Pith. sign in

REVIEW 3 major objections 4 minor 56 references

A Physics-Based Continuum Model for Versatile, Scalable, and Fast Terramechanics Simulation

T0 review · 3 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read A physics-based SPH model of granular terrain is validated against rover and digging tests and shown to run near real time at scales up to 28 km.

desk verdict A serious engineering contribution that mostly delivers on its claims, but the active-domain scalability story needs a sensitivity study and the cone-penetration parameters are likely fitted, so the headline accuracy claims need softening. read the letter →

arxiv 2507.05643 v2 pith:3SO24WIV submitted 2025-07-08 cs.RO

Huzaifa Unjhawala , Luning Bakke , Harry Zhang , Michael Taylor , Ganesh Arivoli , Radu Serban , Dan Negrut This is my paper · ORCID
classification cs.RO
keywords terramechanicsSmoothedParticleHydrodynamicsgranularcontinuummodeldeformableterrainsimulationpersistentneighborlistsactivedomainsoff-roadvehicleGPU-accelerated
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Chrono::CRM treats loose soil as a continuum fluid rather than as individual grains, so terrain that would require millions of discrete elements reduces to roughly a million smoothed-particle-hydrodynamics particles. The paper reports that this model reproduces experimental penetration depths, cone-impact depths, rover wheel slip on slopes, and the digging torque of a counter-rotating excavation drum, the last checked against a high-fidelity discrete-element benchmark. Two computational devices carry the speed claim: persistent neighbor lists that are reused for several time steps, and “active domains” that freeze and unload every soil particle outside a moving box around each body. With both enabled, the solver is said to match or beat a widely used semi-empirical terrain model in runtime while retaining a physics-based soil description, and to keep near-interactive rates on terrain strips up to about 28 km before GPU memory runs out.

What carries the argument

The central object is the continuum representation of granular media in SPH form: soil is discretized as particles carrying density, velocity, and stress, with the momentum balance discretized so that local linear momentum is conserved, and the stress is returned to a Mohr–Coulomb $\\mu(I)$ yield surface with a cohesion intercept through a radial-return map. Material phases (elastic, plastic, and cohesion-limited) emerge from where the trial stress sits relative to that envelope. The two performance mechanisms are persistent neighbor lists, rebuilt at a user-set frequency, and “active domains”: object-oriented bounding boxes attached to each body, inside which particles are active, with an extended-active buffer within $2h$ of each box, while everything farther away is frozen and freed from GPU memory.

What would settle it

Re-run the published MGRU3 or RASSOR setup with the same active box but place a rigid wall or obstacle inside the frozen zone, and compare wheel slip or drum torque against a full-domain run; any difference beyond benchmark noise shows frozen terrain is not inert. A cheaper check is to shrink the active box progressively and report the box size at which the claimed no-loss-in-accuracy property starts to fail.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central claim is that a continuum SPH representation of granular media, coupled to a multibody dynamics engine through boundary-condition-enforcing markers, can serve as a general-purpose terramechanics simulator: accurate enough to match dedicated physical tests and DEM benchmarks, versatile enough for digging, grading, and deformable wheels, and fast enough to compete with semi-empirical models in large-scale off-road simulation. The paper does not introduce a new soil law; it argues that the existing continuum approach, once reorganized around persistent neighbor lists and active domains, retains its validated accuracy while gaining roughly 1.8–3.9x speedups from neighbor-list caching, a further 2–3x from active-domain culling, and sub-linear scaling with terrain length until GPU memory is exhausted.

Load-bearing premise

The active-domains device assumes every soil disturbance stays inside a moving box around the vehicle, so all terrain outside can be safely frozen and forgotten; if a pressure wave or a pushed pile reaches beyond that box, the computed forces are no longer trustworthy.

Editorial extensions

If this is right

  • Digging, grading, and grouser or tread interactions, which semi-empirical Bekker–Wong models cannot represent, become available in a physics-based solver at near-real-time rates.
  • Simulation speed reaches or exceeds a semi-empirical soil-contact model on identical vehicle benchmarks, so physics-based terrain need not be traded away for speed.
  • Terrain strips up to about 28 km on an H100 GPU, and about 5.9 km on a consumer GPU, can be traversed before memory is exhausted, with sub-linear runtime scaling over that range.
  • The model supports rigid and flexible bodies, including ANCF deformable tires, tracked vehicles, and full rover models, all co-simulated with the same soil description.
  • Because the software is open source and the validation cases are public, the reported benchmarks can be rerun and extended by others.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The active-domain assumption that soil disturbance is local is plausible at the low speeds tested, but nothing in the paper bounds the error once a disturbance front reaches a frozen region; high-speed travel, repeated passes over the same lane, or deep loading could quietly invalidate the computed forces.
  • The 28 km scaling result is really a statement about the size of the active region rather than about total particle count, so comparisons across studies should quote the active-box dimensions and the fraction of particles actually processed.
  • A natural and testable extension is to make the active box adaptive, grown according to the measured stress or velocity front speed, which would preserve the speed gains while covering scenarios where distant terrain must respond.
  • Speed parity with semi-empirical models suggests that large-scale controller training could be done on physics-based soil interaction rather than on simplified terrain models, a direction the paper's leveling case study only begins to explore.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The paper presents Chrono::CRM, a GPU-accelerated SPH-based continuum model for simulating deformable terrain in vehicle and robot terramechanics applications. The model is built on the Chrono FSI framework and uses a mu(I) rheology with a cohesive Mohr-Coulomb yield envelope. The authors validate the solver against three physical experiments (sphere cratering, cone penetration, and NASA MGRU3 single-wheel tests) and one DEM benchmark (RASSOR digging), then report performance optimizations including persistent neighbor lists and an 'active domains' scheme that deactivates and frees terrain particles outside moving bounding boxes. They demonstrate speedups relative to a prior SPH implementation and to the semi-empirical Chrono::SCM model, and claim scalability to terrain lengths up to about 5.9 km on a consumer GPU and about 28 km on an H100. The paper also includes two demonstrations: a Polaris RZR with rigid/flexible tires on rigid/deformable terrain, and a learned controller for an autonomous Gator leveling task.

Significance. If the central claims hold, the paper is a significant contribution to practical terramechanics simulation: it combines a physics-based continuum description with the ability to run large-scale, near-real-time vehicle-terrain interaction problems, and it is released as open source with a reproducibility repository. The quantitative validation in the sphere-cratering test (slope 0.1336 vs. the empirical 0.14, R2 = 0.9714) and the cone-penetration tests (relative errors 3.5% and 7.4%) provides concrete evidence for the solver's basic fidelity. The active-domains approach is an interesting and potentially powerful idea for long-traverse simulations. However, the validation gaps in the MGRU3/RASSOR comparisons and, more importantly, the absence of any accuracy check for the active-domain method in the regime used for the headline scalability claim mean that the strongest claims in the paper are not yet supported.

major comments (3)
  1. [Sec. 4.2 and Sec. 5 (Fig. 16)] The 'no loss in accuracy' claim for active domains is validated only in two low-speed, large-box settings (Sec. 4.2.1: MGRU3 at 0.8 rad/s with a 0.6x0.6x0.8 m box in a 5x0.8x0.25 m bin; RASSOR at 0.15 m/s with a 0.5x0.4x0.7 m box), but the scalability study in Sec. 5 uses 0.4x0.4x0.4 m boxes around 0.33-m-radius RZR tires and reports only RTF versus terrain length, with no comparison to a baseline without active domains. Since particles outside the moving boxes are frozen and removed from memory, a disturbance front that reaches the box boundary cannot influence the vehicle; this is exactly the regime (small boxes, fast or large vehicles) in which the 5.9 km and 28 km scaling claims are made. The paper should either report an accuracy study for the RZR configuration (e.g., wheel forces, slip, or trajectory error versus a no-active-domain baseline) or provide a quantitative bound on the disturbance-front argument that justifies the chosen box sizes.
  2. [Sec. 3.2] The cone-penetration validation uses rheology parameters (mu_s, mu_2, I_0) that the paper states were 'determined empirically' because specific experimental material data was not available. The resulting 3.5% and 7.4% errors are therefore partially a curve-fit result, not an independent predictive validation. Given that this is one of only three physical validation tests, the paper should either obtain independent parameter measurements (e.g., from a separate experiment) or explicitly reframe the study as a calibration demonstration and temper the claim that the solver is 'validated against physical tests' in a predictive sense.
  3. [Sec. 3.3 and Sec. 3.4] The MGRU3 single-wheel and RASSOR drum comparisons are presented only visually (Figs. 5a and 7a), with qualitative statements of 'good agreement' and 'close agreement'; no numerical error metrics, such as RMS error, maximum deviation, or correlation, are reported for the slip-versus-slope curve or the torque time history. Without quantitative measures, the reader cannot assess the actual fidelity of these system-level validations, which are central to the paper's claim of experimental and DEM-verified accuracy. Please add numeric error metrics for both benchmarks.
minor comments (4)
  1. [Throughout] The abstract and several body passages have missing spaces between words (e.g., 'Thispaperdiscusses', 'baring minor differences' in Sec. 2, 'thess active boxes' in Sec. 4.2), which should be corrected.
  2. [Sec. 2.5] The BCE stress extrapolation equation appears garbled in the manuscript (missing Greek symbols and indices), making it impossible to verify the formula; please repair the typesetting.
  3. [Sec. 5] The RZR tire width is given as 0.21 m in Sec. 5 but 0.33 m in Sec. 6.1; please reconcile these values.
  4. [Sec. 6.1] The statement that 'the CRM solver completes a step in a significantly faster time than the multibody system' is supported by Table 6, but the comparison mixes CPU (multibody/FEA) and GPU (CRM) timings; a brief clarification of where the co-simulation bottleneck lies would help the reader interpret the RTF values.

Circularity Check

1 steps flagged · score 4.0 of 10

Cone-penetration validation appears to use empirically fitted rheology parameters, making that specific test a calibration check; the remaining validation and performance claims are not circular.

  1. fitted input called prediction [Sec. 3.2, Cone Penetration test (rheology parameter setting and Fig. 3 comparison)]
    "For the rheology model, we set the parameters as follows: μs = 0.70, μ2 = 0.80, and I0 = 0.08 for the glass beads; and μs = 0.80, μ2 = 1.00, and I0 = 0.08 for the Ottawa sand with cohesion (c in Eq. 15) set to 0 Pa. These parameters were determined empirically, as specific experimental material data was not available. ... We obtain mean relative errors across all drop heights H of 3.5% for glass beads and 7.4% for Ottawa sand."

    The same cone-penetration experiments provide the validation metric (penetration depth vs time). The rheology parameters are described as 'determined empirically' without citation to an independent material-characterization source, so the paper gives no external grounding for them. Reporting mean relative errors of 3.5% and 7.4% against those same experiments therefore measures how well the calibrated parameters reproduce the calibration data, not an independent predictive test. The fitted degrees of freedom (μs, μ2, I0, cohesion) are inputs to the model, and the agreement is partly enforced by the empirical determination rather than derived from first principles.

full rationale

The central solver derivation is not circular: the continuum equations, SPH discretization, and return-mapping plasticity (Eqs. 1-19) follow from standard mechanics and the externally published μ(I) rheology of Dunatunga & Kamrin [13]; no target validation quantity is an input to those equations. The sphere-cratering comparison uses an external empirical relation (Eq. 20) and external experimental data; the MGRU3 wheel test compares directly to NASA SLOPE lab measurements; the RASSOR study compares CRM against a DEM solver [56] that, while from the same group, is a methodologically independent benchmark. Those results retain independent content. The one substantive circular element is the cone-penetration validation, where the rheology parameters are 'determined empirically' with no external source, and the same experiments are then reported as agreement. This is a fitted-input-called-prediction in one of the four validation studies. Self-citations [21], [23], and [56] are present and provide context and prior validation, but they do not force the paper's central accuracy or performance claims by themselves. The active-domain accuracy discussion is limited in scope but is an extrapolation issue, not circularity. Overall, the paper's core physics and performance claims stand independently; the cone-penetration step is a partial circularity, so a moderate score of 4 is appropriate.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central claim rests on the continuum assumption, the specific rheology closure, the SPH discretization, and the active-domain localization assumption. No new physical entities are introduced; active domains, persistent neighbor lists, and BCE markers are algorithmic or existing constructs.

free parameters (4)
  • Artificial viscosity coefficient gamma_a = 0.01, 0.2, 0.02, 0.02 per test
    Table 1 lists a different gamma_a for each validation test; the paper calls it empirically determined, so the stabilization term is a tuned input that can influence penetration depths and wheel forces.
  • mu(I) rheology parameters (mu_s, mu_2, I_0) for cone penetration = 0.70, 0.80, 0.08 (glass beads); 0.80, 1.00, 0.08 (Ottawa sand)
    Sec. 3.2: 'These parameters were determined empirically, as specific experimental material data was not available.' They directly set the yield envelope used to match the cone penetration data.
  • Rheology grain diameter d = 1 mm, 3 mm, 7 mm depending on test
    Secs. 3.1-3.2: d enters the inertial number I in Eq. 17 and is chosen from the experimental particle sizes or from Ambroso's size-insensitivity finding; it is not measured within the simulation.
  • Active-domain memory parameters G, S, SI = 1.2, 0.75, 50
    Sec. 4.2: set empirically for memory management. They affect performance and memory behavior, not the physics, but the scalability claims rely on them.
assumptions (5)
  • domain assumption Granular media can be represented as a continuum 'fluid' with linear hypoelasticity plus mu(I) rheology (Eqs. 1-6, 17-19).
    Secs. 2.1-2.4: the constitutive model is inherited from refs. [13] and [23]; it assumes the macroscopic stress-strain-rate relation captures grain-scale behavior without resolving individual grains.
  • standard math SPH kernel interpolation with schemes F2/F3 converges for these flows and is stable with the chosen artificial viscosity.
    Sec. 2.2: standard SPH discretization; no convergence study is provided for the terramechanics cases.
  • domain assumption BCE markers impose an adequate no-slip boundary and stress extrapolation at fluid-solid interfaces.
    Sec. 2.5.1: coupling relies on extrapolation formulas from refs. [1], [53], and [23]; accuracy is tested only through the overall validation cases.
  • domain assumption Particle size does not significantly affect penetration depth, so the rheology length scale d can be set independently.
    Sec. 3.1: based on Ambroso et al. [2]; used to justify d = 1 mm for cratering.
  • domain assumption Terrain outside active boxes can be frozen without affecting the simulation outcome.
    Sec. 4.2 and Algorithm 3: justified by slow stress-wave propagation in soil; validated only for MGRU3 and RASSOR test cases with generously sized boxes.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Physics-Based Continuum Model for Versatile, Scalable, and Fast Terramechanics Simulation." pith.science (2026). https://pith.science/paper/3SO24WIV

@misc{pith2026250705643,
  author       = {Pith},
  title        = {Pith review of: A Physics-Based Continuum Model for Versatile, Scalable, and Fast Terramechanics Simulation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3SO24WIV}},
  note         = {Machine review of arXiv:2507.05643}
}
read the original abstract

This paper discusses Chrono's Continuous Representation Model (called herein Chrono::CRM), a general-purpose, scalable, and efficient simulation solution for terramechanics problems. Built on Chrono's Smoothed Particle Hydrodynamics (SPH) framework, Chrono::CRM moves beyond semi-empirical terramechanics approaches, e.g., Bekker-Wong/Janosi-Hanamoto, to provide a physics-based model able to address complex tasks such as digging, grading, as well as interaction with deformable wheels and complex grouser/lug patterns. The terramechanics model is versatile in that it allows the terrain to interact with both rigid and flexible implements simulated via the Chrono dynamics engine. We validate Chrono::CRM against experimental data from three physical tests, including one involving NASA's MGRU3 rover. In addition, the simulator is benchmarked against a high-fidelity Discrete Element Method (DEM) simulation of a digging scenario involving the Regolith Advanced Surface Systems Operations Robot (RASSOR). Being GPU-accelerated, Chrono::CRM achieves computational efficiency comparable to that of semi-empirical simulation approaches for terramechanics problems. Through an ``active domains'' implementation, Chrono::CRM can handle terrain stretches up to 10 km long with 100 million SPH particles at near interactive rates, making high-fidelity off-road simulations at large scales feasible. As a component of the Chrono package, the CRM model is open source and released under a BSD-3 license. All models and simulations used in this contribution are available in a public GitHub repository for reproducibility studies and further research.

Figures

Figures reproduced from arXiv: 2507.05643 by the authors.

Figure 1
Figure 1. Sphere Cratering test: (a) Schematic showing a sphere of radius 𝑅sphere = 0.0125 m dropped from a height 𝐻 onto granular media. (b) Results showing penetration depth correlation with empirical model. (c) Velocity field visualization for a sphere with density 𝜌sphere = 2200 kg∕m3 dropped from 𝐻 = 0.2 m. SPH particles are rendered as spheres with diameter 𝑑 = 𝑑0 = 2.5 mm. Similarly, the BCE markers are rendered with a… view at source ↗
Figure 2
Figure 2. Cone Penetration test: (a) Schematic of the test setup with cone geometries. (b) Velocity field visualization for the 60◦ cone on glass beads (Left) and the 30◦ cone on Ottawa sand (Right). Huzaifa Mustafa Unjhawala et al.: Preprint submitted to Elsevier Page 8 of 32 [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Simulation results for cone penetration depth versus time show good agreement with experiments [51], with relative errors below 10% for all cases. (Top) Data for a 60◦ cone in glass beads. (Bottom) Data for a 30◦ cone in Ottawa sand. Results cover three drop heights (0, 𝐿∕2, 𝐿, left to right) for each scenario, where 𝐿 is the cone length. We obtain mean relative errors across all drop heights 𝐻 of 3.5% for glass bea… view at source ↗
Figures from the paper (18 more)
Figure 4
Figure 4. Figure 4: Schematic of the MGRU3 single wheel tests. The gravity direction is modified to simulate ramps of angles from 0 ◦ to 30◦ . The wheel is given a constant angular velocity of 𝜔 = 0.8 rad∕s and is constrained to move in the 𝑥 direction. (a) Slip vs slope curve for the MGR…
Figure 5
Figure 5. Figure 5: Results obtained for the MGRU3 single wheel test: (a) The slip vs slope curve obtained compares the CRM results with the DEM results from [56] and the experimental results from the NASA SLOPE lab. The CRM results show close agreement with the DEM results and the experi…
Figure 6
Figure 6. Figure 6: RASSOR single drum tests: (a) Figures (1) to (6) depict the mechanism with which the RASSOR is expected to excavate lunar regolith. (1) to (3) show the left drum rotating counter-clockwise and the right drum rotating clockwise to fill the RASSOR drum. (4) to (6) show R…
Figure 7
Figure 7. Figure 7: Results obtained for the RASSOR single drum test: (a) The driving torque for the RASSOR single drum test compares the CRM results with the benchmark DEM-E results [56]. The CRM results show close agreement with the DEM results. (b) A snapshot of a cross-section view of…
Figure 8
Figure 8. Figure 8: Comparison of results using different neighbor list update frequencies (𝑝𝑠freq) ranging from 1 (update every step) to 10. Subfigures show results for: (a) MGRU3 (Sec. 3.3), (b) RASSOR (Sec. 3.4), and (c) Sphere Cratering (Sec. 3.1). The close agreement across different…
Figure 9
Figure 9. Figure 9: Additional simulation results for the cone penetration benchmarks using different neighbor list update frequencies (𝑝𝑠freq) ranging from 1 (update every step) to 10. (iii) 𝑁𝑎 and 𝑁𝑒 denote the number of active and extended-active particles, respectively, computed using…
Figure 10
Figure 10. Figure 10: Speedup factor (RTF1/RTF𝑝𝑠) for different simulations at varying neighbor list update frequencies. Higher values indicate better performance, with the baseline being RTF1 (updating the neighbor lists at every time step) [PITH_FULL_IMAGE:figures/full_fig_p017_10.png]
Figure 11
Figure 11. Figure 11: An illustration of the active domain in 2D. The “active box” OOBB (green box) moves with the solid body. SPH particles within the “active box” are flagged as Active. For computing properties of SPH particles at the boundary of the “active box”, the algorithm considers…
Figure 12
Figure 12. Figure 12: Illustration of the active domain partitioning for the validation test cases. The yellow “active box” is shown surrounding the primary object, MGRU3 wheel and RASSOR drum, respectively. Particles inside this box (Active) and those within 2ℎ of its boundary (Extended-A…
Figure 13
Figure 13. Figure 13: Comparison of results using active domains (dashed green lines) and without active domains (solid green lines) for the MGRU3 and RASSOR test cases. The results closely match the baseline approach, indicating minimal loss in accuracy due to the active domain optimizati…
Figure 14
Figure 14. Figure 14: Rendered views of the benchmark test cases used for performance comparison. (a) Baffle Flow, (b) RASSOR, (c) Tracked Vehicle, (d) Flexible Cable, (e) VIPER. • SCM Configurations: Chrono::SCM includes an optimization also called “active domains”, which limits contact d…
Figure 15
Figure 15. Figure 15: Comparison of RTF performance across benchmarks for different simulation configurations. The RTF-axis uses a logarithmic scale; lower bars indicate better performance (faster execution). Bars represent the Baseline [23], Chrono::CRM variants ((a), (b), and (c)), and S…
Figure 16
Figure 16. Figure 16: Scaling analysis of the Chrono::CRM solver using a Polaris RZR vehicle simulation. (a) Visual setup for the shortest terrain length. (b) Log-log plot of RTF versus terrain length, demonstrating the solver’s performance scalability on different GPUs. The dashed tangent…
Figure 17
Figure 17. Figure 17: Illustration of the four simulated test scenarios for the Polaris RZR: (Top Left) Rigid tire on rigid terrain, (Top Right) Flexible tire on rigid terrain, (Bottom Left) Rigid tire on deformable terrain, and (Bottom Right) Flexible tire on deformable terrain. Deformabl…
Figure 18
Figure 18. Figure 18: Comparison of vehicle longitudinal velocity over time for the four simulated scenarios: rigid tire on rigid terrain, flexible tire (110 kPa) on rigid terrain, rigid tire on deformable terrain, and flexible tire (110 kPa) on deformable terrain. The plot illustrates tha…
Figure 19
Figure 19. Figure 19: Progression of soil leveling towards a desired state (ℎdesired). The top panel shows the target heightmap. The subsequent panels illustrate the initial heightmap at 𝑇 = 0 s and the evolving actual heightmap at 𝑇 = 1 s, 𝑇 = 2 s, and 𝑇 = 6 s, as achieved by the autonomo…
Figure 20
Figure 20. Figure 20: Renderings of the Gator vehicle autonomously leveling a soil pile at different time instances. (Top) Initial state at 𝑇 = 0 s. (Middle) At 𝑇 = 1 s, the controller has lowered the blade to engage the soil. (Bottom) At 𝑇 = 2 s, the controller has rotated the blade upwar…
Figure 21
Figure 21. Figure 21: Time evolution of the Gator blade’s state during the autonomous leveling task. The plots compare the desired control inputs (pitch angle 𝑝 and vertical displacement 𝑑) generated by the controller with the actual achieved states of the blade. These inputs correspond to…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

56 extracted references · 49 canonical work pages

  1. [1]

    Ageneralizedwallboundaryconditionforsmoothedparticlehydrodynamics

    Adami,S.,Hu,X.,Adams,N.,2012. Ageneralizedwallboundaryconditionforsmoothedparticlehydrodynamics. JournalofComputational Physics 231, 7057–7075

    work page 2012

  2. [2]

    Penetration depth for shallow impact cratering

    Ambroso, M., Santore, C., Abate, A., Durian, D.J., 2005. Penetration depth for shallow impact cratering. Physical Review E 71, 051305

    work page 2005

  3. [3]

    The material point method for granular materials

    Bardenhagen, S., Brackbill, J., Sulsky, D., 2000. The material point method for granular materials. Computer methods in applied mechanics and engineering 187, 529–541

    work page 2000

  4. [4]

    Introduction to terrain-vehicle systems

    Bekker, M.G., 1969. Introduction to terrain-vehicle systems. University of Michigan Press, Ann Arbor, MI

    work page 1969

  5. [5]

    Learningtodrivefromsimulationwithoutrealworldlabels, in: 2019 International Conference on Robotics and Automation (ICRA), IEEE

    Bewley,A.,Rigley,J.,Liu,Y.,Hawke,J.,Shen,R.,Lam,V.D.,Kendall,A.,2019. Learningtodrivefromsimulationwithoutrealworldlabels, in: 2019 International Conference on Robotics and Automation (ICRA), IEEE. pp. 4818–4824

    work page 2019

  6. [6]

    Using simulation and domain adaptation to improve efficiency of deep robotic grasping, in: IEEE International Conference on Robotics and Automation (ICRA), IEEE

    Bousmalis, K., Irpan, A., Wohlhart, P., Bai, Y., Kelcey, M., Kalakrishnan, M., Downs, L., Ibarz, J., Pastor, P., Konolige, K., Levine, S., Vanhoucke, V., 2018. Using simulation and domain adaptation to improve efficiency of deep robotic grasping, in: IEEE International Conference on Robotics and Automation (ICRA), IEEE. pp. 4243–4250. doi:10.1109/ICRA.201...

    arXiv 2018

  7. [7]

    Lagrangian meshfree particles method (SPH) for large deformation and failure flows of geomaterial using elastic-plastic soil constitutive model

    Bui, H., Fukagawa, R., Sako, K., Shintaro, O., 2008. Lagrangian meshfree particles method (SPH) for large deformation and failure flows of geomaterial using elastic-plastic soil constitutive model. Int. J. of Numer. Anal. Meth. Geomech. 32, 1537–1570

    work page 2008

  8. [8]

    Smoothedparticlehydrodynamics(SPH)anditsapplicationsingeomechanics:Fromsolidfracturetogranular behaviour and multiphase flows in porous media

    Bui,H.H.,Nguyen,G.D.,2021. Smoothedparticlehydrodynamics(SPH)anditsapplicationsingeomechanics:Fromsolidfracturetogranular behaviour and multiphase flows in porous media. Computers and Geotechnics 138, 104315. doi:10.1016/j.compgeo.2021.104315

    arXiv 2021

Show all 56 references
  1. [9]

    On the use of simulation in robotics: Opportunities, challenges, and suggestions for movingforward

    Choi, H., Crump, C., Duriez, C., Elmquist, A., Hager, G., Han, D., Hearl, F., Hodgins, J., Jain, A., Leve, F., Li, C., Meier, F., Negrut, D., Righetti, L., Rodriguez, A., Tan, J., Trinkle, J., 2021. On the use of simulation in robotics: Opportunities, challenges, and suggestio...

    2021 doi

  2. [10]

    DualSPHysics: Open-source parallel CFD solver based on Smoothed Particle Hydrodynamics (SPH)

    Crespo, A.J., Domínguez, J.M., Rogers, B.D., Gómez-Gesteira, M., Longshaw, S., Canelas, R., Vacondio, R., Barreiro, A., García-Feal, O., 2015. DualSPHysics: Open-source parallel CFD solver based on Smoothed Particle Hydrodynamics (SPH). Computer Physics Communications 187, 204–216

    2015

  3. [11]

    Neighbour lists in smoothed particle hydrodynamics

    Domínguez, J., Crespo, A., Gómez-Gesteira, M., Marongiu, J., 2011. Neighbour lists in smoothed particle hydrodynamics. International Journal for Numerical Methods in Fluids 67, 2026–2042

    2011

  4. [12]

    CARLA: An open urban driving simulator, in: Proceedings of the 1st Annual Conference on Robot Learning, pp

    Dosovitskiy, A., Ros, G., Codevilla, F., Lopez, A., Koltun, V., 2017. CARLA: An open urban driving simulator, in: Proceedings of the 1st Annual Conference on Robot Learning, pp. 1–16

    2017

  5. [13]

    Continuum modelling and simulation of granular flows through their many phases

    Dunatunga, S., Kamrin, K., 2015. Continuum modelling and simulation of granular flows through their many phases. Journal of Fluid Mechanics 779, 483–513. doi:10.1017/jfm.2015.383

    2015 doi

  6. [14]

    Farchy, A., Barrett, S., MacAlpine, P., Stone, P., 2013. Humanoid robots learning to walk faster: From the real world to simulation and back, in: Proceedings of the 2013 international conference on Autonomous agents and multi-agent systems, pp. 39–46

    2013

  7. [15]

    Error estimation in smoothed particle hydrodynamics and a new scheme for second derivatives

    Fatehi, R., Manzari, M.T., 2011. Error estimation in smoothed particle hydrodynamics and a new scheme for second derivatives. Computers & Mathematics with Applications 61, 482–498

    2011

  8. [16]

    Towards scientific machine learning for granular material simulations – challenges and opportunities

    Fransen, M., Fürst, A., Tunuguntla, D., Wilke, D.N., Alkin, B., Barreto, D., Brandstetter, J., Cabrera, M.A., Fan, X., Guo, M., Kieskamp, B., Kumar, K., Morrissey, J., Nuttall, J., Ooi, J., Orozco, L., Papanicolopulos, S.A., Qu, T., Schott, D., Shuku, T., Sun, W., Weinhart, T....

    2025 arXiv

  9. [17]

    Fast,generic, andreliablecontrolandsimulationofsoftrobotsusingmodelorderreduction

    Goury,O.,Duriez,C.,2018. Fast,generic, andreliablecontrolandsimulationofsoftrobotsusingmodelorderreduction. IEEETransactions on Robotics 34, 1565–1576

    2018

  10. [18]

    Lift and drag forces on a moving intruder in granular shear flow

    He, H., Zhang, Q., Ottino, J.M., Umbanhowar, P.B., Lueptow, R.M., 2025. Lift and drag forces on a moving intruder in granular shear flow. Journal of Fluid Mechanics 1008, A5. doi:10.1017/jfm.2025.87

    2025 doi

  11. [19]

    Smooth particle hydrodynamics simulations of low Reynolds number flows through porous media

    Holmes, D.W., Williams, J.R., Tilke, P., 2011. Smooth particle hydrodynamics simulations of low Reynolds number flows through porous media. International Journal for Numerical and Analytical Methods in Geomechanics 35, 419–437

    2011

  12. [20]

    A study demonstrating that using gravitational offset to prepare extraterrestrial mobility missions is misleading

    Hu, W., Li, P., Rogg, A., Schepelmann, A., Chandler, S., Kamrin, K., Negrut, D., . A study demonstrating that using gravitational offset to prepare extraterrestrial mobility missions is misleading. Journal of Field Robotics URL: https://onlinelibrary.wiley.com/doi/abs/10.1002/...

    doi

  13. [21]

    Using physics-based simu- lation towards eliminating empiricism in extraterrestrial terramechanics applications

    Hu, W., Li, P., Rogg, A., Schepelmann, A., Creager, C., Chandler, S., Kamrin, K., Negrut, D., 2024a. Using physics-based simu- lation towards eliminating empiricism in extraterrestrial terramechanics applications. URL: https://arxiv.org/abs/2405.11001, arXiv:2405.11001

    arXiv

  14. [22]

    Calibration of an expeditious terramechanics model using a higher-fidelity model, Bayesian inference, and a virtual bevameter test

    Hu, W., Li, P., Unjhawala, H.M., Serban, R., Negrut, D., 2024b. Calibration of an expeditious terramechanics model using a higher-fidelity model, Bayesian inference, and a virtual bevameter test. Journal of Field Robotics 41. URL:https://doi.org/10.1002/rob.22276

    doi

  15. [23]

    Modeling granular material dynamics and its two-way coupling with moving solidbodiesusingacontinuumrepresentationandtheSPHmethod

    Hu, W., Rakhsha, M., Yang, L., Kamrin, K., Negrut, D., 2021. Modeling granular material dynamics and its two-way coupling with moving solidbodiesusingacontinuumrepresentationandtheSPHmethod. ComputerMethodsinAppliedMechanicsandEngineering385,114022

    2021

  16. [24]

    How simulation helps autonomous driving: A survey of sim2real, digital twins, and parallel intelligence

    Hu, X., Li, S., Huang, T., Tang, B., Huai, R., Chen, L., 2023. How simulation helps autonomous driving: A survey of sim2real, digital twins, and parallel intelligence. IEEE Transactions on Intelligent Vehicles , 1–20doi:10.1109/TIV.2023.3312777

    2023

  17. [25]

    Taichi: a language for high-performance computation on spatially sparse data structures

    Hu, Y., Li, T.M., Anderson, L., Ragan-Kelley, J., Durand, F., 2019. Taichi: a language for high-performance computation on spatially sparse data structures. ACM Transactions on Graphics (TOG) 38, 1–16

    2019

  18. [26]

    The analyticaldetermination ofdrawbar pullasa functionof slipfortracked vehiclesin deformablesoils,in: Proc of the 1st int conf mech soil–vehicle systems

    Janosi,Z., Hanamoto,B., 1961. The analyticaldetermination ofdrawbar pullasa functionof slipfortracked vehiclesin deformablesoils,in: Proc of the 1st int conf mech soil–vehicle systems. Turin, Italy

    1961

  19. [27]

    A literature review of simulation fidelity for autonomous-vehicle research and development

    Johnson, C., Graupe, E., Kassel, M., 2023. A literature review of simulation fidelity for autonomous-vehicle research and development. SAE Int. J. Aerosp. 16, 253–261. URL:https://doi.org/10.4271/01-16-03-0021, doi:10.4271/01-16-03-0021

    2023 doi

  20. [28]

    Design and use paradigms for Gazebo, an open-source multi-robot simulator, in: IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Sendai, Japan

    Koenig, N., Howard, A., 2004. Design and use paradigms for Gazebo, an open-source multi-robot simulator, in: IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Sendai, Japan. pp. 2149–2154. doi:10.1109/IROS.2004.1389727

    2004 arXiv

  21. [29]

    NASA’s VIPER prototype motors through moon-like obstacle course.https://www.nasa.gov/feature/ames/ nasas-viper-prototype-motors-through-moon-like-obstacle-course

    Laboratory, S.L.O., . NASA’s VIPER prototype motors through moon-like obstacle course.https://www.nasa.gov/feature/ames/ nasas-viper-prototype-motors-through-moon-like-obstacle-course . Accessed: 2023-04-02

    2023

  22. [30]

    Cyberbotics ltd

    Michel, O., 2004. Cyberbotics ltd. webots: professional mobile robot simulation. International Journal of Advanced Robotic Systems 1, 5

    2004

  23. [31]

    Granulargym: High performance simulation for robotic tasks with granular materials

    Millard, D., Pastor, D., Bowkett, J., Backes, P., Sukhatme, G.S., 2023. Granulargym: High performance simulation for robotic tasks with granular materials

    2023

  24. [32]

    Smoothed Particle Hydrodynamics

    Monaghan, J.J., 2005. Smoothed Particle Hydrodynamics. Reports on Progress in Physics 68, 1703–1759

    2005

  25. [33]

    Shock simulation by the particle method SPH

    Monaghan, J.J., Gingold, R., 1983. Shock simulation by the particle method SPH. Journal of Computational Physics 52, 374–389

    1983

  26. [34]

    A refined particle method for astrophysical problems

    Monaghan, J.J., Lattanzio, J.C., 1985. A refined particle method for astrophysical problems. Astronomy and astrophysics 149, 135–143

    1985

  27. [35]

    Mueller,R.P.,Cox,R.E.,Ebert,T.,Smith,J.D.,Schuler,J.M.,Nick,A.J.,2013.Regolithadvancedsurfacesystemsoperationsrobot(RASSOR), in: 2013 IEEE Aerospace Conference, IEEE. pp. 1–12

    2013

  28. [36]

    Design of an excavation robot: regolith advanced surface systems operations robot (rassor) 2.0, in: ASCE Earth & Space Conference, American Society of Civil Engineers

    Mueller, R.P., Smith, J.D., Schuler, J.M., Nick, A.J., Gelino, N.J., Leucht, K.W., Townsend, I.I., Dokos, A.G., 2016. Design of an excavation robot: regolith advanced surface systems operations robot (rassor) 2.0, in: ASCE Earth & Space Conference, American Society of Civil En...

    2016

  29. [37]

    Isaac SDK: NVIDIA’s Robotics Platform

    NVIDIA Corporation, 2023. Isaac SDK: NVIDIA’s Robotics Platform. URL:https://developer.nvidia.com/isaac-sdk. available: https://developer.nvidia.com/isaac-sdk

    2023

  30. [38]

    Design of a generic Chrono framework for fluid-solid interaction

    Serban, R., Bakke, L., Unjhawala, H., Negrut, D., 2025. Design of a generic Chrono framework for fluid-solid interaction

    2025

  31. [39]

    Real-timesimulationofgroundvehiclesondeformableterrain

    Serban,R.,Taves,J.,Zhou,Z.,2023. Real-timesimulationofgroundvehiclesondeformableterrain. JournalofComputationalandNonlinear Dynamics 18, 081007

    2023

  32. [40]

    Chrono::Vehicle template-based ground vehicle modeling and simulation

    Serban, R., Taylor, M., Negrut, D., Tasora, A., 2019. Chrono::Vehicle template-based ground vehicle modeling and simulation. International Journal of Vehicle Performance 5, 18–39. doi:10.1504/IJVP.2019.097096

    2019

  33. [41]

    Dynamics of Multibody Systems

    Shabana, A.A., 2020. Dynamics of Multibody Systems. Fifth ed., Cambridge University Press, Cambridge, England

    2020

  34. [42]

    AirSim: High-fidelity visual and physical simulation for autonomous vehicles, in: Field and service robotics, Springer

    Shah, S., Dey, D., Lovett, C., Kapoor, A., 2018. AirSim: High-fidelity visual and physical simulation for autonomous vehicles, in: Field and service robotics, Springer. pp. 621–635

    2018

  35. [43]

    Gen ai-powered synthetic data generation in robotics

    SoftServe Inc., 2024. Gen ai-powered synthetic data generation in robotics. URL:https://www.softserveinc.com/en-us/blog/ ai-powered-synthetic-data-for-robotics . accessed: 2025-04-15

    2024

  36. [44]

    Gradient-based trajectory optimization with learned dynamics, in: 2023 IEEE International Conference on Robotics and Automation (ICRA)

    Sukhija, B., Köhler, N., Zamora, M., Zimmermann, S., Curi, S., Krause, A., Coros, S., 2023. Gradient-based trajectory optimization with learned dynamics, in: 2023 IEEE International Conference on Robotics and Automation (ICRA). Final version available athttps: //crl.ethz.ch/pa...

    2023

  37. [45]

    Chrono: An open source multi-physics dynamics engine, in: Kozubek, T

    Tasora, A., Serban, R., Mazhar, H., Pazouki, A., Melanz, D., Fleischmann, J., Taylor, M., Sugiyama, H., Negrut, D., 2016. Chrono: An open source multi-physics dynamics engine, in: Kozubek, T. (Ed.), High Performance Computing in Science and Engineering – Lecture Notes in Compu...

    2016

  38. [46]

    An efficiency comparison of different ANCF implementations

    Taylor, M., Serban, R., Negrut, D., 2023a. An efficiency comparison of different ANCF implementations. International Journal of Non- LinearMechanics149,104308. URL: https://www.sciencedirect.com/science/article/pii/S0020746222002785,doi: https: //doi.org/10.1016/j.ijnonlinmec....

  39. [47]

    Implementation implications on the performance of ANCF simulations

    Taylor, M., Serban, R., Negrut, D., 2023b. Implementation implications on the performance of ANCF simulations. International Journal of Non-Linear Mechanics 149, 104328. URL: https://www.sciencedirect.com/science/article/pii/S0020746222002980, doi:https://doi.org/10.1016/j.ijn...

  40. [48]

    Low-speedimpactcratersinloosegranularmedia

    Uehara,J.,Ambroso,M.,Ojha,R.,Durian,D.J.,2003. Low-speedimpactcratersinloosegranularmedia. PhysicalReviewLetters90,194301

    2003

  41. [49]

    Reinforcement learning from simulation to real world autonomous driving using digital twin

    Voogd, K.L., Allamaa, J.P., Alonso-Mora, J., Son, T.D., 2023. Reinforcement learning from simulation to real world autonomous driving using digital twin. IFAC-PapersOnLine 56, 1510–1515

    2023

  42. [50]

    Piecewise polynomial, positive definite and compactly supported radial functions of minimal degree

    Wendland, H., 1995. Piecewise polynomial, positive definite and compactly supported radial functions of minimal degree. Advances in Computational Mathematics 4, 389–396. doi:10.1007/BF02123482

    1995 doi

  43. [51]

    Cone penetration results for 3mm glass beads and 20-30 Ottawa sand

    Williams, K., 2016. Cone penetration results for 3mm glass beads and 20-30 Ottawa sand. Technical Report TR-2016-04: http: //sbel.wisc.edu/documents/TR-2016-04.pdf. Simulation-Based Engineering Laboratory, University of Wisconsin-Madison. URL: http://sbel.wisc.edu/documents/TR...

    2016

  44. [52]

    Prediction of rigid wheel performance based on the analysis of soil-wheel stresses Part I

    Wong, J.Y., Reece, A.R., 1967. Prediction of rigid wheel performance based on the analysis of soil-wheel stresses Part I. Performance of driven rigid wheels. Journal of Terramechanics 4, 81–98

    1967

  45. [53]

    Three-dimensional modeling of granular flow impact on rigid and deformable structures

    Zhan, L., Peng, C., Zhang, B., Wu, W., 2019. Three-dimensional modeling of granular flow impact on rigid and deformable structures. Computers and Geotechnics 112, 257–271

    2019

  46. [54]

    Data-driven bulldozer blade control for autonomous terrain leveling

    Zhang, H., Arivoli, G., Unjhawala, H., Bakke, L., Serban, R., Negrut, D., 2025. Data-driven bulldozer blade control for autonomous terrain leveling. IEEE Robotics and Automation Letters submitted

    2025

  47. [55]

    Simulation Reproducibility for Different Path Following Policies

    Zhang, H., Caldararu, S., Young, A., Ruiz, A., Unjhawala, H., Ashokkumar, S., Mahajan, I., Batagoda, N., Bakke, L., Negrut, D., 2024a. Simulation Reproducibility for Different Path Following Policies. https://github.com/uwsbel/sbel-reproducibility/tree/ master/2024/PathFollowi...

    2024

  48. [56]

    A GPU- accelerated simulator for the DEM analysis of granular systems composed of clump-shaped elements

    Zhang, R., Vanden Heuvel, C., Schepelmann, A., Rogg, A., Apostolopoulos, D., Chandler, S., Serban, R., Negrut, D., 2024b. A GPU- accelerated simulator for the DEM analysis of granular systems composed of clump-shaped elements. Engineering with Computers , 1–21. Huzaifa Mustafa...

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.