How to cyber-engineer race cars
Within the racing game genre, engineering artists such as Casey Ringley are in high- demand—it’s a role that takes years to master (Casey got his start back in the late 1990s), incorporating a passion for mechanics, a deep knowledge of both engineering and physics, and—perhaps most crucially—a test-driver’s ‘feel’ for how a car is reacting on the limit. This is how he engineers the feel for in-game cars from the most iconic automakers on the planet.
Casey Ringley is Lead Technical Vehicle Artist at Slightly Mad Studios; he leads the engineering process on all vehicle art production and in-game visual-effects systems—in other words, building them and then ensuring that they handle correctly.
For those who are interested in his day-to-day job, Casey gave a fascinating interview with Honda, which you can read here.
Creating a car for a high-end simulator is about more than simply inputting raw numbers into the physics engine; no matter how accurate the engine is, there is always the ‘human’ touch that is needed to ensure the car handles as it does out in the real-world. The same, of course, applies to laser-scanned tracks: you can scan a track to ensure it replicates the real-world track to almost the millimetre, but only a driver with experience can pin-point why one turn will have just a little less grip than the rest of the track, for reasons no-one will ever know.
During the lead-up to the release of Project CARS 2, a series of videos were produced (#BuiltbyDrivers) highlighting this close working relationship between auto manufacturers, drivers, tyre manufacturers, and the studio. For Project CARS 2, drivers were not simply endorsing the product; new motorsports such as rallycross needed the sure-touch of professionals in order to lead engineers such as Casey toward unrivalled accuracy.
The simulation engineers at Slightly Mad Studios need to get it all right in order to get a car to behave in-game as it would in real-life—so accurate that many racing drivers use SMS’s simulation technology to test for their real-world races. Here’s how that process works …
BUILDING A CAR
A good starting point for building any car is to look at the suspension geometry. With modern cars, we often receive detailed CAD models from our automotive partners. Pickup points from these models can usually be applied directly into our dynamic suspension model to be sure that things such as caster angle, kingpin inclination, roll centers, etc., and changes of these with respect to suspension travel, all match the behavior of the real car.
The steering geometry here also plays a direct role in calculating forces which come through on FFB steering wheels, giving drivers a ‘feel’ for how the car is reacting, so this a key area of focus and a main driver in what makes an RWD prototype have different cornering feel to an FWD touring car.
This data often comes from the manufacturer. Things like homologation forms for modern GT3 cars include a vast amount of information for spring rates, damper force vs. velocity plots, and anti-roll bar dimensions. These values generally plug directly into our model or can easily be approximated with best-fit curves to match the real car over its range of setup options.
The most important aspects of the chassis itself are mass, center of mass, and moments of inertia on each axis. The former pairing is relatively simple to measure on a real car; moment of inertia (MOI), on the other hand, is a much more complicated affair.
To measure it accurately requires a large turntable and setting up a car with no fluids and rigid suspension to remove dynamic effects—all of this along with a great deal of test rig calibration in order to produce accurate values.
It’s quite an ordeal, and although we have received this kind of data on occasion, it is very rare. A much more common approach, and the one we use, is to break the car down into seven or eight component boxes such that their size and mass create a reasonable approximation for the known specifications and dimensions of the real car. The result from this approach can be remarkably close to MOI measurements from a real car.
Other chassis elements include aspects such as the fuel-tank size and position, which also play a part in vehicle handling. Modern race car designers put a premium on locating the fuel tank as low as they can and as near to the Centre of Gravity as possible so that handling does not change significantly as fuel-load burns off.
But race car engineering is not an exact science, of course, which is why you will also find superb racers such as the Aston Martin DBR1 fitting their fuel-tank right out at the other end of the spectrum—a 182L tank hung way out behind and above the rear axle. Weight distribution therefore changes all the way from 40:60 to 50:50 in this car, depending on fuel load, so the chassis dynamics will evolve significantly over a long run. Getting this kind of accuracy in our simulated cars is vital—not only for us, but also for our manufacturer partners.
Engine & Gearbox
Dyno plots are the gold standard for engine model creation. We use them whenever possible to match the torque curve shapes, and we then add to that minor calibration for known power levels (because all dynos have some calibration factor in-built and are more about ‘torque curve shape’ and relative differences rather than absolutes).
Our engine model works on a volumetric system—amount of air flowing through it—so setting up the intake system is the next step in making sure throttle response is correct, while also pay- ing close attention to the effect of things like engine air restric- tors being correct.
Turbocharged engines get a boost pressure curve mapped over the full engine RPM-range, which is especially important for modern race cars as balance of performance is largely done through controlling how much boost turbo cars can use at various engine RPM. This ensures both power level and throttle response is correct—something that a few of our drivers are not too thrilled with when we accurately model the turbo-lag inherent in some of the early 1970s monsters!
Our hybrid power unit system is essentially adding electric motors in parallel with the ICE engine. We give it basic specs such as maximum torque and power output, storage capacity, recharging properties, etc., and then hook it into the drivetrain to assist in driving the car.
The modern LMP1H field are a challenge in this respect, mainly because so much of the tech’ is highly-secretive, but the regulations dictate enough detail (max power output, total energy per lap at Le Mans, for example) that it is possible to observe on- board telemetry and reverse-engineer the systems so our game model is as accurate as possible. This actually does bring up a point about our working relationships with auto-manufacturers: they’re happy to share detailed information, but some racing technology will always remain a secret, sometimes even years after a car has been retired from competition.
Our driveline model is a modular system where each element of the car, from engine crankshaft through to the wheels, is assembled from a range of building blocks: clutches, differentials, bearings, gearboxes, brakes, couplings, shafts and so on to link everything together.
Each component of each car is tuned for inertia, stiffness and damping properties, and special functions such as the various types of differential (clutch & ramp, geared torque biasing, viscous, locking). This, for a ‘simple’ car, might be modeled as:
A complicated AWD car might send the GEARBOX output through center differential, one shaft of the center differential’s output going to another differential on the front axle while the other shaft goes through a handbrake-activated disconnect before reach- ing a third differential on the rear axle. Our system allows for a completely arbitrary design of the driveline models using as few or as many of those building blocks as is necessary to connect the engine to the wheels. This lets us create accurate models for everything—from your average FWD hatchback to the 1968 Lotus 56 AWD turbine-powered IndyCar to the Ford Bronco ‘Brocky’ with a transfer-case 4WD system which can toggle between 4WD and RWD modes.
When we’re really lucky, a car’s reference data will include a complete aero’ map showing how it responds to ride height and setup changes for drag, downforce, and center of pressure. The task then is one of matching the behavior in our system which is composed of 8 individual aero’ elements, each with unique response to setup changes, ride height, chassis rake, yaw effects, and non-forward motion.
Cars usually begin their development life on ‘donor’ tyres from a similar car. As other elements fall into place, focus shifts to the tyre construction so the carcass handles vertical and cornering loads appropriately. We monitor deflection in the model to adjust the sidewall and tread construction so the magnitude of translational and torsional deflection is in the correct ballpark and occurs in a way that suits the tires being modeled—radial, cross-ply, or some hybrid in-between. The rubber compound is then tuned for adhesion, thermal, and wear properties to suit what is used on the real car. This process then iterates to finer detail until the performance aspects all match the real data as close as possible.