By the early 1980s, computer technology had advanced far enough in two separate areas that the concept of combining them together usefully became possible for the first time. Those areas were that of simulation (specifically vehicle simulation) and of networking. In the U.S., the Department of Defense was the largest proponent and sponsor of both technologies. Military training was made safer and cheaper through the use of simulators, and all manner of military activity relied on computer-to-computer communication.

The most significant limitation of simulator technologies of the day was that they were primarily useful for either learning procedures, or (at a stretch) for solo combat training. Fighter pilots or single aircraft crews could quite usefully train in sims, but for other branches of the military where combat was a much more integrated team activity, simulators (which up until that time were standalone) weren't much use.

In the mid 1980s, DARPA began to investigate the possibility of networking simulators together to allow different personnel (and, more importantly, separate teams of personnel) to train together. A project which included the simulator units themselves as well as the network connecting them was initiated, and named SIMNET (SIMulation NETwork). Eventually, work on SIMNET was contracted out to three separate firms; DARPA served as the project integrator and coordinator, so no 'prime contractor' was named. Those firms were Delta Graphics, Perceptronics, and Bolt, Beranek & Newman. Delta Graphics was responsible for the then-cutting-edge graphics systems used in the simulators as well as for the terrain models which the simulators would run in. Perceptronics built the physical simulation units themselves. BB&N produced the vehicle simulation software and the networking software and protocols required to link them up, as well as several ancillary software packages which interacted with the simulation network to increase its realism. As an example, BB&N developed a software add-on which produced realistic artillery fires inside the sim.

The actual 'customer' for SIMNET was the United States Army. Eventually, SIMNET contained three types of simulator units: tank simulators (generally modelled on the U.S. M1 Abrams tank), attack helicopter simulators based on the AH-64 Apache, and aircraft simulators which modelled the A-10 Thunderbolt ground attack fighter. Although there were other entities simulated (in one experiment, a docked U.S. Navy destroyer had its Combat Information Center displays 'wired into' SIMNET in order to take part in a scenario) those were the three primary training types which had dedicated simulators built.

The network itself was originally a system of multiplexed 56Kpbs dial-up telephone lines (SIMNET predated the commercial Internet as it's known today). As a military system, it ran on a closed and dedicated set of communication lines. Eventually, when SIMNET was first 'deployed' in 1987, the Army built simulator facilities in five places: Fort Benning, Fort Rucker, Fort Knox, Fort Leavenworth and Grafenwoehr, Germany - all locations with U.S. Army armor units on permanent or temporary station.

SIMNET allowed Army tank crews to train together, practicing platoon and company tactics in each location and allowing multiple companies to fight 'together' in the networked battle simulation no matter which of the five locations they were stationed at. Along with Army helicopter pilots and Air Force A-10 pilots, they could practice for what at the time was the primary mission of the U.S. Army - standing off a Warsaw Pact armored attack on the Central Front. Best of all, they could learn the tactics without risking any lives or expending any fuel, ammunition or parts, and without tearing up big chunks of landscape (tanks are not environmentally friendly, believe you me.)

Although the 'official' simulators were available at the above locations, there were other units. Each of the three contracting firms had units available at their laboratories for development and demonstrations. I was personally familiar with the four units which in 1988 were present at the Bolt, Beranek & Newman facilities in Cambridge, Massachusetts. There was a 'full tank simulator' in the 'front' of the building for demonstrations (by 'full' I mean physically identical to the ones the Army had - fully enclosed duplicates of the vehicle cockpits). There was a prototype tank simulator in the lab - it was enclosed and had the full set of systems, but not all the various props on the inside that made it look like a tank. There were also, at the time, two aircraft simulators which could be set to be either A-10s or AH-64s.

Understand, of course, that to the various computer scientists and technicians responsible for designing and building them, these were no less than the world's best video game. I'll give you an example. The 'full' tank simulator was intended to make actual U.S. Army tank crews 'forget' that they were in a sim. As a consequence, all of the controls and displays in the simulator were duplicates of (or actual instances of) the systems used in real Abrams tanks. The seats, optics, instruments and more were all high-quality duplicates of those used in the tank itself. As part of the design process, Perceptronics had taken high-quality recorders out to the field and recorded every sound that the Abrams tank made, from engine noises to gun loading noises to turret rotation to gunfire. There were no less than (I think) 18 Boston Acoustics speakers embedded in the walls of the simulation compartment (as well as in the seats) and each played a separate dedicated audio channel. Once you got into the simulator and closed the hatches, it was really really hard to tell that you weren't in a tank other than the missing smells of jet fuel and cordite (yes, I've been in a real M1 Abrams while moving; no, I wasn't a crew member).

The sound these simulators made when you took a gun hit was...impressive. Perceptronics ended up having to convince the Army training units to let them turn the sound down, because hearing loss wasn't a desirable outcome. I found out later that the military had been trying to simulate the effects of a Soviet laser blinding weapon on pilots, and their first proposal had been to actually install high-power strobe lights inside the sim's helmet visors. Again, they had to be convinced that actual eye damage was...undesirable.

The graphics weren't much to speak of from a modern point of view. Any PC tank sim from the mid 1990s on probably had better graphics. As moeyz points out to me, there are free-to-play online multiplayer tank games now which have graphics and features that SIMNET would have killed for. However, SIMNET was convincing not only because of the well-appointed interiors. There were four (I believe) vision displays in each tank simulator - the driver's periscope, the gunner's weapon station, the commander's weapon station and the commander's turret ring. There are armored quartz viewports set into the ring around the commander's hatch, so you can 'look out' of the tank by spinning the hatch ring. The SIMNET simulator had a spinning hatch as well, and small displays mounted outside the viewblocks that rotated with the turret. Each simulator was responsible for maintaining 'coordinated' displays for each of those three positions. In addition, the weapon stations could offer magnification, thermal imaging and night vision modes.

One problem with using these as a video game was that there was no provision for short crewing. In order to fight the tank, you needed at least two people - a driver in the driver's compartment and another person at the commander's position. This was only possible because you didn't need to actually haul shells to load the main gun; instead, you hit a control to open the armor plate which protected the ready ammunition behind the main turret, then hit a button on a painted-on 'round' to select it (it might be a HEAT round, an APFSDS round, a smoke round, etc.) and then hit a button on the main gun breech to 'load' it. If you were limber, you could do all three of those from the commander's high chair with your left toe. Once the gun was loaded, you could fire it from either the gunner's position or the commander's position, but if you needed magnification or to use the main ballistic computer, you needed to switch to the gunner's weapon station. The ideal 'team size' when playing around with the sims was three; the buttons meant you didn't need a loader/radio operator, so a driver, gunner and commander could fight the tank.

There were voice intercoms between all the stations in the tank just as on the real vehicle; in addition, there were networked communications systems which simulated the various radio nets the tank could use. These could of course be used from 'radio' stations outside the simulators. There were also control stations - computer terminals outside each simulator - which could communicate with the sims, and could also be used to cause 'failures.' For example, the simulator operator could 'knock out' the tank's radios, or main gun, or turret rotation, or whatever by entering commands on the control console.

Generally, the sim units in the BB&N building were on their own dedicated local network. When I was there (I had a good friend who worked there) there was an informal 'after lunch tank league' which would spend odd hours playing tanks in the simulator. My friend and I, being video gamers of long standing and relatively young compared to the staff with commensurately better reflexes, were used to coming out on top of these informal matches. One day (I recall vividly) we went in to play, and whoever had gotten out to initialize the simulators did something incorrectly - because suddenly those of us in the tanks were looking 'out' of the viewports at dozens of tanks, and listening to the chatter of dozens of transmitters. Our simulators had been linked up to the 'live' SIMNET, and we'd been dropped into the middle of an exercise being run at Fort Knox. Before we could even figure out what was going on, some unknown Army tank crew had decided we weren't friendlies (it's possible that their system was painting us as 'enemies' since we were on a separate network) and we were deafened by the sound of our tank exploding.

We re-initialized the sim and 'ran' for a forest. We didn't make it.

We re-initialized the sim to put us a kilometer or so from the Army units and inside the treeline. The screens stabilized, and four seconds later our tank exploded.

We re-initialized the sim to put us five klicks from the nearest Army tank. We oriented ourselves, found a riverbed, and began sneaking along it towards the trees. Then our tank exploded.

Essentially we learned that yes, when people are trained to drive and fight tanks, it doesn't matter how good a video gamer you are.

The Apache simulator was nearly impossible to fly. One problem of course is that flying a helicopter is much much easier with 'physical' feedback and the aircraft sims at BB&N didn't have motion systems on them - they were just essentially fancy computer terminals. I remember trying to hover - out of my heading, yaw and altitude I could usually keep one or maybe two consistent, but never three. None of the comp sci crew there could do much better. I was told by one of the techs that when they'd gotten the simulators mostly complete, they'd asked the Army to send over some pilots. Ten or twelve Ph.Ds clustered around as two twenty-year-old gum-chewing pilots came in, strapped in, and fired up the simulators. Within ten seconds, the pilots were hovering behind treetops, popping up to engage, flying sideways and backwards down valleys while firing at targets, and performing similarly flabbergasting maneuvers which none of the computer geeks had ever managed to get right. They turned to the techs and said "Hey, this thing flies just like the real deal!"

I was told that while proud of that, several of the techs nevertheless were miffed at the ease with which the pilots mastered their baby.

SIMNET used dead reckoning to handle the motion of units in between network updates. However, at least in the early versions, it didn't correct for terrain or elevation. This meant that when communications between the simulators glitched or dropped out, other tanks would (from the glitched sim's point of view) just keep going in the direction they had been going at the same speed. If they were, however, driving uphill at the time, they might be seen to 'soar' out into the air above the hill crest if the glitch hadn't vanished by the time they reached the top. 'Flying tanks' were a staple joke. There was an attempt to build an add-on for SIMNET called OPFOR (Opposing Force) which would control a number of enemy tanks using artificial intelligence - I watched some tests of that, and OPFOR tanks were very very good at suddenly 'taking flight' whenever the machines running the OPFOR got overloaded. In addition, when communications would resume, the 'flying' tank would suddenly 'teleport' to a new location (its proper location) on the ground. Tanks which had slowed down during the break would 'jump back' to where they were 'really' supposed to be. This exact effect became quite familiar to players of early online FPS games such as Quake, where they were dubbed 'teleporting' and 'rubberbanding'. The basic problem of communications interruption causing position problems was already, at BB&N, known as the infamous 'lag.'

SIMNET was used to rehearse the operations and tactics used in the first Gulf War. By the mid 1990s, however, it had been made obsolete in several areas of technology, and new programs such as the Close Combat Tactical Trainer had taken its place. With ARPANET morphed fully into the Internet (and its 'secure' cousin the SIPRNET) SIMNET's dialup links were replaced with more modern networked simulation systems.

Modern gaming owes at least two major things to SIMNET - the concept of online multiplayer simulation, and the use of Z-buffer real-time graphics techniques to render objects in a 3D environment. And in a classic case of "what goes around, comes around" the United States Army's newest training system DSTS - a networked simulator - in fact is being built by a gaming company, and runs on a popular gaming engine. It looks like SIMNET after being hit in the head by a large supercomputer and a regiment of FPS gamers.

Log in or registerto write something here or to contact authors.