees, so converted them into radians
gencood_rad(:,5:7)= NUM(:,5:7); % PelvisTx,Ty,Tz are not angles but translation
gencood_rad = lowpass(gencood_rad,6,60);
v = (gencood_rad(2:(nRows-1),:) - gencood_rad(1:(nRows-2),:))/0.016666;
a = (v(2:(nRows-2),:)-v(1:(nRows-3),:))/0.016666;
%% Inverse Dynamics
import org.opensim.modeling.* % Import the OpenSim library
model = Model('subject01_simbody.osim'); % Import the lower body model
path='E:\Opensim\OpenSim 4.1\Geometry'; % Path to the folder where all the geometry files of the model is stored
ModelVisualizer.addDirToGeometrySearchPaths(path); % This path is given to the Model visualizer so that visualization of the model can be done
model.setUseVisualizer(true); % Visulaizer is true for the model
state = model.initSystem(); % Initialize the state of the model
% Setting the muscle property in ID_model such that no force is applied by them
for mn = 1:54
model.getMuscles().get(mn-1).setAppliesForce(state, 0);
end
% Following code snippet is necessary to run the visualizer
viz = model.updVisualizer().updSimbodyVisualizer(); % Start the visualizer for the given model
viz.setBackgroundColor(Vec3(1)); % Set the background colour of the model using RBG.
viz.setGroundHeight(0); % Set the height at which model will function
ID = InverseDynamicsSolver(model); % Inverse dynamics solver needs the model
u_dot = Vector(model.getNumCoordinates(),0); % desired acceleration of all the joints
for i = 1:size(a)
timed = time(i+1)-time(i); % Calculate the time step between two instances
for j = 2:nCols % For every instance, desired angle and speed
model.getCoordinateSet().get(j-2).setValue(state, gencood_rad(i,j));
model.getCoordinateSet().get(j-2).setSpeedValue(state, v(i,j));
u_dot.set(j-2,a(i,j)); % Desired acceleration of all the joints
end
tau = ID.solve(state, u_dot); % Solves the inverse dynamics
for k = 1 : model.getNumCoordinates()
obtained_tau(i,k) = tau.get(k-1); % 'tau' contains the joint moments
end
newT(i) = time(i);
manager = Manager(model);
state.setTime(0);
manager.initialize(state);
state = manager.integrate(timed);
end