Community Question: Should Racing Simulators Prioritise Soft-body Physics?

Cover Image.jpg
BeamNG's class-leading damage model
Soft-body physics has always been a hot topic in the sim racing world, but which simulators could pull it off? What titles would it ruin?

Soft-body physics, or dynamics in the wider reaches of the technology, is a field of simulation that focuses on visually realistic physical simulations of the motion and properties of deformable objects or better known as 'Soft Bodies'.

The most common use of soft body physics within racing simulators is the damage model. Titles such as BeamNG.drive and its predecessor, Rigs of Rods, both show just how detailed and difficult the technology is to master. BeamNG has been in continual development since its release in 2013.

Pothole Damage.jpg

BeamNG's Soft-body damage model.

BeamNG is the obvious choice for a simulator in which soft-body physics works wonders. Without it, the project would have very little to write home about. But what current sim racing titles and projects would positively benefit from adding in a complex damage model that models both the visual and physical implications of damage realistically?

Could iRacing adopt soft-body physics?​

The sim with the best damage model is often considered to be iRacing, however, their damage model is still far from the level of BeamNG. The iRacing damage model is imposing until you start reaching the deeper levels of detail.

There are not too many chassis dents, just bodywork and mechanical damage that you can feel through your wheel - plus a myriad of parts flying off the cars, as well as wheels being visibly crooked or only hanging on by a tether if they take a solid hit. That being said, the iRacing damage model is certainly a class leader in the world of pure racing simulators currently.

iRacing's new catalogue-wide damage model as of June 2024

So would iRacing benefit from a BeamNG style of soft body physics? We do not think so. The newer damage model suits the simulator perfectly well. Despite lacking detail in some of the more intricate levels of damage detail, iRacing does a fantastic job of simulating what a racing crash looks like - something BeamNG often exaggerates and consequently falls down on.

Why Rally Is The Perfect Candidate​

If you have ever played the iconic rally title Richard Burns Rally, you will know jut how punishing a damage model for this discipline can be. However, modern iterations of rally titles have toned it down a bit when it comes to a comprehensive damage model.

Both DIRT Rally 1 and 2 had a satisfactory damage model with impactful mechanical damage and failures, however, some of the crashes that would have ended your rally at best simply require you to reverse and continue with a slight crack in the windscreen and a missing front bumper.

1.jpg

Porsche 911, Dirt Rally 2.0. Image: Codemasters

EA Sports WRC
has this same problem. In small and medium-speed accidents, you are punished to a fairly high standard. If you clatter a wall, you will mostly likely damage your bodywork and maybe your suspension. However, a high-speed multi-rollover accident will more than likely not damage your car all that much, and certainly will not cost you a lot of time - even with hardcore damage enabled.

So is the world of rally crying out for soft-body physics? Let's take a short look into BeamNG'drive' s rally world to see if it could work.

BeamNG Rally 1.0​

BeamNG Rally has been created and published by Track Broseff on the BeamNG repository, and this mod breathes fresh new life into the title with custom rally stages on dirt and tarmac based on the real-world American Rally Association ruleset.

Rally Notes 1.jpg

BeamNG's rally pace notes

Over 88 miles/141 kilometres of rally stages are available, and all stages can also be run in reverse within the time trial menus. The maps range from Utah to Jungle Rock Island and everything in between. The real draw to this specific rally mod, however, is the inclusion of pace notes. A revolutionary addition that has put BeamNG on the map for rally fans.

Combine these essential rally additions with BeamNG's soft-body physics and you have a rally game that is as punishing and brutal as it can get in sim racing. Overall, it works brilliantly. The pace notes are mostly accurate with a few further additions that you will not see in any other games apart from Richard Burns Rally.

Should 'Realistic' Damage be the aim?​

But how about circuit racing? Often when the subject of damage models in sim racing is brought up, you will hear the take that the most realistic is the best. This is not always the case. Again, let's look at BeamNG. Your PC's required processing power to simulate your car is immense. Imagine trying to do that for twenty-plus cars simultaneously while also producing the driving physics and running an online server.

190E Merc Damage.jpg

Medium-speed wall impact damage in Automobilista 2

Most simulators skip a high-detail damage model in favour of a more immersive and detailed driving experience. This is not uncommon and is a tried and true method of creating an accurate racing simulation - it is the core of the experience, after all. Racing titles, especially pure simulators, should not entirely sacrifice their damage model; there is no question about it.

With that being said, accurately realistic damage should not be a priority for a lot of simulators. The relatively forgiving damage model of RaceRoom, for instance, could be considered a part of why the title can be run by lower-end PCs as well as forgiving small mistakes online, a key addition for teaching new players what is good and what is not without ending their race.

Which simulators do you think should include soft-body physics? Should it be kept to titles similar to BeamNG? Let us know in the comments below!
About author
Connor Minniss
Website Content Editor & Motorsport Photographer aiming to bring you the best of the best within the world of sim racing.

Comments

The difference between Steel, Aluminium, Glass Fiber, Carbon fiber, they all react differently in an accident... are you gonna do them all, will the Marcos and Morgan still have Wooden chassis? to what point do you take the sim?
Will the next big thing be all moving parts take on their own physics, will the proshaft rip through the floor if the bolts sheer, the 10 kg flywheel smash through the soft cast ally bellhousing and make a bid for freedon across the fields at 6000RPM, it's great to be able to do stuff like that, but, I don't think it's really that important.

And as has been mentioned the object is to not damage the car in sim racing, I can understand that certain sections of the community (VR) might find it interesting to get out of the car and walk round...yeah, ok but that's like the grass and leaves on the trees... I don't care about that too much as I drive past at 90mph, I simply don't see it on my pancake screen... unless I'm upside down in a roadster!
It will be a lot better to have a well-structured and involving single player career mode like Gran Turismo rather than modelling the best crash physics ever. Just my 2 cents.
 
Premium
Reading through some of the comments here, it’s clear that there is a wealth of insight into the diverse and varied personal preferences of individual Sim Racers. The range of opinions and perspectives shared highlights the unique tastes and interests that make each person’s viewpoint valuable and interesting. This diversity enriches the conversation and provides a broader understanding of the topic at hand.:thumbsup:

Personally, I am very pleased with the current range of Sims. With most of them installed and regularly played, I find myself thoroughly satisfied with each, giving a great experience in their own way.:)
 
Last edited:
> BeamNG has good physics for offroad driving

> sounds brilliant, let's look for whether there's anything for offroad racing with AI

> search Youtube

> there's nothing but page after page of crashes created in BeamNG

> just crashes and launching the car off a cliff

> crashes and cliffs

> occasional rally video by Jimmeh

> then crashes again

> and cliffs

> rock climbing! how original

> crashes again

> and again crashes and cliffs

> motorbikes in BeamNG? yeah, again crashes and nothing else

> just foking crashes all day long

> crashes in the morning, crashes in the afternoon, crashes in the evening

> just nothing but crashes as far as the eye can see
 
Is it possible to have only the damage calculations according to the soft bodies physics without concretizing them in visual effects? Like a fighter pilot with master caution panel of defects and warning lights that tells him what's wrong, he doesn't necessarily see the breakage. A visual indication and a strong influence on driving could be enough. This could completely change the way multiplayer games could be played. You break your car, it's over.
 
Last edited:
Damage physics creates a huge variation in a race, from a tire stall to debris damaging other cars, a run-off and slight collision delighting a car, debris puncturing or blowing out tires, debris causing other cars to spin or puncture radiators, tires, breaking for windshields, collisions that even end in engine bursting, these bursts dirtying the windshield of those who are behind, can create so many variants in a race that I think it is impossible for someone who loves simulation to be against it, even if this can be turned off for those who don't like it, with me it has happened to lose an Indy race in Rfactor 2, because pieces of cars broke my suspension with 12 laps of 200 to go, disappoints, but you know it can happen, that's simulating, I'm totally in favor of it, as long as it's done well.
 
I think the damage is a punishment, no one runs here to make collisions in the race, this is the consequence of your mistake or the opponent's mistake that hits you.

A full damage model that punishes you even if you scrape the guardrail is cool, after all, the guardrail isn't meant to be scraped.

If a race with realistic damage has a lot of YF disrupting the race, it's the virtual racers' fault, right? Of course it's going to be a boring race, but that's down to bad driving, not the law of physics.

I personally like to see the more real the better, especially in an online race broadcast that I watch, but the other side of having to train to be good in a very real simulator is that for me it gets in the way, goodbye real world.

For those who don't have time to train, run online, or a super PC (I include myself in all of this), the middle ground is enough.
 
I'm all for soft body physics if it's simulated properly. In BeamNG steel behaves like tinfoil and in Wreckfest the damage is only cosmetic. You can still drive the car even when it has been squished from all sides and looks like a moped.
Soft body physics simulating every major structural part of the vehicle could offer superior simulation and a larger variation in terms of handling between cars, other than the most commonly simulated difference of weight and center of gravity.
 
Premium
Skipping over the actual soft body physics discussion for a second, online racing is arguably an important aspect of any racing simulation. Ultimately it's a lot more satisfying to beat real people (your friends) than AI who, more often than not, don't even use the same physics to drive their virtual cars as you.

Now online racing poses a big problem when it comes to collision detection and the resulting damage calculations (soft body or not). Driving at around 200 km/h, over 50 m/s, a latency of 100 ms means the opponent position you receive is already 5 meters behind where that car is now. I know prediction can do a very reasonable job to figure out where he probably is now, but even if the algorithm is 90% accurate in its prediction, you are still looking at about 50 cm of margin. Cars are quite heavy so at those speeds they carry a lot of energy and even small collisions between cars can already create large forces. Imagine two cars rubbing paint and the algorithm getting it wrong by 50 cm, making the cars "overlap" by that amount. That would create quite a dramatic force, resulting in a lot of damage. This would probably not be perceived as realistic at all.

My point: latency in online races ruins your chances of accurate damage physics because you simply do not know exactly where the other cars are. And latency is a problem we can't really solve, unless someone either shrinks the earth or speeds up light, or both.

Now obviously this would only be true for online car to car contact. Slamming your car into barriers would be fine.
 
Premium
Wouldn't simply making the hit box smaller then the model compensate for the latency?

It would be no more unrealistic then cars causing damage without actual contact, and would allow for close racing while still be activated from a proper collision.
 
Premium
How much smaller then? If 100 ms at 200 km/h means you really do not know where the car is (as it has moved 5 meters) then your safest bet would be to reduce the hit box by 5 meters, which is probably the whole car?! Taking my own 90% accuracy point above, it would still reduce it by 50 cm (which means a significant overlap for cars without bumping into each other). Racing someone with a worse connection (I've seen 200+ ms latency in international races) means that would go to 1 meter. I don't think that would work very well either. It's not an easy problem to solve and in the end prediction is guesswork. Where those opponents get rendered is probably not exactly where they actually are.
 
Premium
If I were calling the shots I'd go for 50cm and be done with it unless real world use dictated a better alternative.

Once the ping is over 200ms then there are going to be a multitude of issues that affect the experience that people either have to live with or go do something else.
 
Last edited:
Why are we still using hit-boxes in the 2020s. We don't even have to have soft-body physics but at least get rid of hit-boxes. Can't all the physical objects, at least vehicle and circuit, be actual physical entities in terms of at least contact with each other?

Or can't devs make the hit-boxes closely follow the shape of each object instead of literally just being a box?

Excuse my ignorance on this subject.
 
Soft body physics simulating every major structural part of the vehicle could offer superior simulation and a larger variation in terms of handling between cars, other than the most commonly simulated difference of weight and center of gravity.
I'd rather just do FEA and make a more empiric model based on that. Maybe then it'd actually run in realtime and would be stable enough to use. BeamNG runs fast enough to have accurate compliances in a rigidbody model and the softbodies still vibrate profusely.
 
Premium
We don't.

And box or not is totally irrelevant in the discussion about latency.
Steady...
he did say "excuse my ignorance", however, hit box, colision zone etcetra as far as I see have more in common with soft body for contact than latency* surely,
but conversations and threads evolve and often take on a different trajectory while still going farward :thumbsup:

*I'm no expert, but I understand what Marcell was saying, and it's relevace!
 
Premium
It was me and my ignorance that introduced hit box's into the conversation.

Guess this just leave shrinking the size of the earth and increasing the speed of light.

If anyone wants my 2 cents on how this can be achieved then just yell out, I'm on a roll.

And yes, It involves hit boxes.
 
Premium
Can't all the physical objects, at least vehicle and circuit, be actual physical entities in terms of at least contact with each other?

Or can't devs make the hit-boxes closely follow the shape of each object instead of literally just being a box?
I think the issue is most sims what you see has no physical properties and what has physical properties you don’t see in game. So it is going to be performance saving to not have so many verts in the the physical mesh. And also if just the Armco or other wall or guide rails that don’t touch the ground, I think there would be issues with getting stuck more easily under physical mesh if just what you see was the collider. But I am no expert either just my thoughts.
 

Latest News

Article information

Author
Connor Minniss
Article read time
4 min read
Views
3,994
Comments
73
Last update
Back
Top