Dear Dan, Marcel,
Thank you both for your responses. Dan – I truly appreciate the pointer to *The Last Garage*. It’s an impressive piece of work, and I fully understand that such results are the outcome of sustained effort—not something to be achieved in a short afternoon. I certainly don’t underestimate the complexity involved in building robust real-time simulations. Marcel – your practical insights are very helpful, especially the emphasis on decoupling rendering from the physics loop. That’s indeed the architecture we follow as well. In our case, the 1 kHz target is not due to fast motion dynamics, but rather stems from the requirements of the motion cueing system and real-time hardware integration. While we employ classical motion cueing algorithms (e.g., washout filters and tilt coordination), which operate in the lower frequency range (typically below 10 Hz), the simulation loop needs to run at high frequency to ensure low-latency signal delivery and stable control of the motion platform. Our setup involves full dynamic simulators with a human operator onboard, so the control interface must respond smoothly and predictably to external disturbances. By way of background: I’ve been developing simulation software myself for many years and have been involved in scientific research in multibody dynamics and real-time applications. We’ve developed our own simulation kernel that achieves real-time performance for rigid MBS, but it currently lacks in tire modeling and terrain interaction. This is where Chrono appears promising—especially due to its open architecture and extensibility. Your replies have given me the confidence that a prototype based on Chrono is a worthwhile path forward. I’m particularly interested in exploring how customizable the tire-terrain interaction can be for our application. Thanks again for your openness and support. Best regards, Stefan Oberpeilsteiner [email protected] schrieb am Donnerstag, 31. Juli 2025 um 19:00:20 UTC+2: > Hello Stefan, Dan, > > Let me chime in here as that's "my" project. It indeed uses Chrono to run > real-time vehicle physics (with a lot of work on top as well as Dan > mentions, I've been working on this full-time for over three years now). > > Let me answer your questions though: > > As to how to achieve "stable" simulation at 1000 Hz (to be honest, you > don't always need 1000, 500 might be fine too, but that's another > discussion), what I've done is completely removed the graphics rendering > from the core simulation loop as a first step: you certainly don't need to > render every 1 ms, most GPUs won't be able to keep up with that, no > monitors can. So decouple that and make a physics loop that does just that: > calculate the physics. Well, there is one other thing you want to do there > and that's read controller inputs (and possibly provide force feedback > effects to wheel and pedals). Then you simply need that thread to keep up > with real-time and pace it as well as you can. Most modern computers will > be fast enough. Most operating systems you run this on won't be (hard) > real-time OS'es and that means you have to settle for a "best effort" > synchronization with time. In practice that is very do-able though in my > experience. > > As for the solver, I agree with Dan's feedback, start with the defaults. I > found them pretty good. > > As for performance bottlenecks on a non-deformable, triangulated mesh, I > would say you might run into some bottlenecks there if you end up with a > big mesh, but like so many things in Chrono, the "tire vs mesh" interaction > is something you can optimize and plug in your own implementation and there > are many ways to replace what's there now with something more specialized > for your use case. > > All in all I would say what you're trying to do is certainly feasible. > Chrono is a great platform for this. > > Greetings, Marcel > > > On 26-Jul-25 19:39, 'Dan Negrut' via ProjectChrono wrote: > > Hi Stefan – here’s an example of real-time simulation using Chrono: > https://thelastgarage.com/ . > > Disclaimer: the link above is not to a “one afternoon” project, but what > you want is very likely possible. > > As about your three questions, I would start with the defaults in Chrono > and then tweak the model parameters to get the level of performance you > need. > > Dan > > --------------------------------------------- > > Bernard A. and Frances M. Weideman Professor > > NVIDIA CUDA Fellow > > Department of Mechanical Engineering > > Department of Computer Science > > University of Wisconsin - Madison > > 4150ME, 1513 University Avenue > > Madison, WI 53706-1572 > > 608 772 0914 > > http://sbel.wisc.edu/ > > http://projectchrono.org/ > > --------------------------------------------- > > > > *From:* [email protected] <[email protected]> *On > Behalf Of *Stefan Oberpeilsteiner > *Sent:* Friday, July 25, 2025 5:08 AM > *To:* ProjectChrono <[email protected]> > *Subject:* [chrono] Real-Time Capability of Rigid Body Vehicle Model on > Rigid Terrain > > > > Dear Chrono Team, > > I am currently implementing a simplified tractor model in Chrono for use > in a real-time simulation environment. The model consists of a rigid > chassis, a fixed rear axle with steered and driven wheels, and a front axle > modeled as a rigid beam suspended via a central pivot (as commonly used in > tractor front axles) with additional steering. The goal is to capture the > essential dynamic behavior of the vehicle—particularly in heave, pitch, and > roll—sufficiently well for deployment on a motion platform. > > The vehicle operates on a rigid, non-deformable terrain, represented by a > tessellated triangular mesh. The primary excitation arises from > terrain-induced vertical disturbances with dominant frequency components up > to approximately 10 Hz. > > *My core question is:* > What is the expected feasibility of running such a model in real time > using Chrono? In particular: > > - Are there benchmarks or best practices for achieving stable > simulation at control loop frequencies ~1 kHz? > - Which solver/integrator is recommended for this type of setup? > - Are there known performance bottlenecks when using non-deformable > terrains modeled as triangular meshes? > > Any insights—especially from prior experience with real-time vehicle > simulation or motion cueing applications—would be highly appreciated. > > Best regards, > Stefan Oberpeilsteiner > > -- > You received this message because you are subscribed to the Google Groups > "ProjectChrono" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > To view this discussion visit > https://groups.google.com/d/msgid/projectchrono/ddfef702-fd37-4479-aceb-72186eeac1f0n%40googlegroups.com > > <https://urldefense.com/v3/__https:/groups.google.com/d/msgid/projectchrono/ddfef702-fd37-4479-aceb-72186eeac1f0n*40googlegroups.com?utm_medium=email&utm_source=footer__;JQ!!Mak6IKo!OiK8ujXNh3As54pZIhOOwQOT6cgJDkI2-J8LI6YfZlY5nGUWl67isPd8wIg--H1NPXJfblqL_ZWseE9tE_UgWcwN3w$> > . > -- > You received this message because you are subscribed to the Google Groups > "ProjectChrono" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > > To view this discussion visit > https://groups.google.com/d/msgid/projectchrono/DM8PR06MB770344053BDBC5518E6BCB10B158A%40DM8PR06MB7703.namprd06.prod.outlook.com > > <https://groups.google.com/d/msgid/projectchrono/DM8PR06MB770344053BDBC5518E6BCB10B158A%40DM8PR06MB7703.namprd06.prod.outlook.com?utm_medium=email&utm_source=footer> > . > > -- You received this message because you are subscribed to the Google Groups "ProjectChrono" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion visit https://groups.google.com/d/msgid/projectchrono/f9a30e4d-4643-4ed3-b82d-188d7c7fbd1dn%40googlegroups.com.
