The Driving Simulator
Monday, November 08 2004 @ 05:39 PM EST
Contributed by: thelusiv
The Driving Simulator is a project of the Psychology Department at Clemson University, headed by Dr. Fred Switzer. On Friday October 29, I interviewed Dr. Switzer and test drove the simulation. You can see the pictures I took, and read the interview and my impressions of the driving simulator below.
Chris Guirl: First off, could you tell me a little about yourself, the project and its goals?
Dr. Fred Switzer: Absolutely. I'm an industrial organizational psychologist by training, which means I look at human behavior and cognition in the workplace. When I was in graduate school I worked for the Aviation Research Lab at the University of Illinois, so I've had some training and interest in human factors kinds of issues too.
Long story very short, about three years ago, I was looking around [and] we didn't have any kind of aviation simulation facilities here. Turns out we've gotten a little of that since then, but at the time we didn't. I was interested in doing some work on human judgment, decision-making, and risk perception and things like that. Talking to a couple of my colleagues downstairs in human factors, it turns out they had been doing some work in automotive related stuff, and so I had the bright idea that maybe if we got ourselves a driving simulator, maybe we could get some multiple research lines going on this thing.
So we found some funding, and then scouted around for a research-level driving simulator. We looked at lots of stuff and finally found this one vendor. There are several research efforts around the country using this particular simulator - University of Illinois for example just got a $6 million grant from GM, and this is the simulator they bought (or an updated version of ours).
We didn't have that kind of budget, so we basically talked the vendor into selling us sort of a do-it-yourself kit. [laughs] So what happened was, they sent us a whole bunch of parts. Basically they sent us the computers and the software, and we put everything together. For a car, we found [one] in a junk yard. Over one Christmas break I stripped it down to the chassis in my driveway (my neighbors loved that). [laughs] Then I cut it up with a saw, and we (meaning the Psychology Department faculty) loaded it into the back of my pickup, drove it up here, carried it up the stairs piece-by-piece and bolted it back together. [laughs] So anyways, now we have a 1989 Mitsubishi Galant on the third floor of Brackett Hall.
It's a really nice sim - it's got a 150° field of forward view and a 50° rearward view so you can see the cars behind you in the rearview mirrors and all that good stuff. It's OpenGL and Linux-based. It basically consists of at least six or seven local-area-networked computers. The actual simulation software itself is all on the Linux machines, but there is a Windows machine where we do the scenario authoring. What makes this different from say, Gran Tourismo 3, is that it's completely user-programmable. 
It's got a drag-and-drop system to create the virtual world, including all kinds of traffic scenarios, roadways, [and] visibility levels; you can implement fog and things like that. Every single vehicle, every entity period - vehicles, pedestrians, and even a dog and a deer - are user programmable. That, and the fact that every variable that you can imagine, that you can measure in a sim - everything from latitudinal and longitudinal accelerations, position of the brake pedal - you name it, that thing measures it.
We've done multiple research projects on this - everything from cell phone studies (distractability studies), to studies of car following in fog - over the last three years. Right now we're working on several projects, one of which has to do with testosterone in male drivers. We're doing that kind of stuff, and we're doing the effects of sleep deprivation on driving performance, just finished a study on that. Right now there's a study that's running using the in-vehicle information system - onboard maps, and things like that - we have a generic in-vehicle screen that we've done various kinds of studies on, looking at the effects of in-car navigation systems, and other kinds of driver notifications and things like that. That's running Delphi 6 on a Windows machine. We've got the biggest mix of platforms, programming environments, you name it - it's just kind of a real - I don't know - mongrel. [laughs] But it works, and that's what we care about.
CG: That's not as uncommon as you might think.
FS: Oh is it? OK, good, I feel better - I figured when you have this homemade system, that's kind of what you end up with - so it's kind of like Frankenstein's monster, but it does live! [laughs] And that's what counts. We're getting some good research data out of it, I mean it's really turned out to be a great research tool.
CG: What made you choose to use Linux?
FS: Well, the vendor we got it from. Basically, that was where that started, but we've been really happy about that since then. I know why they chose Linux - it doesn't have the kind of limitations we would've run into with a Windows machine. We've had to upgrade this thing several times. It's got to be very flexible - because the vendor doesn't know what kind of research questions we'll be investigating with this thing, so it needs kind of ultimate flexibility. I know that's why they did it, and it's worked out great for us too, so we're happy about it.
CG: What is the name of the software you use?
FS: The software is currently called HyperDrive, by a company who's changed their name several times. Currently they are called Drive Safety and they're based out of Salt Lake City, Utah. Originally they were a group called Hyperion Technologies, which I think was started by some folks who had been doing research at the University of Iowa. Iowa was an early leader in drive sim research, in fact the National Advanced Driving Simulator, which is this multi-million dollar mass of facilities right outside their campus.
The folks who did the driving simulator company got bought up by another company that actually did crane simulations to train operators. The branch we deal with is the research simulator branch. There are probably a grand total of about ten, maybe fewer, simulators of this kind in North America. I do know that they used a spinoff of their system to make driver training systems for the state of Hawai'i, so some of our landscapes are actually Hawai'ian, [laughs] some of the billboards say "Bank of Hawai'i" and stuff like that.
CG: What are some of the biggest problems you've run into and how did you solve them?
FS: Our biggest problems are actually hardware problems. We've had, it seems to me, an unusual number of hard drive failures. Of course, that brings everything to a screeching halt, while we grab a new hard drive. We never can tell which channel of the system - it's a five-channel system - we never know ahead of time which one's going to fail, and we don't have five hard drives formatted, so we have to load all the software on to the hard drive (which takes a while), then replace it, so that usually brings the sim down for a day, so that's been kind of a pain.
Other than that it's been really reliable. We've had a few gripes about the licensing system the vendor uses. For a while there they used a timed key, so if you didn't get your software key updated, the visual systems would go down. They never seemed to get us the key - they were trying to do it in the nick of time, and they never quite got it! [laughs] So we'd spend a day or two with the system down while we get on the horn to get the new key.
CG: What are your future plans?
FS: Basically, we want to upgrade this thing. We want to get a more sophisticated force feedback system. I'd like a little bit better sound system, at least a shaker. We've got a subwoofer on it right now, but a shaker bolted to the chassis would be pie in the sky.
At this point my laptop's battery died, ending my recording. Fred went on to mention that he'd like to try TORCS, The Open Race Car Simulator, on the system sometime. I thanked him for his time, and we headed to the lab so I could test drive the sim.
Before starting, I looked around at the equipment a little. The car itself was just as he said, an automatic '89 Galant. Under the hood one finds nothing one would expect in a regular car. Missing are the engine, transmission, gas tank, all the usual tubes, wires and other components. Instead there is a lot of sensor wiring and circuit boxes, a subwoofer that simulates engine noise, a few weights hanging from the steering column and the pedals, and the cinderblocks on which the car body rests.
There are four projected 800x600 video screens around the car - three in front and one behind. Each one represents a different camera angle. Inside the car, there is a movable touchscreen LCD connected to a computer running Windows and some Delphi applications that Psychology researchers wrote.
The visual system is driven by several drawer-mounted RedHat Linux 7.x systems. One was open because a hard drive had just been replaced. There is also a "master" system (also running Linux) which orchestrates the other four while the simulation is running. The scenario authoring is done on a Windows machine. When the simulation starts, it sends scenario data over the network to the Linux machines. Another computer, also running Windows, monitors a sensor in the car that checks to see if the drivers' eyes are open or not.
After checking everything out I got in the car to start my test drive. Dr. Switzer loaded a scenario and I started driving. The first thing I noticed was that the steering wheel was somewhat loose, as were the brake and gas pedals. The steering wheel did return to the center when turned too far, however. As soon as I got up to about 60 MPH, I came up on a group of police cars, so I slammed on the brakes - a little too hard. I didn't hit anything, but my car spun out of control into the grass, and I narrowly missed sliding into a monstrous ravine that ran next to the road. The police didn't seem to care, however, so I turned around and got back on the road.
Other drivers rudely tailgated me, and tapping the brake didn't help any. A dog ran out in front of my car. Another driver almost ran me off the road when he came into my lane. Then, Dr. Switzer changed to another scenario where I was driving through fog. I couldn't see very far ahead, but I managed to stay on the road this time. I also noticed that if I went over the speed limit by five miles per hour or more, a voice would suddenly remind me, "Slow down. Slow down. Slow down."
It felt at moments as if I was actually moving in the car - it fooled my sense of balance because the video filled most of my field of vision, and the rest of the room was dark. There was a whooshing noise the other cars made when they went by, and a rough approximation of my RPM rumbled from the subwoofer when I pressed the gas. Otherwise, there was not much sound. Another thing that made it feel slightly less authentic was lack of seatbelts.
The car itself was no racer - it felt slow and heavy. I'm not sure how much the performance of the car can be changed in the simulation. Most of the important mechanisms for driving the car worked within the simulation - the brake and gas pedal, the steering wheel, the tachometer and the speedometer. The turn signals, lights, handbrake, transmission selector, and windshield wipers did not work. I wonder if they could be used, if the simulation software supports them.
The steering wheel and pedals all felt very natural, since they were actual car parts, but they lacked the resistance usually present in a car, since all the mechanics those parts are usually attached to are missing. One suggestion I'd like to make is that rather than (or perhaps in addition to) using weights, the pedals could have springs attached to them, which could provide a much more realistic feel. A spring may not work for the steering wheel, however - the only thing I can think of that would make it feel more realistic would be to have the steering column work a hydraulic pump. This would be very much like a power steering system, however it might be somewhat difficult to get working just right, and would introduce more chance for breakdown.
All in all, I thought the project was a good one. The simulator was fairly realistic and while it has lots of room for improvement, it seems like it has most of the things the researchers need. They seem to use it pretty heavily - before I went in, there was someone testing, and more were scheduled after me. Most of all this project is impressive because since the researchers didn't have a lot of money, they built it themselves. The projector stands, the screens, the supports for the car chassis, the car itself, and probably countless other parts were all put together by hand. This, to me, is better than spending millions of dollars on a fancy simulator, even if it's not quite as good - it represents the hard work and ingenuity of the researchers involved. They had to use what they could get for a relatively small amount of money, and this usually means that they put more thought into it than they might have otherwise. The finished product may not be the prettiest, but it works quite well and they are happy with it, and that's what matters most.
Before leaving, I told Dr. Switzer that if he ever had any trouble with his Linux systems to drop me a line, and I would try to get him some help through CLUG. He said he probably would have called us in the past, if he had known about us. Likewise, I would have checked out the driving simulator long ago, had I known about it. There are a lot of other neat projects headed by Clemson faculty and students that get very little attention - in fact, so little that many people who have been here for a long time have no idea they even exist. This was the case with me and the driving simulator. If you know about any other interesting projects which involve Linux at Clemson please email me.
Don't forget to check out the rest of the pictures...
8 comments
http://clemsonlinux.org/article.php/20041108173932895