Modelling the human-exoskeleton interaction interface - Benchmarking MuJoCo's deformable model
Authors: Lisette Masselink, Xiaolong Zhu Thomas George Thuruthel, Mohamed Irfan Refai
Organizations: Chair of Neuromuscular Robotics, Department of Biomechanical Engineering, University of Twente, Enschede, The Netherlands · School of Medical Technology, Beijing Institute of Technology, Beijing, China · Department of Computer Science, University College London, London, U.K.
The human-exoskeleton interaction interface, consisting of the exoskeleton attachment and the underlying human soft tissue, can affect the transfer of assistance. MuJoCo's flexcomp has been previously used to model the soft tissue, but its accuracy has not been benchmarked against a finite element model. We benchmarked the flexcomp Saint Venant-Kirchhoff (StVK) and Stable Neo-Hookean (SNH) models against COMSOL's Yeoh model in three cases: case 1) spherical indentation, case 2) compression, and case 3) compression of a thigh. We compared force and strain energy as a function of indentation depth, hysteresis and computation time. MuJoCo's models capture force trends with a correlation of 0.97-0.99, but showed a stiffer response, which could be attributed to the volumetric locking of the linear finite elements. The MuJoCo models systematically underestimated hysteresis as MuJoCo's Rayleigh damping could not substitute a viscoelastic model. Moreover, MuJoCo's computational advantage was speed dependent: it was 2-2.5 times faster for spherical indentation, but COMSOL was faster for the other cases, which have more degrees of freedom and contacts. Although MuJoCo's flexcomp offers a combination of soft-tissue modelling and multibody dynamics, the limitations regarding the stiffening and hysteresis should be addressed to be used for the analysis of the human-exoskeleton interaction interface.
Figures & tables
Fig. 1 : Overview of the three cases and the methods to compare them. a) Indentation of a silicon block using a spherical indentor, b) compression of the block using a rigid plate, and c) compression of a thigh model with a rigid plate. Arrows indicate the direction of predescribed movement. Simulations are perfomed using MuJoCo’s St. Venant-Kirchhoff model and Stable Neo-Hookean model, and COMSOL’s Yeoh model.
Tool
Contact-Rich
MSK
Validated
Computational
MBD
Modelling
Soft Tissue
Speed
COMSOL [ 23 ]
✗
✗
✓
✗
MuJoCo [ 6 ]
✓
✓
✗
✓
TABLE I : Comparison of Simulation Tools for Modelling the Human–Exoskeleton Interaction Interface
Parameter
Symbol
Value (Unit)
MuJoCo’s settings
Flexcomp parameters
Young’s modulus
E
40kPa
Poisson’s ratio
ν
0.485
Rayleigh damping coefficient
β
0.002s
Solver model
TABLE II : Material Properties and Simulation Parameters used in MuJoCo (top) and COMSOL (bottom)
Fig. 2 : Snapshots of the MuJoCo (top) and COMSOL (bottom) simulations at initial contact and maximum indentation. Case 1 (left): a spherical indentor indents the soft tissue, which rests on a fixed plate. Case 2 (middle): a plate compresses the soft tissue from above. Case 3 (right): a plate indents a deformable thigh (outer cylinder), which is fixed to the femur (inner cylinder). Top: arrows indicate the magnitude of contact forces, and discs indicate the contact points. Bottom: colours indicate the von Mises stress.
Fig. 3 : Force-indentation (right) and strain energy (bottom) as a function of indentation depth for case 1 (left), case 2 (middle) and case 3 (right). The blue line indicates an indentation speed of 1 mm/s, the orange and green lines indentation speeds of 2 and 5 mm/s, respectively. The solid line is the COMSOL simulation, the dashed line the MuJoCo StVk model and the dotted line the MuJoCo SNH model.
Setup
η (%)
MuJoCo (StVK)
MuJoCo (SNH)
COMSOL
Case 1: 1 mm/s
1.81
0.30
7.67
Case 1: 2 mm/s
2.32
0.61
11.02
Case 1: 5 mm/s
3.73
1.49
15.90
Case 2: 1 mm/s
1.54
0.88
12.09
Case 2: 2 mm/s
2.48
1.93
17.65
TABLE III : Hysteresis ratio relative to the loading work for different setups, comparing MuJoCo and COMSOL simulations under nominal conditions.
Setup
twall (s) (RTF)
MuJoCo (StVK)
MuJoCo (SNH)
COMSOL
Case 1
426 (22.4 × )
536 (28.2 × )
1131 (55.5 × )
Case 2
473 (43.0 × )
483 (43.8 × )
172 (14.3 × )
Case 3
5336 (266.0 × )
5194 (258.9 × )
327 (16.4 × )
Fransen et al. [ 15 ]
156 (39.0 × )
–
–
TABLE IV : Wall-clock time and real-time factor (RTF) for simulations at an indentation speed of 1 mm/s.
Accurate soft tissue simulation is essential for surgical training, pre-operative planning, and haptic feedback systems. While learning-based surrogate models trained on data using the finite element method (FEM) offer a promising path to real-time inference, their reliability depends on well-calibrated constitutive models. Existing approaches neither provide systematic guidance on model selection across stiffness levels, nor generalize across different tissue stiffnesses or geometries. We perform a comprehensive calibration of hyperelastic constitutive models in the SOFA Framework using gravity-loaded silicone beams with different stiffnesses. Using calibrated simulations as training data, we use a softness conditioned equivariant graph neural network, enabling deformation and force prediction across multiple tissue types and unseen geometries. Our model achieves sub-millimeter mean deformation accuracy at 0.010s inference time, while showing that force prediction quality is directly tied to upstream calibration consistency.
Madina Kojanazarova, Sidaty El Hadramy, Philippe C. Cattin
University of Basel, Department of Biomedical Engineering, Allschwil, Switzerland
Soft wearable robots must commit to assistance around movement onset, when the available signals are ambiguous. Surface electromyography (sEMG) detects neuromuscular drive early but degrades under sweat and electrode shift, and kinematic sensors respond only once the limb moves. This letter presents a knitted textile interface that co-samples a three-channel mechanomyography (MMG) array and an anterior strain sensor on a single synchronous 500 Hz clock. Across five participants, MMG activation preceded strain-detected knee motion during sit-to-stand by per-channel medians of 322-344 ms. Sampled on the same clock, the sign of the strain slope resolved movement direction on 649 of 650 completed transitions with no training or per-participant calibration, raising four-class balanced accuracy from 90.3% to 95.6%. The fused signals drove a two-stage confirm-then-assist policy for a soft pneumatic knee. Across 240 trials in four participants, the onboard controller entered the assist state on 78 of 80 sit-to-stand trials and on none of the 80 isometric co-contractions. It also entered the assist state on 27 of 80 stand-to-sit trials and left it at the hold timeout, motivating an antagonistic flexor gauge to resolve this direction at onset. Co-sampling muscle acoustics with textile strain on one clock lets the inflation command be issued around movement onset.
Zheng Ma, Xiao Jin, Zhenhua Yu
Department of Computing Science, University of Aberdeen, Aberdeen AB24 3UE, U.K. · Dyson School of Design Engineering, Imperial College London, London, U.K.
Physically connected humanoids need consistent dynamics for internal actuator loops, inter-robot forces, and changing ground supports. We extend RoboCompiler with member-wise dynamics assembly, compiling reduced dynamics from rigid bodies, moving actuator parts, and reflected rotor inertias. Compiled body Jacobians and analytic constraint curvature yield an explicit Coriolis factor satisfying the power identity for floating-base quasi-velocities. A two-level reduction retains reusable robot models while composing their supports and hand connections; local inertial updates have rank at most six, and prescribed-acceleration constraint wrenches are affine in actuator efforts. These interfaces support actuator-aware tracking, coupled force allocation, payload updates, and support-mode switching for five hand-linked humanoids. Across 264 feasible states in eight humanoid configurations, normalised acceleration discrepancies remain within 1.6×10−10 against a hard-constraint reference and 1.4×10−10 against a compiler-independent reference; MuJoCo simulations achieve 0.73~mm root-mean-square (RMS) pelvis error in the ring dance and complete side steps through 20 support-mode changes.
Mehdi Heydari Shahna, Jouni Mattila
Faculty of Engineering and Natural Sciences, Tampere University, Tampere, Finland