Interview Of Valor
We talk with Andy Taylor of 2015 about the AI in Men of Valor
Tags: Men Of Valor Categories: Features, PC Features, Xbox Features
Posted by Daniel "monk" Pelfrey on Aug 6th, 2004
I had the opportunity to exchange words with Andy Taylor from 2015, where the studio is working on Men of Valor to be published by Vivendi Universal later this year for the Xbox and PC. I could introduce him by listing off a bunch of credentials trying to impress readers with the experience this man brings to the project, but why not let him do it himself…
So, Andy, tell me a little bit about yourself.
 |
My first game was published 20 years ago in the heady days of the 8-bit home computers. I worked for a number of companies freelance until around 15 years ago when I finally came in house. I’ve worked for companies including Eidos, (then Domark Software), SCI and Sega in the UK, Spain, Germany, Hong Kong and briefly Japan. Around 8 years ago I finally made the move to the states and have since worked for Bethesda Softworks, Legend Entertainment, Pandemic Studios and now 2015. I have over 30 published titles on almost every platform from the last 20 years, including the most recent crop of consoles. I’ve been working as a lead programmer for around 10 years, and also as Technical Director at Pandemic. I started to specialize in game AI about the time I moved to the states.
When did you move to the states and why?
I moved here almost 9 years ago from the UK. I’ve always enjoyed traveling and working abroad, I’ve worked in Spain, Germany, Japan and even Honk Kong for 6 months. An opportunity opened up in the states so I took it. What I didn’t realize then is that I’d meet my future wife and “settle down” here. I haven’t left the US yet and I’m quite happy staying put.
Tell me about the project you’re working on now – Men of Valor.
 |
It’s about a grunt called Shepard who is basically just a “regular Joe” trying to survive Vietnam. The story weaves through most of the major conflicts in Vietnam in order to cover as much ground as we could; in reality it would be very hard for a soldier to have taken part in all the operations we cover, (if not impossible). We really wanted to do justice to the scope of the Vietnam War though; you have to remember that the conflict was almost twice as long as WWII so a lot happened over the years.
 |
So the story follows Shepard and his squad through Vietnam. Most of your squad follows you across levels, each of which has specific “personalities” in-game, and of course in the cut-scenes we re-enforce them as ‘real people’ as it were with difference backgrounds and accents. There are a few deaths, I don’t want to spoil things but as always in war Shep loses someone close to the squad at the end of the first op in a poignant scene. Some missions you’ll do with the whole squad, sometimes the squad is cut into smaller teams and you’ll go off to do something with perhaps one other person while the rest do something else. The game is set across as many scenarios as we possibly can, so you’ll be fighting in dense jungle, urban settings, the grand palaces at the end of the War as well as the wide spaces of the paddy fields. You’ll also get to fight on a number of vehicles; APC’s are used extensively for transport to a lot of levels, (which gives us a nice natural break to tell some more of the story), and on other levels you’ll actually fight with the vehicle. For example we have missions on board a patrol boat, on a jeep, as a helicopter gunner as well as in-game insertions an extractions by chopper, tanks, F-4’s, (although you don’t get to fly one of those), Mule’s, Ontos and so on.
You mention that the Vietnam War was approximately twice as long as WWII – how else are they different, and what does that mean for the game?
 |
What the team decided to do early on in the game’s design was portray the war through the eyes of a regular Joe. It would have been quite easy to focus on the political aspects of the campaign, (as some of the Vietnam games have recently), the geo-political stage at the time, American foreign policy, the domestic violence that surrounded the war etc. But that aspect of the conflict was really hammered out in the many, many movies about the war and it just gets in the way of the game. We wanted MoV to be less about the shock and controversy surrounding Vietnam and treat it in much the same way as WWII has been covered in games. We also wanted to cover the team/squad aspects of the war. Because of the difficulties of fighting in the jungle, plus the guerilla-like tactics employed by the VC, perhaps more than any other modern conflict you had to rely on your comrades for support. You needed to cover each others backs, and be sure your squad mates were close by. You also needed to listen to each other and watch for hand-signals. So much of Vietnam was fought behind cover and in almost complete silence; it was a real departure from previous theaters.
 |
Another thing that set Vietnam apart was the stop-go-stop nature of combat. Often operations consisted of being ambushed, or setting up an ambush; you rarely knew ahead of time where the enemy was and the ‘front-lines’ were very fluid. That’s one of the things I think we’ve managed to cover pretty well in MoV. There are levels were you might walk through dense jungle searching for the enemy, but all you do is scare a flock of birds or find the odd booby trap – then, all of a sudden all hell breaks lose as mortars and heavy machine gun fire rips the jungle apart around you as you dive for cover.
What is your job at the studio?
 |
I’m Chief Programming Idiot, it’s my job to the biggest idiot at 2015 – but I’m also in charge of all things AI here. I was brought on board last year with another AI expert Eric Dysand. Unfortunately, Eric died a few months ago which was a great shock to us all. He was a very clever and generous guy, everyone liked him – a big proponent of GDC, he chaired many of the AI round tables there and wrote lots of articles for the Gems/AI Wisdom series. 2015 is really committed to AI; they want to lead the charge of a new revolution in AI over the coming years. Eric and myself were hired to architect the AI Engine the company will use after MoV for all of 2015’s games. I can’t talk too much about what we’re planning, but our future AI system really is revolutionary in design. In the meantime though, 2015 had to let a few people go during MoV’s development and the AI for the game was in a bit of a mess to tell the truth. The company hired Eric and myself late in development, (I’ve been here less than a year), and we basically completed a 50% re-write of the entire AI system in a matter of months. I think we really turned things around.
Can you give a good example?
 |
Sure. We believe that just like a graphics engine, the industry should also apply the same paradigm to AI technology. So much of the code we use for AI, especially for the simpler stuff such as path-finding, object and mapping systems etc. can be reused and like rendering technology 10 years ago, isn’t being reused across games. I’ll be architecting our ‘AI engine’. As Eric’s no longer with us, I’m also building the team to do that work with me. I’ll also be the principle AI guy on one of our projects after MoV, (as anyone on the AI engine team will be), so it’s not as if I’ll be disappearing into tool/architecture land. We’re actually going to architect 50%-75% of the engine during development of our next project here, so the technology won’t be completely finished until the game after MoV is published.
You mentioned that you’ve rewritten much of the AI code, and that you will be assembling a tool set to be used in future games. Where does Men of Valor fit in?
 |
Men of Valor really represents the ultimate in the more ’scripted’ approach to AI. There is a lot of autonomic AI in there, don’t get me wrong, but the game is ultimately driven by scripted objectives, story elements and the agents are basically “nudged” every now and again to keep things rolling. In terms of what we want to do next, MoV doesn’t really fit in. Our future AI is designed along “fluid” lines; that means much more autonomic, motivationally driven, much more natural. Whereas MoV has really tight story elements and works like an epic war movie, next we want to do something that evolves more naturally. Group behavior is a big consideration – not just in squad behavior for example, the agents in the new engine will naturally communicate and share information between each other, organizing themselves according to overall as well as localized strategy. We’re also heading towards the fully independent agent who doesn’t need to be ‘nudged’ by a scripting or objective based system. In the new system the designers will tailor the properties and personality of the agent, give him and then balance a set of motivations and then they let him loose in the world. From frame 1 until the game is won, lost or the agent is killed he lives and works in the world completely independently. If I had to sum it up, think about it this way; we want to put the most fluid and human like Quake/Unreal bots into a single player game.
How have you seen AI change in the last 10 years? In the last 2?
 |
Over the last 10 years AI has changed tremendously; not just in the games industry but in the academic field too. Most of that change has been spurred on by more processing power – and I’m sorry to say this, but hardware more than innovation in games. A lot of the academic principles usable in games used to be impossible simply because the hardware wasn’t faster enough, but that limitation has drastically fallen away over the last decade. As well as the academic field there has been some fantastic work done in the games industry too, almost all genres pushed forward new methods during the 90’s. I think the last decade has at last brought an era of true information sharing too, and I’m not just talking about the internet. Books like the Game Gems series, AI wisdom and many others by authors who “really” work in the industry exploded in the 90’s, they’ve all added to a bigger sense of research, application and sharing of results in the field. It’s great to come up with new methods of doing things, but you need the backup of the community behind you to work out the pitfalls and the advantages of new code. Although we all strive to break new ground, most of the stuff we do has at least been touched before in some other published game. That’s where the internet has become valuable. There are now a number of good forums/message boards around where programmers talk openly about the things they’ve done, things that worked, which potential disasters you should steer clear from etc.
In the last two years? I’m sorry to say that I think most recently things have started to stagnate once more. There have been a few exceptions, but overall new advances in game AI have been few and far between in the last few years. That’s part of the reason behind 2015’s push to further the field in its future games.
How does that apply to Men of Valor?
 |
Eeerm, well, not as much as I’d like. One of the reasons I believe things have stayed still in the industry is the more ‘scripted’ approach recent games have taken. MoV was designed as probably one of the very best of those types of game – probably the farthest in the genre you can go. MoV and other games like it are designed to be story and objective driven – that doesn’t always suit more autonomic AI. With that in mind, there are some tasters of what’s to come in MoV. We’ve certainly got some eye-opening AI in there, but it’s tightly controlled so things don’t get in the way of the story and scripted set pieces.
 |
The friendlies for example are almost entirely autonomous. They’re given prompts now and again to head them towards the next objective or a specific cut-scene, but on the whole they’re free to think and act on their own. They constantly communicate with each other behind the scenes, sharing common information such as who they’re fighting, how well they’re doing etc. They’ll position themselves with the rest of the squad in mind, not only finding the best cover for themselves, but the best place to attack from that gives the whole squad an advantage. They’ll lay down covering fire for each other, defend each others positions, advance and retreat depending on the current situation, and they’ll also attempt to flank positions that pin-down others in their squad.
 |
Of course a lot of that autonomous behavior wasn’t expected in the original design – if you make the friendlies too intelligent, the enemies become too dumb. Instead of dumbing down the friendlies, Eric and myself revisited the VC AI and tailored as much autonomic behavior as we could that didn’t break the flow of the story or the scripted objectives. We came up with a system late in development where enemies between scripted points or objectives could be allowed a lot more freedom than those nearer to them. For long stretches of a level that allowed the LD’s to drop in more really difficult guys that think and perform much of the same complex group logic the friendlies do. The VC flanking and advance AI is intentionally a lot stronger than the friendlies to really challenge the player. They constantly monitor their own position and cover in relation to their surrounding comrades. They look at the size of their squad and the number of recent deaths to decide whether to advance or attack. They also analyze the positions and average out the movement of all friendlies, (including the player), in real-time. That means that they actually formulate a “front-line” of attack for your squad’s positions and know the rough direction the entire group is moving in. That was tough to do without the usual cheating, (by looking where the next objective is for example). Once we had that information, we could get the VC to flank left and right of your positions, and when the attack was repelled, retreat realistically away from friendly positions.
As I said, this is all in the context of a story driven game in the traditions of the original MoH, so it’s probably the best compromise of both worlds; scripted versus autonomic.
2015 has done two expansion packs – this is the first stand alone, and original game the studio has worked on. What scripting from previous 2015 games are you using and how have you changed them for Men of Valor?
 |
MoV was built 100% from the ground up. It was almost 25% re-written again when Eirc and myself came on board. For a start, MoV is the first 2015 game to use the Unreal engine so none of the code from previous games was applicable. There’s also a really strong sense here at 2015 that we want to continually break new ground. MoV to some degree continues in the same vain as MoH because that’s what the publishers expected of us. But, we still injected as much new stuff as we could into MoV and I’ll think people will be surprised at the results. I believe that we’ve taken this scripted style, story driven game to the absolute limits of where anyone could possibly go. After two games like this however, the whole team is looking forward to doing something different.
There have been a lot of military squad based styled first person shooters lately, though most take place during WWII. How is the Vietnam War different from a programmer’s perspective?
 |
From a programmers perspective, it’s a pretty unique conflict. The graphics guys had a lot of headaches with the rendering technology. They ended up heavily modifying almost every core system of the Unreal engine. For a start, in order to purvey the closeness and denseness of the jungle you need to push a lot of polygons. Not only that, but the type of art you’re rendering for a Vietnam game is almost entirely irregular shapes – trees, palms, shrubs, ferns, versus WWII which, granted has a lot of rubble lying around, but is mostly nice rectangular buildings. The collision requirements were also upped, especially LOS collision. It’s impractical to have the player or bots collide with a lot of the underlying vegetation, but you still need to perform a lot of LOS collisions through that stuff; it’s not good if the VC can ‘see’ you through a lot of dense foliage.
 |
For the AI, cover was a massive problem area. Vietnam was all about cover; hiding from attack, maintaining cover while moving, using cover while advancing. We had to come up with a system that maps and analyzes in real time 1000’s of cover locations per level. It all adds up to a lot of processing. One of the most obvious areas was in choosing safe cover for fallback and healing. We decided early on to have the friendlies pick up fallen comrades, throw them over their back and carry them to a safer cover location so they can heal. It looks great to see the guys dynamically doing that stuff, but the whole illusion is shattered when the wounded soldier is plonked down in full view of a VC mounted machine gun. It took us a long time to get that stuff looking right, and not slowing the game down.
Certain sections of the game seem to move along at a crawl, while at other times things happen so quickly that it’s near impossible to tell what’s going on. What tricks do you use to adjust the pacing of the game?
 |
One of the major systems in the game is the objectives system. Basically the LD’s can specify almost anything to spawn, trigger, start-up or shut-down at any place in the game. They can make the player mount/dismount vehicles, start enemy movements and ambushes, trigger air-strikes, music, explosions, etc. etc. The pacing of the levels was taken very seriously by the designers. Each level has been played through a million times, refined, replayed and refined again. The programming team has worked closely with the LD’s, supplying them with every piece of code and functionality they needed in their levels no questions asked. The LD’s also have ownership of several levels each and that’s really all they concentrate on. With that amount of focused attention, replaying and refining I think the design just worked out. The level designers here put in a massive amount of work over the last few years.
We also adopted a very tight approach to level design. Any level that didn’t work was totally scrapped and restarted. We must have completely re-done 50% of the levels at least once and some of them three times or more. The other 50% have gone through so many revisions and refinements they’re really nicely tuned I feel.
 |
The frantic explosions of action, the ambushes and the sense at times that you don’t know what the hell is going on is exactly what we wanted to portray. It’s nice to see that all that work wasn’t in vain. The real Vietnam War was exactly as you described it – a patrol would move along a crawl, and then suddenly the whole jungle would just erupt. It was only in the latter part of the war when it spread to urban areas that the VC couldn’t fall back on gorilla tactics. Their key strategy in-country was the ambush. They didn’t have the technology or the air support the US had, but they did have a lifetime of knowledge about the jungle and they used to that in every way they possibly could.
You mentioned earlier that the team is looking forward to “something different” – care to elaborate? Can you tell me anything about the next project 2015 has?
 |
Well, I’m a new comer here but the general consensus seems to be that everyone is tired of war games. MoH was a big, big project that the guys poured their hearts into, and then MoV which basically took as much heartbreak. I think the team is worried that yet another modern close combat game will be less of a labor of love. As I said though, we’re still looking at a few different design ideas and at least one of those is another war game. The problem is that the team is obviously now really good at these types of game. On one hand we’d like to try something different, but on the other it’s temping to carry on and see just how far we could take the genre.
We’re tossing around a few ideas, so no I can’t be very specific at all. The great thing about working for an independent developer like 2015 is that we get to work on original projects. Really, the world is on a plate as it were. There are a few ideas surfacing for another war game but really the team is looking to do something different this time around. You’ll be the first to know.
Thank you for your time.
[ Post the first comment | View related posts ]
Tags: Men Of Valor
Posted by Daniel "monk" Pelfrey on Aug 6th, 2004 and is filed under Features, PC Features, Xbox Features. You can follow any responses to this entry through the RSS 2.0 feed. You can post a comment, or trackback from your own site.
