High-fidelity fakes give armchair racers quality seat time.
Imagine going head to head with some of NASCAR's finest and 11 of your best friends in a simulated race that's being fed live action. This is the vision of Bill Donaldson and his company NASCAR Silicon Motor Speedway, Indianapolis, and one that could soon be reality. Donaldson became interested in NASCAR simulation, having spent 19 years with Indianapolis Motor Speedway where he helped organize the NASCAR Brickyard 400.
Of course, race simulation, even at this level of sophistication, is only as good as the simulator itself. "Entertainment simulator" may bring to mind balky amusement park rides that excel at jerking occupants about but fall short on mirroring reality. Not so with the SMS NASCAR device. It closely models the underlying physics of a race car navigating banked turns at high speed and faithfully transfers those sensations to the driver.
Then start-up company Silicon Entertainment later acquired by SMS tapped electrical engineer and former Atariarcadegame designer, Rick Moncrief, for the simulator design. Moncrief is the creator of titles including Asteroids, Lunar Lander, and the world's first driving simulation game, Hard Drivin'. For the NASCAR simulator, Moncrief literally began with a blank sheet of paper and figured simulator-system development costs at $2.88 million. He got only $1.75 million by the time the product launched but still managed to deliver the simulator design on schedule for the target retail of $95,000/copy. Elan Motorsport Technologies Inc., Braselton, Ga., maker of Panoz G-Force Indy Racing League cars, builds the simulators, which are about two-thirds the size of a NASCAR racer. Moncrief is currently putting finishing touches on a scaled-down version called SMS Reactor. It will retail for about $35,000 and is being contracted to a yet unnamed manufacturer.
Like any engineered product, the SMS NASCAR simulator is a series of compromises. But one uncompromising aspect underpinning the design was to keep the driving experience as real as possible.
A big problem with automotive simulation is how to produce relatively smooth motion reversals along a single axis. Real automobiles generally impart to their occupants smooth accelerations coincident with direction changes. This ruled out low-cost, arcade-type motion technology that relies on electric-motordriven gear trains. Motion reversals in such systems are unnaturally abrupt.
At the other end of the spectrum were hexapod platforms, ala aircraft simulators. These incorporate special (and prohibitively expensive) hydraulic cylinders with calibrated leak rates. Electromagnetic actuators were also candidates but discounted, again because of cost and their inability to accurately deliver the range of forces needed for the job. And heat buildup can be a problem under high-duty cycles.
This pointed to conventional hydraulic systems. But four-way valves for controlling reversible flow to the cylinders have difficulty at the crossover point between lift and lower modes. A valve spool transitioning between modes causes a pressure spike ("clunk") at zero flow or a sink if the spool is a make-beforebreak type. These characteristics prevent four-way valves from performing uniformly under widely varying loads. A 75-lb person would get a markedly different ride than, say, a 200-lb person, which was unacceptable. It turns out that simple proportional valves, applied in a novel way, give the desired result. A single-acting cylinder on the acceleration sled's heave or lift axis, for example, receives a pair of series-connected proportional valves. One valve goes between the hydraulic pump and cylinder inlet, the other between the cylinder inlet and oil storage tank. Similarly, the double-acting cylinder that drives the sway axis receives two pairs of series-connected proportional valves.
In operation, the patented control system compares commanded position with the actual position of hydraulic actuators based on encoder feedback. The resulting error signal feeds to a conditioning circuit that outputs a pair of valve-drive signals and what is called a quiescent drive signal. It is this quiescent drive signal that develops, or nearly develops, a quiescent fluid flow through a valve pair.
"The signals automatically adjust the spool valves so they are in position to drive the acceleration sled such that it gently 'floats' as would an actual racecar going over undulations on a track surface," explains Moncrief. "Motion reversals are imperceptible, or nearly so. But the sled also has the dynamic range to deliver a wallop when a driver hits another car or the retaining wall." The hydraulic system/sled mechanicals respond to control inputs in about 50 msec (20 Hz), considered slow by some measures. "Professional race-car drivers are able to sense the delay. They know it's not there," says Moncrief. "For the average person, however, the experience is overwhelming."
But even professional race drivers agree, the SMS simulator is miles ahead of video games in its realism. This is due largely to the accuracy of the mathematical model on which the simulation is based. That model comes from automobile-dynamics gurus at Milliken Research Associates, Buffalo, N.Y.
One of the most difficult aspects of modeling car dynamics is what happens between the tire patch and road, and how the tire sidewalls comply under varying loads. Bill Milliken and his son Doug wrote what is considered the "Bible" on the subject. Moncrief's engineer friend Max Behensky and Doug used the information to create a free-body model for the car. In addition to the tire data are equations describing moments of inertia, shock compliance, caster, and camber. Tire wear, a critically important metric in real racing, was initially included in the simulation but later disabled.
The model calculates tire forces and sends them through virtual steering gear to the simulator steering wheel. A large dc motor, driven by an analog amplifier, connects to the steering wheel and provides force feedback. "When an event in the simulation (free-body model) moves the modeled car, we try to impart only the initial portion of acceleration into the sled and differentiate out the movement," explains Moncrief. "We keep the peak or onset motion cues so the driver feels like they are moving." Surprisingly, axes on the large sleds move just ±6 in. stop to stop. For the Reactor, that number is a mere ±2 in. Both sleds limit accelerations to 0.25 g, though the rate of change of acceleration or jerk can be quite high. Moving a driver in the simulator is a side effect, not the goal.
Also modeled are the inevitable collisions between cars and the retaining walls. Car-to-car collisions, for instance, are handled by networking together all of the cars and determining which ones are in contact and where, and how "deep" a contact is. A Dynamic Force Balance (DFB) computer runs a physical model of the collision and calculates the associated forces and application points. The DFB then sends this information to the free-body models of simulators involved in the crash.
Banked turns require a little trickery. Stop the car on a banked turn and the sled rolls about its long axis in the same direction as the bank, as you'd expect. At racing speed, the sled rolls opposite the bank angle to
simulate centripetal force throwing the car and driver toward the outside of the turn. The display then angles the other way so drivers don't notice the opposing sled roll.
Making the simulation "feel" real was an iterative process. SMS brought in professional race drivers including Michael Waltrip and Jeff Gordon to critique the simulation and point out problems. Fixing it necessitated tweaking of the simulation system's computer code, which is written mostly in C. The code is tens of thousands of lines long and needs lots of computing power to run. For that, Reactor simulators use a single, dual-core Linux PC with a 2.4-GHz processor. This is a big improvement over the large simulators that originally ran 6 DOS PCs per car.
Reactor simulators will use CANbus for all but the most critical inputs such as the steering wheel and actuator-position sensors. Those get special-purpose USB I/O boards. CANbus eliminates some of the wiring harnesses that were needed in the large simulators.
Graphics displays are another mission-critical item. The large simulators initially used three CRT projectors, which worked well but cost about $20,000. Reactor simulators will instead incorporate Digital Light Processing (DLP) displays. Compared with CRTs, DLPs are considerably brighter, have better contrast, take up less space, and cost about one-tenth as much. But DLPs have one drawback: latency. With a CRT, latency defined here as the time it takes a camera to process and project an image on screen is on the order of a few nanoseconds. In contrast, latency for DLPs is measured in the tens of milliseconds.
"The delay, if not properly managed, can make it begin to feel like you're driving a boat instead of a car," says Moncrief.
On the plus side, it is easier to align or register images from the DLP units so they correctly appear as one image. High-end graphics cards adjust to compensate for distortions in projection geometry. The correction process involves using what is called a screen-object grid (basically a 3D object with equidistant spacing). The grid is projected to appear as a piece of grid paper. The grid lines don't converge, as would, say, a set of railroad tracks. A system that is not properly corrected may severely distort the grid spacing. But spacing remains equidistant when the projection is true. Then, simulated and real scenes can look remarkably similar.-In motion, images converge as they should, pixels fill edges in a natural manner, and textures appear lifelike.
Realistic rendering has another important upside: it lessens the chance of "blue flags." Blue flags are when operators have to clean up after a motion-sick driver. A person highly susceptible to simulator sickness may last only a few seconds before feeling queasy in a simulator with an improperly corrected display system. That same person can go 10 min or longer with no negative affects when images are undistorted.
A typical race lasts 5 to 7 min and fields 32 cars, of which up to 12 are human-driven simulators. Computer-driven drones make up the rest of the field and act as pacesetters. Programming of the drones involved first recording the path taken around a track by the best available human simulator driver. A fast "line" is sampled for each simulated track. The drones then follow the fast line, or try to. Just to make it interesting, some of the drones are given a little more tire grip, and others, a little less horsepower. So-called "rules of offset" forbid drones from holding the fast line when a human-driven car assumes that position. Drones use the same DFB model as humandriven cars, so crashes with them seem real.
"The drones appear 'rock solid' and give a dynamic quality and realism to the experience that we believe is unmatched," says Moncrief. Further adding to what Moncrief terms the "suspension of disbelief" is a threshold-of-pain (120-dB) sound system that pumps out the roar of unmuffled V8 engines, along with lap-by-lap commentary.
If all goes as planned SMS will have the Reactor simulators ready for market later this year. The combination of a smaller footprint (30 72 in.) and lower cost should help the company penetrate markets Silicon Entertainment and its larger simulators couldn't. The large simulators still operate in several hightraffic shopping malls, including the Mall of America in Bloomington, Minn. Bill Donaldson sees other venues such as entertainment centers embracing the smaller Reactor, perhaps starting Tuesday-night racing leagues. It's not unreasonable to think it can happen. Research shows over 90% of customers who try the SMS NASCAR simulator five times get hooked.
But the real kicker could be remote racing. The idea isn't new. Online gaming is big business, though only the well heeled among us could afford to shell out five or six figures for a NASCAR simulator; someone like a sheik in the United Arab Emirates capital Abu Dubai, for example. He bought four of the large simulators. But then a neighbor wanted to race too so he bought two more and SMS connected the two palaces with a fiber-optic line.
Eventually, SMS hopes to make its simulators work on a T1 or DSL line, or possibly wireless site to site. "This would let a professional race-car driver like, say, Tony Stewart sit in his basement with a simulator and compete with other simulator players offsite," says Donaldson. Then there is the Holy Grail of racing simulation: "We are now looking at piping telemetry data from NASCAR cars on the track to our racing centers in real time so players can participate in a live race," says Moncrief. SMS' "Wheel to Real" racing could be ready for testing at the NASCAR Speed Park in Myrtle Beach this year.