AMMR 2.3.3 (2021-03-24)


  • Fix a problem single leg models in AnyMocap. The file OptimizeAnthropometricsOnOff.any preventing models with only one leg included from being loaded.

  • Removed two unused measures with references to thorax in the scapula and clavicula segments. The references could cause problems for ´AnyKinCOM` which uses search when including segments. Thanks to Handa Kensuke from Terrabyte helping fix this error.

  • Fix bug in initial wrapping guess for Triceps LH muscle wrapping. This bug could cause the left side Triceps LH muscle to wrap incorrectly at shoulder.

  • Fix calibration of of element 4 and 5 of the latissimus dorsi muscle. These two stands of the muscles where not included correctly in the calibration study when using the three element muscles model.


AMMR 2.3.2 (2021-01-21)


Model results can change!

A bug related to scaling was fixed in AMMR 2.3.2. This affect models using the length-mass-fat and uniform scaling, when only Body Height is specified. The results can change as models more accurately match the requested body height.


  • Fixed tibialis anterior/posterior variables in the SelectedOutput section. The force values were represented as the max() of the muscle elements instead of the sum(). Thanks to Dr. Enrico De Pierie from University Children’s Hospital Basel for fixing this issue.

  • Fixed a bug when using _SCALING_LENGTHMASSFAT_ and _SCALING_UNIFORM_ and only specifying the body height as input. The segment lengths were set to slightly smaller values, which caused the total body height to be less than what was used as input. The bug occurred since a standard body height in some ‘AnyMan’ files was assumed to be 1.8 m instead of the correct 1.75 m. This fix aligns the results with _SCALING_XYZ_ (used in the MoCap models) and _SCALING_STANDARD_ (50 percentile) which were both correct.

  • Fixed a problem with foot support in the “Spine Pressure validation model” lying flat on the back. The foot is now supported in the right direction.

  • Fixed a problem with the Standing Model template which was using an alternate mode where foot constraints were disabled.

  • Fixed a problem with the ADL gait MoCap example, where the batch processing script would report some trials as failed even if they succeeded correctly.

  • Use correct times in the Xsens example MoCap (trial “S01-Trial2”). The model stopped walking after a number of time steps because tEnd was set after the last data point in the BVH data file.

  • Fixed/improved the wrapping surface for the subscapularis muscles. The muscles can no longer slide off its wrapping surface in extreme range of motions.

  • Fixed infraspinatus muscle wrapping for high arm positions. This fix adds a torus for infraspinatus wrapping on the humerus. It fixes a problem where infraspinatus could slide of its wrapping surface in high arm postures.

  • Fix problem with latissimus dorsi muscle wrapping in over-head arm postures. A new wrapping ellipsoid has been added at the humeral head.


  • The rectus femoris muscle wrapping surface at the hip has been adjusted to prevent collision between the rectus femoris origin and the wrapping surface. Further, the surface is now used for both the straight and reflected head of the rectus femoris.

  • Increase the discretization (StringMesh) on approx 30 muscles to improve wrapping.

  • The results for the “Wilke Spine Pressure” validation model has been updated due to the fix for the _SCALING_UNIFORM_. The updated model improved the validation results slightly.

  • The new large scale ” MoCap model (ADL gait)” has been updated so the results of running all standing reference trials have been added to the model. This means that all the dynamic trials will work even if the user forgets to first run the standing reference trials. It is still necessary to run the standing reference trials again if any changes are made to the model.

  • The wrapping for the “Triceps Long Head” muscle has been updated at the humeral head to avoid the muscle wrapping incorrectly in over-head arm postures. This fix also adds a via point at the middle of the humerus to make the wrapping solution more robust.

  • The VideoLookAtCamera camera class template in the AMMR now saves videos as they look in the model view. I.e. if things are hidden in the model view they will not show up in the video. This feature is enabled by a new setting AnyCameraLookAt.RenderUserInterfaceViewState in AnyBody version 7.3.2. The old behavior can be restored by setting Camera.RenderUserInterfaceViewState=Off; in the class template.

AMMR 2.3.1 (2020-09-30)


  • All models now run with the new wrapping solver by default, which makes all models with wrapping muscles run much faster. Hence the AnyBody73_ExperimentalShortestPathSolver setting is deprecated and no effect. The old wrapping algorithm can be enabled (not recommended) by setting:

    System.Compatibility.ShortestPathSolverVersion = 2; // 4 is default (new wrapping)
  • New dedicated wrapping ellipsoid for Triceps at the glenohumeral head. This addition solves a few problems with overhead arm postures.


  • Improvements to the new “ADL Gait” example. The model now has muscles enabled and the marker protocol has been adjusted for a better shoulder posture. Also, the batch processing script now runs all trials without hickups.

  • Improve the shoulder rhythm. Scapula position when using the shoulder rhythm is now determined by the Conoideum ligament driver, like how scapula kinematic is controlled with no shoulder rhythm. The rhythm is now controlled based on the plane of elevation and humerus elevation from the ‘ThoraxHumerus’ measures in the interface folder. This tracks the original specification of the shoulder rhythm in “De Groot, J. H. The shoulder: a kinematic and dynamic analysis of motion and loading. (1998)”. One difference is still shoulder protraction/retraction where regression coefficients needed to be tweaked for the shoulder to work in the full range of motion. De Groot et al (1998) had only based their shoulder rhythm on arm position in front of the body (e.i. positive elevation planes).

  • Removed metabolic power (Pmet) contributions from artificial muscles defined as AnyMuscle derived classes. Pmet contributions from tools, such as GRF prediction, modelled as AnyGeneralMuscles will no longer be part of studies’ Pmet calculations.

  • Adjusted the wrapping segments for deltoid posterior so it doesn’t affect the muscle in overhead arm posturess.

  • Teres minor now uses a torus for wrapping at the humeral head.


  • Fixed incorrect values in “SelectedOutput” for four reaction moments variables, which were wrong due an implementation change of AnyRevoluteJoint. The following variables in SelectedOutput.Trunk.JointReactionForce were incorrect.

    • C2C1FlexionExtensionMoment

    • C2C1LateralMoment

    • C1C0AxialMoment

    • C1C0LateralMoment

    The values were wrong because they referered to the internal generalized force output of AnyRevoluteJoint. (i.e. Constraints.Reactions.Fout). The values Fout[3] and Fout[4] can not be interpreted as real physical reaction moments because AnyRevoluteJoint is implemented using euler angles.


    The use of generalized forces (Fout) is discouraged, and will be removed in the AMMR. Instead use the force and moment variables in AnyRefFrameOutput subfolder instead.

  • Fixed incorrect volume of Satorius muscle in TLEM1 and TLEM2. Thanks to Dr. Adam D. Sylvester from Johns Hopkins School of Medicine and Dr. Patricia A. Kramer from the University of Washington for pointing out the error. The error occurred since the satorius in the original TLEM1 paper was muscles in series with a pseudo insertion/origin on the femur. Both of these muscle elements were therefore listed with the full PSCA of Satorius. This detail was missed in the AnyBody TLEM1 and TLEM2 implementation where Satorius was implemented two parallel elements (TLEM1) and a single element (TLEM2). Hence the Satorius had twice the correct volume.

  • Fixed a problem when C3D files are malformed and doesn’t contain a FORCE_PLATFORM.ZERO variable. The C3D standard specifies that the variable should exist but since it not used in the models the reference could just be remvoed.

  • Fixed an problem which prevented the ANSUR scaling plugin example from working correctly 7.3.0

  • Fixed incorrect output of ThoraxHumerus interface measures. The mesure wasn’t used by other parts of the AMMR until now so the bug had no consequences in previous AMMR versions.

  • Fixed mass calculation of the foot and talus segments in the GaitVaughan validation model.


  • Removed unused BM_TRUNK_CERVICAL_LIGAMENTS switch which had no effect since there were no ligaments defined in the neck model.

AMMR 2.3 (2020-07-07)


  • Added a new Full Body “ADL Gait” MoCap Example based on the “Rehazenter Adult Walking Dataset” by Schreiber and Moissenet (2019). The model is configured to run all 50 subjects and 1145 trials in the data set. However, you must download the actual data separately from FigShare where it is hosted under a Creative Commons license.

  • New GUI plugin which uses statistical information from the ANSUR database to set anthropometric values in AnyBody. A small example model is added Examples/StatisticalScalingPlugin which shows how to use the plugin.

  • Added a new option to optimize “head width” / “head depth” / “neck length” / “Tibia torsion” / “hand breadth” as part of the OptimizeAnthropometricsOnOff class_template in the AnyMoCap framework.

  • Added new rotation vector based measures in the interface folder for the arm degrees of freedom. These measures (Interface.Right.RotVectorMeasures.*`) for joint angles do not have a clinical meaning but are useful when exporting and importing joint angles.

  • Added a new option for setting custom muscle calibration. I.e. an option which disables calibration so the user can add their own code.

    #include "<ANYBODY_PATH_BODY>\HumanModel.any"
    // Manually include calibration code.
    Main.HumanModel.Calibration = {
      #include "<ANYBODY_PATH_BODY>/LegTLEM/Calibration/CalibrationSequenceRight.any"
      #include "<ANYBODY_PATH_BODY>/LegTLEM/Calibration/CalibrationSequenceLeft.any"
      #include "<ANYBODY_PATH_BODY>/Arm/Calibration/CalibrationSequenceRight.any"
      #include "<ANYBODY_PATH_BODY>/Arm/Calibration/CalibrationSequenceLeft.any"


    If you need to modify the calibration code, copy it to the application folder and include it from there.


  • All application examples use the new and fast experimental wrapping algorithmn in the AnyBody Modeling System 7.3.0. A notice is shown in the output window to remind the users that the algorithm is enabled. The experimental wrapping algorithm can be used in all models by setting:

    System.Compatibility.AnyBody73_ExperimentalShortestPathSolver = On;

    See also

    The release notes for the AnyBoy Modeling System has more information on the new and faster experimental wrapping algorithm.

  • Split the SternoCleidomastoid muscle up into two muscle branches. This makes the two heads of the muscle able to work independent.

  • Add extra muscle-via points on for the Latissimus dorsi on the Thorax segment. This change fixes wrapping problems where the Latissimus dorsi muscle strands would slide off the wrapping surface representing the thorax segment.

  • Renamed all deprecated muscle class names. AnyViaPointMuscle-> AnyMuscleViaPoint and AnyShortestPathMuscle-> AnyMuscleShortestPath

  • The AnyMoCap framework now uses a more robust way of saving optimized parameters from the parameter optimization studies, when using AMS 7.3.

  • Pectoralis Major has been split from 5 into 10 branches to better represent the real fiber direction of the muscle. Similarly the origins of the new Pectoralis Major muscles have been moved to align with real anatomy of the muscle.

  • Changed the format used by the AnyMoCap models when transferring joint angles from the mrker tracking step to inverse dynamics simulation. The joint-angles for the arms and pelvis rotation are now written in terms of rotation vectors instead of Euler angles. This fixes potential singularity problems. The temporary joint angle files produced by the MoCap models have been renamed to indicate the data is no longer Euler angles.


    This means that the data in these temporary files can not easily be interpreted from a clinical point of view. Create your own output files with physiological joint angles if you need to export that data.

  • The body model configuration plugin have been updated. This fixes a number of smaller issues and bugs.

  • The via points of PeroneusLongus and PeroneusBrevis have been adjusted to ensure the muscles maintain moment arm in a larger range of motion.

  • Rectus abdominis muscle has been split into eight branches to better represent geometry, than what is possible with one branch. The strength between the new branches has been split evenly. Additionally, the PCSA has been increased from \(2.6 cm^2\) to \(7.9 cm^2\) according to McGill et al 1988, the \(2.6 cm^2\) originates from a study with elderly subjects, the McGill study used mid age subjects.

  • Adjusted the origin of Serratus Anterior 1 so it aligns with the geometry of the rib on which it originates.

  • Changed the way the deltoid wrapping surfaces move relative to humerus. This changes makes it wrapping more robust to deltoid flipping over to the wrong side of the wrapping surface in extreme postures.

  • The BVH virtual marker protocol have been adjusted. If older BVH models is used with AMMR 2.3 a warning/error is triggered to warn the user that the protocol needs to be updated.


  • Fixed problems with the following variables in the SelectedOuput folder: HipAbductorMuscleForce, HipExtensorMuscleActivity, HipExtensorMuscleForce, KneeFlexorMuscleActivity and KneeFlexorMuscleForce.

  • Markers on BVH stick-figure model in the AnyMoCap framework now respects the setting: Main.ModelSetup.BVHFileData.ModelDrawOnOff.

  • Fixed occasional kinematics problems with FullBody MoCap models during inverse dynamics simulation. This was fixed by changing the internal representation of the shoulder joint angles when transferring kinematics from marker tracking to the Inverse Dynamics simulations.

  • Fixed a problem with the Body Model configurator plugin showing reporting: “No Human Model” when a configuration file already exists.

  • Corrected the location of via points of Flexor_Digitorum_Superficialis_Digit2 to avoid wrapping surface collisions in certain hand positions.

  • Corrected the location of via points of Flexor_Digitorum_Superficialis_Digit3 and 4, Palmaris_Longus, Extensor_Carpi_Radialis_Brevis and Extensor_Digitorum_Digit2 to avoid wrapping surface collisions in certain hand positions.

  • The muscle via point representing the bicipital groove have been adjusted to ensure the Biceps Caput Longum wraps the humeral head correctly in

  • Fixed a bug in the FreePosture model, where input for the Left/Right arm drivers was switched in some places.

  • Fixed missing visual color indication of force plate forces for type 3 force plates in the MoCap models.

  • Fixed a bug with in the built-in studies to evaluate arm joint strength. The range of motion for the left arm elbow pronation/supination were not correct.

  • Fixed a bug preventing the model from loading with with _SCALING_XYZ_ and both legs excluded.

  • Fixed the load-time position of the head segment when neck scaling is changed.

  • Fixed a problem with visualization in GRF prediction class which prevented GRF prediction to be used with moving base frames (i.e. treadmills). Thanks to Marco Antonio Marra (University of Twente) for fixing this.

  • Fixed the marker node orientation for BVH (inertial based) MoCap models in the CreateMarkerDriver class template. Marker orientations are now handles identical to normal C3D based markers.```

  • Fixed incorrect wrapping of teres minor and teres major at extreme shoulder range of motion. New wrapping surfaces have been added. One in Humerus and one on scapula. The meet at the glenohumeral joint to prevent the muscles from sliding to the wrong side of the humerus.


  • Removed unused support files for muscle contact. Tools/ModelUtilities/ContactMuscles/*

AMMR 2.2.3 (2019-11-13)


  • Add new option to set the type of wrapping surface used in the deltoid wrapping implementation. The following will force the use of cylinders (which is faster) instead of ellipsoids.


    See BM_ARM_DELTOID_WRAPPING for possible options.


  • Fixed problems with BVH models that have jumping angular data due to multiple solutions to the Euler angle equations when gimbal locks occur. The fix requires a new structure of the BVH model, where virtual marker trajectories are calculated by the AnyInputBVH class. See the updated BVH_Xsens example, and port your existing BVH models to the new example structure.

  • Fixed a problem in AnyMoCap models, where model view operations in Main.ModelSetup.Views.SetViewMacros would not trigger a redraw of the model view when executed directly.


  • Increase the time offset when running MoCap inverse dynamics simulations. The offset was 2 frames which were skipped at the start/end of inverse dynamics simulations. But the accelerations of 4th order b-spline are only stable on the fourth frame, and hence the offset has been increased to three frames instead of two.

  • ScalingXYZ scaling law has been extended to respect the following new variables in the HumanModel.Anthropometrics.SegmentDimensions folder: (HeadWidth, HeadDepth, TrunkDepth, NeckLength, FootWidth)


  • Remove the joint muscles’ strength dependency on the range of motion. Joint muscles are the torque actuators embedded in the joints when the model is loaded without muscles. Previously, joint muscles would reduce their strength to 10% when the joint was outside the range of motion. The lower strength outside the range of motion was originally implemented to prevent inverse-inverse simulations finding non-physiological solutions. However, this problem is now better solved with kinematic joint limits, and the change in joint strength could sometimes cause models to fail with a muscle recruitment failure.

  • Remove the warning for modified Body model files. The warning can still be enabled by setting the switch #define ANYBODY_ENABLE_MODIFIED_AMMR_NOTICE


  • Added a warning to MoCap models when the number of frames is below 8. Since the low number of frames causes problems with both low-pass filters and b-spline interpolation. Note: C3D files with only a single frame works fine since that is handled specially and often used for static trials.

AMMR 2.2.2 (2019-09-12)


  • Fixed kinematic problems when driving specific wrist-flexion/abduction values. The definition of wrist flexion in BodyModel.Interface now measures the flexion between the radius and small artificial WristJointSeg segment between the radius and the hand segment. This ensures that there is always unique solution when driving the wrist flexion and abduction.

  • Fixed a bug in the AnyMoCap examples when setting BM_SCALING to _SCALING_LENGTHMASSFAT_.

  • Fixed a number of smaller bugs with Body Model configuration plugin.

  • Fixed a bug in calibration of 3 element leg muscles in the TLEM1 and TLEM2 model. The problem was that the first calibration study (LegCal1) included muscles which were also calibrated specifically in later calibrations. This could in some cases (certain scaling) lead to a problem where LegCal1 would be forces to change the fiber-length even though later specific calibration of the muscle did not need that. Now, LegCal1 only calibrates muscles which are not included in the later calibration studies.


  • Changed the posture of the push-up model example to something which looks more natural, and reduces the max muscle activity.

AMMR 2.2.1 (2019-05-13)


  • Fixed a symmetry issue with Scapula scaling in the _SCALING_XYZ_ scaling law.

  • Fixed a bug in the Knee Simulator example regarding an extra tibiofemoral AnyForceSurfaceContact.

  • Fixed a bug in how the Conoid driver scales with scapula size.

  • Fix bug in reference axes for SacrumPelvis Moment measures in selected output. Thanks to Divyaksh Chander from Politecnico di Torino for fixing this.

  • Attempt to fix muscle wrapping errors with Latissimus dorsi when the model bends forward. This was done by increasing the Augmented Lagrangian coefficient SPLine.AugLagCoef value and the using full wrapping solver.

  • Fix some muscle wrapping problems with subscapularis muscle by increasing the number of points in the string mesh.

  • Fixed minor symmetry issue in the head markers of the Full Body MoCap example.


  • Changed the recruitment criterion to the power of 2 (quadratic) in the two MoCap examples using GRF prediction and weak human-ground residuals. (link 1 and link 2). This greatly increases the robustness of recruitment solver and prevents many muscle recruitment failures.


  • New options to optimize pelvis-, trunk- and head- widths and depths in the OptimizeAnthropometricsOnOff class template.

  • Simple library of Python utility functions which can be used from AnyScript. This first version only adds a few functions from the Python’s os.path module. They are used by including the Python.Utils.any file:

    #include "<ANYBODY_PATH_MODELUTILS>/Python/Utils.any"
    AnySwitchVar file_exists = PyUtils.os.path.exist("some-file")

AMMR 2.2 (2019-04-03)


  • New implementation of the Deltoid muscles that uses wrapping instead of previous rake-segment/via-point approach. The new wrapping deltoid was developed by Marta Strzelczak from Ecole de technologie superieure, Montreal, CA. The new implementation is used by default but the old implementation can be enabled with the BM_ARM_DELTOID_WRAPPING switch. Please see the Shoulder-Arm Documentation for more information.

  • Added new example model of a Knee Simulator using a knee implant model and force-dependent kinematics (FDK). See the Knee Simulator example for more information.

  • New model plugin (BodyModel Configurator) which provides a graphical user interface to configure the Body Model (i.e. setting the BM_* parameters). Please see the blog post introducing the BM plugin.

  • Added a new Posture prediction model based on standing posture. The model can predict posture as a consequence of applied loads in hands. It does this by minimizing joint torques and applying balance drivers which account for external applied loads.

  • Added new standing model example which uses soft constraints and GRF prediction. See this link. This new model is good starting point for making standing/balancing models, and the corresponding standing model in the template generator has also been updated.

  • Added new kinematic rhythm to determine the Sterno-Clavicular axial rotation, effectively removing the degree of freedom from the shoulder model. The rhythm distributes the axial rotation equally between SC and AC joint. The rhythm is active by default, but can be disabled with BM_ARM_CLAVICULA_ROTATION_RHYTHM switch.

  • More documentation on the AnyMoCap framework and the settings available .

  • New measures for shoulder angles based on recommendation from International Society of Biomechanics (ISB). The measures are available in the BodyModel.Interface folder. The new measures are:

    • AcromioClavicular (Wu et al. 2005 section 2.4.3)

      • Protraction

      • MedialRotation

      • PosteriorTilt

    • ScapulaHumerus (Wu et al. 2005 section 2.4.4)

      • PlaneOfElevation

      • Elevation

      • InternalAxialRotation

    • ThoraxScapula (Wu et al. 2005 section 2.4.6)

      • Protraction

      • MedialRotation

      • PosteriorTilt

    • ThoraxHumerus (Wu et al. 2005 section 2.4.7)

      • PlaneOfElevation

      • Elevation

      • InternalAxialRotation

  • Three new BM switches to control add default drivers for the individual neck degrees of freedom:

  • A default center of mass kinematic measure were added

    • Main.HumanModel.BodyModel.Interface.CenterOfMass

  • Markers in MoCap models now changes color to gray when they drop out (i.e. have a weight of zero).

  • Added new ASIS/PSIS/PT bony-landmark nodes to the PelvisSeg.Right/Left folders. This ensures the bony landmarks are easily available as nodes in all models even if the legs are not included.

  • Added a new setting (#define MOCAP_FILTER_JOINT_ANGLES ON/OFF) to the AnyMoCap frame work. If enabled a filter will be applied to the intermediate set of joint angles generated by the Marker Tracking step. This can be useful to remove high accelerations due to kinematic limit constraints (e.g. Joint angle limits, surface contacts etc.).

  • Added a #define MOCAP_PARAMETER_FILE_PREFIX switch to override the prefix for parameter identification files (containing marker position and scaling). This can be useful in special cases where subjects share a common parameter identification file.

  • Added two new default arguments (MANNEQUIN_FOLDER, BODYMODEL_FOLDER) to the class_template Template_MannequinDrivers, making it easier to reuse the class template as a general-purpose driver for all joint angles.

  • Added a PreAnalysis step to the video recorder template (CameraClassTemplate.any). Operations placed in the PreAnalysis step will be executed before the video recorder starts.

  • Added a new warning if using low-pass filters in the AnyMoCap model without a zero phase filter (e.g. warn if FilterForwardBackwardOnOff is set to Off)

  • Added a new option for choosing a different implementation of the kinematic sliding between scapula and thorax. The new scapula sliding implementation uses multiple points on the ribs and a norm measures to surfaces on the scapula. This work is preparation of a new thoracic model where ribs are individual segments. Note that this implementation is still a work-in-progress. The new thorax-scapula contact can be enabled with the BM_ARM_THORAX_SCAPULA_CONTACT switch:

  • New DEFAULT_PARAMETER_FOLDER class template for creating folders with default values which can be overridden. See the following example:

        NPARAM = 2,
        PARAM_1 = TibialRotation, PARAM_1_TYPE = AnyMat33,
        PARAM_2 = Offset, PARAM_2_TYPE = AnyVar
    ) = {
        // Default values are specified in the defaults folder.
        Default.TibialRotation = RotMat(5*pi/180,y);
        Default.Offset = 0.005;
    // Parameters can now be overridden if desired.
    Settings.TibialRotation = RotMat(10*pi/180,y);


  • Re-implemented the origins for the Semitendinosus and Biceps Caput Longum muscles in the TLEM-2.1 model based on the original digitized insertions and MRI scans. Better correspondence with the original MRI scans increases the moment-arms through range of motion.

  • New wrapping surfaces for Semitendinosus and Biceps Caput Longum muscles based on original MRI scans. New surfaces ensure that muscles don’t penetrate the pelvis bone at high hip flexion angles.

  • Reimplement insertion of Gluteus Medius/Minimus of the TLEM2.1 model based on original TLEM2 dataset. This improves range of motion in which Gluteus Medius and Gluteus Minimus has a moment arm for internal rotation of the hip.

  • Kinematic rhythms for the shoulder, lumber-spine and cervical-spine can now be specified as soft contraints. The setting is controlled by setting the BM control statement BM_XXX RHYTHM to _RHYTHM_SOFT_. For example:

  • Re-implemented the origins for the Semitendinosus and Biceps Caput Longum muscles in the TLEM-2.1 model based on the original digitized insertions and MRI scans. Better correspondence with the original MRI scans increases the moment-arms through range of motion.

  • New wrapping surfaces for Semitendinosus and Biceps Caput Longum muscles based on original MRI scans. New surfaces ensure that muscles don’t penetrate the pelvis bone at high hip flexion angles.

  • Reimplement insertion of Gluteus Medius/Minimus of the TLEM2.1 model based on original TLEM2 dataset. This improves range of motion in which Gluteus Medius and Gluteus Minimus has a moment arm for internal rotation of the hip.

  • The TLEM2 model is now the default leg model unless the model is configured otherwise.

  • All AnyMoCap examples where changed to use the _SCALING_XYZ_ scaling law. This scaling law is similar to _SCALING_LENGTHMASSFAT_, but allows for optimization of all segment dimensions (hence the XYZ name).

  • Scaling law _SCALING_XYZ_ now has a scaling factors to control the neck scaling. Before neck scaling were controlled by the head scale factors.

  • All MoCap model examples now use the option (UseC3DWeightResiduals=ON) in the marker protocols. This ensures that the marker weights are reduced to zero when a marker drops out (i.e. when the residual value in the C3D file becomes negative).

  • AnyMoCap examples are updated with extra margin at the start/end of the analysis compared to the data in the C3D file. This prevents inaccurate kinematics due to low-pass filter transients.

  • Add warning to AnyMoCap models when tStart/tEnd are very close to the start/end times of the data in the C3D file. This can cause inaccurate kinematics at the very start/end of the simulation due to low-pass filter transients.

  • Increased the resolution tibia bones for the TLEM1 and TLEM2 models.

  • All examples now have their body model configuration in a separate file (usually, BodyModelConfiguration.any) to accommodate the new Body Model Configurator plugin.

  • The MoCap models now use joint angles for all degrees of freedom when transferring motion from MarkerTracking study to the Inverse dynamic study. This means that rhythms can now be soft constraints when running the Marker tracking study, and markers can override the kinematic rhythms if needed.

  • Refactored all examples to avoid the confusing paradigm with a reference called Model.HumanModel pointing to HumanModel.BodyModel.

  • Improvements to the accuracy of the wrapping cylinder used by Teres Minor and Teres Major.

  • The interface morphing algorithm (which morphs pelvis of the trunk or leg dataset to match) now includes the ASIS/PSIS/PT bony landmarks as control points. This change ensures that bony-landmarks of the Trunk and Leg datasets map to the exact same point.

  • Models generated from templates now fail with a descriptive error message if they are loaded with an older unsupported version of the AMMR.


  • Fixed problem with the initial guess for Vastus wrapping which could cause the Vastus muscle to wrap incorrectly when the knee started in slightly over stretched posture.

  • Fixed location of bicep longus via points to pass through bicipital groove

  • Fixed a regression in bike model where the foot crank connection did not rotate around the correct point (PedalConnectNode). The point in the foot coordinate system was also updated to a more posteriorly location beneath the meta tarsal joints.

  • Fixed an issue with over-constraint default mannequin drivers when shoulder rhythm was enabled. Now the default mannequin drivers will not create drivers for sterno-clavicular rotation, elevation, protraction when the shoulder rhythm is enabled.

  • Fixed the initial wrapping guess for Teres Minor to prevent incorrect wrapping when starting model in postures which are close to range of motion.

  • Fixed typo in Flexor_Digitorum_Profundus_Digit5 name.

  • Fixed an interpolation issue with BVH based AnyMoCap models when the models were using all frames in a BVH file.

  • Added missing hand length/breadth values needed by Detailed Hand to the CustomScaling.any for custom scaling scenarios

  • Fixed missing possibility for overriding the reaction forces for the Trunk flexion/extension/rotation drivers in HumanModel.DefaultMannequinDrivers.

  • Add missing strength scaling factor to pectoralis_major_thoracic_part_3 in the simple muscle configuration.

  • Fixed issue with the Standing Model example, where the elbow flexion velocity was incorrectly set to a non zero value.

  • Fixed the problem with over-constraint models when adding the shoulder rhythm.

  • Fixed a bug with the Video class (CameraClassTemplate.any) when specifying direct path to ffmpeg.

  • Fixed a problem with the FullBody AnyMoCap example where a relative high weight on the soft clavicula rotation constraints could cause the model to fail kinematically in with the AnyBody Modeling System 7.2.


  • Removed previously deprecated MoCap model which were not based on the AnyMoCap framework.

  • Removed all references to the previously deprecated BM_TRUNK_NECK switch.

  • Removed outdated versions of JointLimits_template.any and KinLimitsDriver_template.any from the AnyMoCap folder. Updated version of the files still exists in the Tools\ModelUtilities\KinematicLimits folder.

  • All uses of the Sterno-Clavicular axial rotation in all examples, since this DOF is now handled by a kinematic rhythm which distributes the rotation equally between SC and AC joint.

  • The BM_MANNEQUIN_DRIVER_INDIVIDUAL_WEIGHTS switch was removed, since the individual weights can be set directly using

    Main.HumanModel.DefaultMannequinDrivers.KneeDriverRight.WeakDriverWeight = 0.1;

AMMR 2.1.1 (2018-06-12)

The AMMR 2.1.1 version is a minor release of the AMMR with smaller changes and bugfixes. The AnyBody Managed Model Repository now has a DOI (10.5281/zenodo.1250764). This is handled by (The European Open Science platform hosted at CERN).


  • Added a Leg.Seg.Foot.GroundJoint RefNode to the foot segment of the TLEM2 model. This RefNode was also present in the TLEM1 model and was used by few applications.

  • New GroundVelocity setting added to the ForcePlateAutoDetection class_template, which makes class usable with instrumented treadmills.

  • All force plate types now have a user-definable ForcePlate.CalMatrix variable. The feature is useful to easily compensate for errors in the c3d file (i.e. swapped channel and wrong sign)


  • Enhanced the multi-trial MoCap example for best practice for MoCap trials with multiple subjects and trials. The example now shows how to deal with multiple subjects and storing the c3d files in a separate folder.


  • Corrected the default mass of the patella segment in the TLEM2 model from zero to 0.025.

  • Add missing unilateral reaction between TS node on Scapula and the Thorax when using the shoulder rhythm.

  • Adjusted the initial wrapping vectors for the semitendinosus, semimembranosus, and biceps caput breve muscles around the knee. This has no influence on the output of the model, and is only done to reduce the risk of the muscles wrapping the wrong direction around the wrapping surface.

  • AnyMoCap:

    • The high memory usage of the AnyMoCap BVH model has been fixed. The high memory usage was due to a reference in the CreateMarkerDriver class template which caused the studies to include the BVH input data multiple times in the output.

    • The BVH and multi-trial examples now use zero phase shift filtering of force plate forces. The zero-phase filter can be important to ensure that forces are synchronized with kinematic data.

    • Fix bug in Type 1 Force plate when data is not in meters (i.e. PointScaleFactor != 1). The PointScaleFactor was applied twice to the z moment component.

    • Fix a problem reading C3D files where force plate meta information is saved with the wrong dimensionality. Such a issue can, for example, happen if the c3d files are created manually in MATLAB.

    • The option to load parameter optimization results is no longer hidden when the MOCAP_CREATE_PARAMETER_ID_SHORTCUT is set to OFF.

AMMR 2.1.0 (2018-03-22)


  • New squat example model which demonstrates a parameterized squatting motion.

  • New #class_template for adding limit drivers to kinematic measures. Can be included with #include "<ANYBODY_PATH_MODELUTILS>/KinematicLimits/KinLimit_template.any"

  • Muscle ColorScale can now be set from the Main.DrawSettings.Muscles.ColorScale. If you port an older model to the new AMMR you will need to update your DrawSettings.any file.

  • AnyMoCap framework:

    • New argument DRAW_SCALE in the CreateMarkerDriver class_template to control the visual size of markers.

    • Added an option to the ForcePlateAutoDetection class template to make the limb1/2 contact detection mutually exclusive. Setting the option ALLOW_MULTI_LIMB_CONTACT=OFF will ensure that both legs cant be in contact with the plate at the same time. This can prevent accidental contact detection for the collateral leg in the swing phase.

    • New option to specify the contact nodes in ForcePlateAutoDetection class template. This makes the class useful for connecting force plate to other limbs than the feet.

    • Add support for the Shoulder Rhythm in AnyMoCap based models.

    • Add support for the Detailed hand model in AnyMoCap based models.

  • New BM_FOOT_MODEL parameter, as preparation for integrating the Glasgow-Maastricht foot model (GM-Foot) back into the AMMR. Currently, the parameter can only be used for excluding the feet (#define BM_FOOT_MODEL _FOOT_MODEL_NONE_). making it easier to work on integrating the GM foot model.

See also

Adding the GM-Foot to the TLEM2 model is a work in progress. An early version is released on GitHub:


  • Default mannequin drivers for the Pelvis are changed to drive the Pelvis anatomical frame instead of the segmental reference frame. This also fixes the discrepancy between the load time position and the mannequin driver position and makes the driver consistent with the interface measures BodyModel.Interface.Trunk.PelvisPosX/Y/Z.


This change will affect models using the default mannequin drivers unless the driver values are updated.

  • The Wilke spine pressure validation examples have been updated and now uses the TLEM2 lower extremity model.

  • In _SCALING_XYZ_ scaling law the definition of length/depth/width of the scapula is corrected to match the anatomical definitions.

  • Updated the AAU Mandible Model introduced in AMMR 2.0.0. By accident the authors did not share the exact same version of the model that was used in the publication by Andersen et al. 2017 This is now corrected and the validation example produces the same results as published version.

  • TLEM 2 model

    • Pectineus implementation has been adjusted based on the original MRI Scans. Only the most lateral element of the pectineus muscle was changed. This was done to prevent it from penetrating the femoral head in some situations.

    • Tweaked the Gluteus Maximum Superior implementation to reduce the risk of collisions between muscle insertions and the wrapping surface. The two most proximal insertion points of the Gluteus Superior are shifted more distally, and the locations of the wrapping surfaces are tweaked.

  • AnyMoCap Framework:

    • Models now uses the default kinematic solver for the inverse dynamic analysis.

    • Changed the set joint drivers used for transferring the joint angles from the over-determinate marker tracking to the inverse dynamic analysis. To improve the stability of the scapula kinematics the Scapula Thorax Elevation angle is used instead of the sterno clavicula elevation angle.

    • Changed the way marker drivers are excluded from the Inverse Dynamic analysis. The drivers are now excluded using the new AnyMechObjectExcluder class instead of the MechObjectExlcude member of the Study. This has no practical effect on model output but means that the MechObjectExlcude can now be assigned by the users.

    • Update to the BVH MoCap example. Pre-processing the BVH data is now a separate operation which saves the virtual marker positions to a file. Thus, this step can be skipped the next time the model is reloaded.

    • Changed the AnyMoCap Multi trial example Plug-in-gait_MultiTrial_StandingRef to make it a better starting point for creating new models. 1. Moved the Human-Ground residuals from the Trunk to the Pelvis segment. 2. Use the TLEM 2 lower body model in the example. 3. Disable upper bounds for the muscle recruitment Criterion.UpperBoundOnOff = Off; to improve the stability of the simulations.


  • Shoulder-arm model:

    • Fixed problems with Pectoralis Major wrapping. Wrapping could fail unpredictively when muscles wrapped in the intersection between the coracoid wrapping and pectoralis minor cylinder. The coracoid wrapping cylinder has been removed, and the others has been made longer. This creates a more robust wrapping for the Pectoralis Major muscle.

    • Better initial position for scapula and clavicula. The initial position is now calculated from the initial position of the chain from thorax through the clavicula to scapula. This will not change the model output but should make the arm model more robust solving the first step.

    • Improve the Teres Minor wrapping by adjusting the wrapping surface, and ensure that the insertion point is not directly on the surface.

    • The “Evaluate joint strength” studies now also work when the shoulder rhythm is enabled.

    • Fixed problems with via points of the wrist extensor muscles colliding with the wrist wrapping surface in normal range of motion. The via point of Extensor indicis Via_Extensor_Indicis_pos has been slightly adjusted and the location of the wrist wrapping cylinder MedialExtensorCyl has been improved.

  • Detailed hand model:

    • Fix a problem causing the hand to scale incorrectly.

    • Fix a problem with the default mannequin drivers for the left thumb, where CMC and MCP abduction was treated as adduction.

    • Fixed compatibility between the detailed hand and scaling law _SCALING_XYZ_.

  • Lower extremity models:

    • Fixed various warnings when running calibration routines with the lower extremity models.

    • Fixed muscle insertions for the old leg model (#define BM_LEG_MODEL _LEG_MODEL_LEG_). The misaligned pelvis muscles insertions was a regression due to the updated Trunk pelvis introduced in AMMR 2.0. The pelvis muscles insertions have been translated and rotated to fit the new Trunk geometry as best as possible.

    • Fix missing left leg calibration operations when right leg was excluded.

  • MoCap models:

    • Weak residuals for GRF prediction: Ensure the same strength is used in all directions for the weak recruited actuators.

    • Fix bug in CreateMarkerDriver class template which prevented the UseC3DWeightResiduals from having any effect.

    • Fix the wrong visualization of the contact area in ground reaction force prediction class template, when the base frame is different from the global coordinate system.

    • Wrong masses in the deprecated “old” MOCAP models. The model was not using the body mass specified in Main.TrialSpecificData.Anthropometrics.BodyMass

  • Other fixes:

    • Wrapping convergence for multiple muscles has been improved. This was done by tweaking the string-mesh of the muscles to improve the wrapping solvers ability to find the solution within the given number of iterations.

AMMR 2.0.0 (2017-11-29)

Major changes:

New lower extremity model (TLEM2.1)

  • The Twente Lower Extremity Model version 2.0 dataset, developed in the TLEMsafe EU project was implemented in the AMMR repository. The model is not the default model, but can be enabled with the BM parameter #define BM_LEG_MODEL _LEG_MODEL_TLEM2_

  • The model is versioned TLEM 2.1, to indicate the number of changes and correction which has been added in the process. The changes and updates to the TLEM2 dataset was done in the Life Long Joints EU research project (paper submitted for publication).

  • The most important changes to the TLEM 2 dataset include the following:

    • Updated wrapping for the Gluteus Maximus, Iliacus, Psoas around the hip.

    • Reworked muscle topology for Gluteus Medius and Gluteus Minimus

    • Updated wrapping for Hamstring muscles, and Gastrocnemius around the knee.

    • Redefined revolute knee axis and patella joint axis and patella tendon length based on the bone geometry. The original TLEM knee axis was estimated using a functional method, and was only valid for very small flexion angles. Thanks to Marco Antonio Marra (Radboudumc) for this for this improvement.

    • The ratio of volume between Gluteus Maximus superior/inferior has been re-estimated based on the original cadaver MRI scans.

    • Update the Sartorius via points.

    • Re-implemented the Hip Joint location for the pelvis and the femur by fitting spheres to the femoral head and the acetabulum. This is in contrast to the original TLEM2.0 implementation where the hip joint center was found using a functional method.

    • Included a more detailed version of the patella bone.

    • Various minor bug fixes from the original implementation used in the TLEMSafe project.

    • Corrected the femoral attachment points for the popliteus muscle

    • Update ankle joint nodes with positions fitted to the bone geometry

    • Small correction to wrong insertion points for Adductor Longus and Vastus Medialis.

  • See the dedicated TLEM2.1 page page for more information.


  • New framework for running MoCap models. The AnyMoCap framework is an effort to create a simple and unified framework for doing any kind of MoCap analysis with the AnyBody Modeling System. See the AnyMoCap example gallery for more information.

  • Algorithms and file for Ground reaction force prediction are added to: ammr/tools/GRFPrediction/. See the MoCap examples for how they are used.

TLEM 1 updates

  • The cumulated smaller bug fixes and updates to the TLEM1 model mean that we now denote the model 1.2.

  • Changed the position of the Heel contact node, to align the heights of the TLEM1 and TLEM2 models.

  • Added missing GeomScale function to the patella segment.

  • Update ankle joint nodes with positions fitted to the bone geometry

  • PSCA is now calculated based on the scaled fiber length. This aligns the TLEM 1 model with the new TLEM 2 model. It also means that scaling the model without changing the muscle volumes will change the strength of the model.

  • Update scaling of the Patella tendon to work with non-linear scaling laws.

  • Added visualization of the Fibula bone.

AMMR restructure

  • Added a top-level Tools folder helper files and other models which don’t belong in Body/ or Application.

  • Moved Body/AAUHuman/Toolbox to Tools/ModelUtilities

  • Renamed the Body/AUHuman to Body/Mandible_AU

  • Restructured the anthropometric (“AnyMan”) folder. The folder HumanModel.AnthroDataSubject is renamed to HumanModel.Anthropometics. The structure of the folder has also changed with a SegmentDimensions and SegmentMasses subfolder.

Configuration parameters

  • The previously used system to configure human body using a BodyPartSetup file was completely replaced with the new body model (BM) parameters.

  • Please see the documentation on: BM parameters.

Scaling laws

  • An additional scaling law based on individual segmental scaling factors was added to the repository. It can be enabled using this BM parameter #define BM_SCALING _SCALING_XYZ_. See the guide on scaling.

  • Default scaling is now _SCALING_STANDARD_ (e.g #define BM_SCALING _SCALING_STANDARD_) which scales all models to default standard 50% male.

  • Introduced _SCALING_NONE_, which disables scaling. E.g. models gets the original unscaled size.

New Mandible model

Spine model

  • BM parameters were updated to have control over each section of the spine and relevant components.

  • The anatomical reference frame of the thorax segment was modified. This change reflects a change in the pelvic anatomical reference, and ensures upright posture for the standing postures, to align C1C0 joint with the hip joint centers.

  • Boney surfaces of both pelvis and sacrum were updated and now correspond better to the relevant muscle attachments. These segments now also share a common scaling function. Hip joint centers were corrected for the old Leg model.

  • Improved wrapping surface for Psoas Major muscles based on the TLEM2.0 MR scans

  • Insertion, via, and attachment nodes of relevant muscles have been updated to match new geometries of pelvis and sacrum.

Arm model

  • The model now facilitates individual personalization for each side using nonlinear morphing schemes in a more consistent manner. Previously the morphing needed to be done on the right side and then reflect to have the left side morphing. This change removes an extra mirroring step.

  • Arm calibration was updated

  • BM parameters have been updated for more convenient use. BM_ARM_DETAILED_HAND and BM_ARM_SHOULDER_RHYTHM are now used instead of individual switches for right and left side, which were deprecated.

  • Muscle wrapping surfaces were updated for more physiological behavior.

  • Scapula reaction contact forces were simplified, and do not longer utilize slider segments.

  • Conoid ligament length now scales along with the scapula width.

  • Add a GeomScale function the Clavicula segment.

Muscle models

  • All muscle models are updated to support the structure of the new AnyMuscleModel3E and AnyMuscleModel in the AnyBody Modeling System 7.1. The following variables are renamed:

    • The “optimal fiber length” variable renamed from Lfbar to Lf0

    • The “Pennation angle” variable renamed from Gammabar to Gamma0

    • The “Tendon strain at F0” variable renamed from Epsilonbar to Epsilon0

  • Restructured the muscle model section of both TLEM1 and TLEM2 models.

    • All the original TLEM based muscle parameter are now located under: Leg/ModelParameters/Muscles

    • All scaled muscle parameters are located in Leg/MusPar/SubjectMusPar. This folder references the TLEM muscle and applies strength scaling etc. The SubjectMusPar folder and all subfolders are implemented with class_template. Thus, all muscle parameter can now be overridden in applications by just assigning the variables a new value: E.g.

      Main.HumanModel.BodyModel.Right.Leg.MusPar.SubjectMusPar = {
        GastrocnemiusMedialis.MuscleVolume = 300; // Volume in mililiters
        GastrocnemiusMedialis.Pennationangle = 15; // (in degrees)


  • Updated calibration for Arms and TLEM legs in the Body Model to

    • include muscles to the calibration study with search functions.

    • drive the postures using the measures from the interface folder to remain anatomically similar throughout future versions.

  • Added new experimental two-parameter calibration, which is based on range-of-motion postures. The calibration type is controlled by the BM_CALIBRATION_TYPE parameter.

    For example:


Minor Changes:

  • Added new initial guess for wrapping muscles, which make the wrapping more when the model starts in extreme postures.

  • Update many examples to use the TLEM 2.1 model. See the example gallery.

  • BM mannequin drivers are now implemented with a class_template allowing all weights and other settings to be customized.

  • A default HumanModel.Mannequin folder is now automatically created with a class_template when no user-defined Mannequin file is set.

  • A default Main.DrawSettings folder is now automatically created with a class_template when no user-defined DrawSettings file is set.

  • Extra Mannequin drivers for the individual shoulder degrees of freedom: Sterno clavicula protraction, Sterno clavicula elevation, Sterno clavicula axial rotation

  • The initial positions of the pelvis now use the anatomical reference frame. This follows the logic from the initial rotation of the pelvis which also uses the anatomical frame.

  • DeltoidMuscleConnector segment loading time positioning now depends on the Humerus segment.

  • Added class template to easily create videos from AnyScript model. The tool requires that FFmpeg is installed. The class template can be found in: <ANYBODY_PATH_MODELUTILS>/Video/CameraClassTemplate.any. See this blog post.

  • In TLEM models make the opacity of the patellar tendon dependent on the opacity of the patellar surface.

  • New AnyDoc classes are added to the different body model, so the GUI can create direct links to the documentation.

  • Simplify the Scapular reactions to the thorax segment.

  • Foot contact nodes are aligned with the AnatomicalFrame

  • Updated the Wilke Validation example to reflect the forces for the AMMR 2.0 repository.

  • Updates to BM parameters:

    • New BM_ARM_DETAILED_HAND parameter for the detailed hand. The old BM_ARM_DETAIL_HAND_RIGHT/LEFT are deprecated.

    • New BM_ARM_SHOULDER_RHYTHM parameter for controlling the shoulder rhythm. The old BM_ARM_SHOULDER_RHYTHM_RIGHT/LEFT are deprecated.

    • Added new BM_JOINT_TYPE_<joint>_<side> parameter for completely disabling joint and associated nodes in the lower extremity models. (See: for example BM_JOINT_TYPE_HIP_RIGHT)

    • New BM_LEG_MODEL parameter for setting the type of leg model used. The BM_LEG_RIGHT/LEFT are now only ON/OFF options.


  • Sign for the plantar flexion variable were reversed in some section of the model. This has been fixed.

  • Bug in Mannequin drivers for the neck, where velocities were not set correctly. (Thanks to Assoc. Prof. Michael Skipper Andersen for reporting this)

  • Fix small bug preventing StandingModelScalingDisplay from loading when using the Leg model.

  • Fixed the opacity of the patellar surface in TLEM models, which pointing erroneously to the opacity of the talus.

  • Fixed wrong symmetry of nodes on the C7 segment of full neck model.

  • Latissimus Dorsi 5 fascicle was missing in MuscleNames.any and thus from many symmetry measures.

  • Fixed a symmetry problem for the Deltoid muscles at the shoulder.

  • Fixed a symmetry problem for the Disc stiffness from L1 to L5

  • Fix white surfaces in examples with flat STL surfaces. For example Cross Trainer.

  • Fixed an issue preventing Standing Model from working with one leg.

  • Fixed a problem with the drawings of the bones in the Arm model which were not always symmetrical.

  • Fixed symmetry issues in scaling laws for scapula and clavicula, and humerus.

  • Fixed a bug where a the Pectoralis wrapping cylinder was not a included in the calibration study.

  • Fixed wrong sign for the AnklePlantarFlexion variable.

  • Added missing GeomScale and AnatomicalFrame for Ulna segment.

  • Fixed ScalingCustom.any to use Thorax folder instead of Trunk.


  • Old MoCap examples have been moved to Application/Examples/Deprecated

  • Removed the deprecated AMMR1.4 hip rotation sequences.

  • The GM-foot model. A new version of this in the pipeline. Contact us if you are interested in this work.

  • All older BodyModels which were deprecated in AMMR1.3