Jan 26

Around the Verse – Engineering Intelligence

In Around the Verse Engineering Intelligence the PU team shares what to expect in 3.0 and beyond.
We also get an update on the MISC Razor, VFX, and a glimpse at upcoming surface outposts.

This post is a transcript of Around the Verse – Engineering Intelligence, material that is the intellectual property of Cloud Imperium Games (CIG) and it’s subsidiaries. INN is a Star Citizen fansite and is not officially affiliated with CIG, but we reprint their materials with permission as a service to the community. INN edits our transcripts for the purpose of making the various show participants easier to understand in writing. Enjoy!

Around the Verse – Engineering Intelligence Full Transcript


CHRIS ROBERTS (CEO, Founder and Director)
SANDI GARDINER (VP of Marketing)

SANDI GARDINER: Hey everyone and welcome to Around the Verse. I’m Sandi Gardiner and with me is Star Citizens’ Game Director, Chris Roberts.

CHRIS ROBERTS: Hey Sandi. I’m so thrilled to be here, of course. I’m heading off to Europe this Sunday, so I’ll be there for a few weeks so you’ll have other hosts here with Sandi going forward for a few weeks until I come back.

SANDI GARDINER: Speaking of Europe…

CHRIS ROBERTS: …Speaking of Europe look what we got from a fan in Switzerland. He’s sent us lots of sugar obviously. Swiss Toblerone, Swiss Caramels, Swiss knife.

SANDI GARDINER: From Zurich, Switzerland…Martin Whip.

CHRIS ROBERTS: Yeah, and sent us a t-shirt from his organization, thank you very much.

SANDI GARDINER: Check this out, bold.

CHRIS ROBERTS: Pretty, I’m not that brave to get a real Star Citizen tattoo on my arm.

SANDI GARDINER: I’m not either, well done.

CHRIS ROBERTS: Well done, very impressive and thank you very much, Martin. Well anyway, back to the show. Today’s show is going to be about something very dear and near to my heart, the game’s AI.

SANDI GARDINER: That’s right we’ve got Tony Zurovec and Francesco to sit down and answer a few questions about how players will be interacting with NPCs in the game, but that will be coming up later in the show.

CHRIS ROBERTS: Yeah, first up we have our weekly studio update from Nick Elms and our U.K. team.

Studio Update

NICK ELMS (Creative Director)

Hi everyone. Welcome to another Studio Update from here in the U.K. My name’s Nick Elms and I’m the Creative Director here at Foundry 42. As you might have imagined since the 2.6 release we’ve been very busy trying to get 2.6.1 out for you. There’s been a lot of bug fixing, a lot of feedback that we’ve taken on board from you guys and John Crewe and Andy Nicholson have been looking at incorporating a balancing pass now with all the feedback we’ve been gathering from the community playing 2.6. They’re making a comprehensive review that hopefully, we’re going to be getting into 2.6.1.

We’re going to look at the MISC Razor next, which I think you saw in November for the concept sale. It’s now with the ship team and, particularly Joe Navel who’s working on it through its white box stage.

As part of the ongoing VFX polish and optimization, we’re going to look at the Drake Caterpillar. We’ve been working on the throttle up and down engine effects and with the Herald we’ve been working on the explosion effects. This is the kind of stuff that we send around the office daily and we thought we would share it would you guys so you can get an idea of what we see. As you’ll see, the VFX guys have shared there’s a roll of all the ship weapons firing simultaneously.

Surface outposts is what we’re going to look at next which I think we’ve released a couple of snapshots of, but it’s starting to come online a lot further now for us, in the form of the building sets that were starting to get through from the artists. The designers are particularly excited about all the gameplay opportunities that these will afford as they can lay them out on the planet surfaces now for you guys to investigate.

That wraps up another U.K. update. As I mentioned earlier there’s a lot of work still going on behind the scenes with bug fixing and optimizations and feedback. All that remains for me to say is: Back to L.A. and see you in the ‘Verse.

In the Studio

CHRIS ROBERTS: Appreciate the update, Nick. Excited to see where the building sets are headed, I mean that’s fantastic, a lot of great stuff.

SANDI GARDINER: No kidding and the work the effects team have been doing is truly impressive. Especially the Herald explosion. That was very cool for me, and as we’re seeing all the ship weapons lined up like that.

CHRIS ROBERTS: Yup, definitely. Seeing things shoot and blow up is always very cool. Nick also mentioned all the work going into 2.6.1. If you want to learn more about the up and coming patch, make sure to head to the website and look at our full production schedule to get a sense of what you can expect. We’ll be updating the schedule every Friday until the release so you can track our progress via our website.

SANDI GARDINER: And of course, a lot of the changes and rebalancing that we’re doing right now are thanks to the fantastic testing that our backers have been doing since 2.6 went live so thanks to all the backers out there.

SANDI GARDINER: Absolutely. Your feedback is invaluable to the project so I want to encourage all backers to keep playing Star Marine, and Arena Commander, and Crusader, and to keep letting us know on the forums, Issue Council, as well as on Spectrum. Please if you haven’t tried out Spectrum, try Spectrum. It’s beta because we’ll be going live with that soon. It’ll help 2.6.1. and 2.6.2. and after that all the subsequent patches to be as good as they can be.

SANDI GARDINER: And as many of you know from our newsletter we recently had many of the directors and leads over at the L.A. office for several weeks of discussions. While that was going on we took advantage of everyone being here to record some deeper conversations about what we’re actively working on.

CHRIS ROBERTS: Yup absolutely, and as the game’s AI system is one of the most complicated and ground-breaking features that is currently in development on Star Citizen, we figured it would be a great opportunity to have lead AI programmer, Francesco Roccucci and Persistent Universe Director, Tony Zurovec have an extended discussion about the systems they’re building.

Star Citizen AI

TONY ZUROVEC (Persistent Universe Director)

TONY ZUROVEC: I’m Tony Zurovec, I’m the Director of the Persistent Universe for Star Citizen.

FRANCESCO ROCCUCCI: So, I am Francesco Roccucci and I am the Lead AI Programmer for Star Citizen, so yeah, we work a lot together.


What do you oversee at CIG?

TONY ZUROVEC: I’m predominantly focused on the architectural side of the stuff, what the basic capabilities are, how we expose that functionality to the designers so that we can get as much leverage from the development process as possible, this includes developing the editor which is the designer’s interface.

FRANCESCO ROCCUCCI: Well, I think about all the architecture from Tony and I make it real in the code, basically. We develop, I put the code, we architect the code and we implement it and then I manage other three people so far on the AI team, so we have Andrega Boni, Mario Sarini, Rich Walsh in the UK studio, so we are a bit in Germany and a bit in UK.

AI in Star Citizen An Update

TONY ZUROVEC: The current state of AI in Star Citizen is that we’re basically, I would say we’re cresting the hill regarding the initial tier of architectural functionality. Francesco has now got a lot of work to do to basically take all of that, distribute it to his guys, get it working within the game and what we’re aiming to do within the fairly near term is to be able to, for the first, in significant fashion, start exposing this functionality to the designers and they should have enough of the base capability to start crafting actual missions within the game.

It’s going to look, to some degree, like the missions you’ve seen before except that they should be able to create them much more quickly which means we won’t be talking about 17 missions, they’ll be able to create, you’ll be talking about much larger potential number of missions. You’ll be talking about a lot larger number of different things you’ll be trying, problems that you’ll be able to present to the player, you’ll be talking about a lot more content that’s going to have an algorithmic component to it.

For example, right now, locations tend to be fixed, and we’re going to be introducing in the not-too-distant future the ability to go down to planets and stuff and we don’t want you to always encounter a derelict space ship with a stranded pilot that needs to be transported to here always in the exact same location.

When you go to pick up that stranded pilot, we don’t want him to always necessarily be friendly; when you go to pick him up, we don’t always want him to be alone, sometimes it’s going to be an ambush, sometimes he’s going to need medical attention.

These are things, all just during that one basic scenario, which is “distress signal, crashed pilot, etc.”, there is a world of opportunity for us to create a range of potential experiences for the player and this is what, as soon as Francesco’s basically integrated the rest of this and I’m sure there is…

FRANCESCO ROCCUCCI: No pressure on me.

TONY ZUROVEC: …Little bits and pieces as we start to mock-up the first tier of the subsumption missions, but as soon as we’ve got that, then we’ll finally start to get a sense as to what the designers can create with this.

FRANCESCO ROCCUCCI: I think it’s always like this first step, right? Of getting the core…


FRANCESCO ROCCUCCI: …there in a state which you’re not only happy with the architecture of the code as well, because for me it’s a lot, I know I can rely on the vision of Tony for that for the architecture of the design and all the other designers that we have, all very talented people and the code as well must be architected in a way you can scale as much as possible.

So, like, you know the big work we’ve done so far is to make sure that subsumption runs and everything with an acceptable amount of memory usage and CPU usage. We have all these concepts of activities that are templates that can be re-used between a lot of characters without creating like instance everywhere or having the mission system as well, reuse as much as possible from the subsumption architecture and so even after we create the same logic all repeatedly in different systems because it uses different systems and it’s also accessing content from one actor to the mission and pass information.

It just becomes very much quicker; you don’t have to go through different DLLs, each DLL can expose different functionality without…You don’t have to add code to be doing complete different structure and then how do I expose this part to this other system, it just becomes everything much more fluid and much quicker to add stuff. So yeah, that sometimes it takes a bit of time but then I think we are starting to arrive to the point where we are very happy with that.

We’re discussing the last bits that you added, for example, this concept of adding some reusable piece of logic that you can recall and for us it’s easy now to implement that and every new functionality seems it’s always quicker to add and you can start to say “We’re doing a good job here” and I don’t like to say it, myself, but I think it’s good when you are happy with the work you are doing.

TONY ZUROVEC: That’s something I referred to before. We’re very focused, and always have been, on the direction we’re going with the concept of “object-oriented content creation”, which is to basically distill all of the different pieces of functionality that we’re going to use to build up these larger missions to break them down into component parts and then, just like you do in an object-oriented language like C++, pass inputs in, to extract outputs so that we can customize them dynamically in the scope of a much larger problem.

And so, what this is going to wind up allowing us to do is, after we’ve got that initial library of functionality, designers are going to be able to craft this stuff much, much more quickly than they’ve could do in the past and what this effectively means to the player is a much larger number of interesting things to see and experience and challenges to face and solve within the game world.


TONY ZUROVEC: The first part of subsumption, the easiest part of subsumption to basically expose is going to be the mission side of it, the mission side is basically built on top of what you call the lower-level AI layer. There’s a lot of commonality in how we do message passing, the data structures, the concept of functions, reasonable parameters, and all this kind of stuff. Task archetypes with which we build up the solutions to these problems. The mission stuff, in general, is a bit more straightforward to expose because it doesn’t directly control any of the animation stuff.

And the animation, when you’re talking about a modern 3D game, animation, even though it’s not technically part of what most people would call the AI, it’s plays a massive role with regards to how people perceive it, if you have a bartender and he’s going about doing his business but the pathfinding is…

FRANCESCO ROCCUCCI: …It’s not smooth or…

TONY ZUROVEC: …the pathfinding doesn’t end very well, he goes right up to other characters and does 90-degree angles around them as opposed to nice splines. If he doesn’t align himself precisely, if things don’t snap right to his hand, if he can’t operate things in a believable fashion then he doesn’t look intelligent, even though under the scenes he’s still able to search for this, he can still basically go over to it, he can still utilize, he can still create items, he can still send messages, received messages, query the environment. All that type of stuff.

And so, one of the big issues that we face before we really expose the AI side of it is we want to make sure that a lot of these, you could call them cosmetic issues because they are really predominantly isolated from the AI itself, although the solution to them is occasionally quite complex because they are still so tightly integrated.

By way of example, one we’ve talked about before is, you’ve got a path find and in combat you’ve got a character and he’s running towards cover and he needs to know in the path find that he’s running towards cover in a combat situation and then he’s basically going to modify the end part of that path find he’s running to that destination, by sliding into cover.

And if you’re talking about a character that’s going to manipulate something, or sit in a chair, they need to know how they are going to need to be aligned. So, while the pathfinding, the predecessor step is in action, it can be taking that into account, because what you don’t want is for them to get to the endpoint and then spin 80 degrees and then basically to play the appropriate animation.

And so, the mission stuff is the first and easiest but given the release schedule and we can’t talk too much about that, we’ve had some conversations about whether or not we might have a small iterative release that might expose some of this for testing purposes but we’ll just have to talk about that as we announce that later.

FRANCESCO ROCCUCCI: Subsumption is a global, general name so I think we’re probably…When the mission system is out, well it’s part of subsumption so there will be a part to experience what subsumption is in general. Right? But, yes, as Tony says, most of the animation issues are the one we are basically working on before we can show something. Or not too much before we “can” show but before we want to show…

TONY ZUROVEC: That’s exactly it.

FRANCESCO ROCCUCCI: …because I think that’s the critical part. It’s like for me, I think we are both very critical to the work we do, so it’s like always say “ah, this doesn’t look very good. I don’t want to show it”.

TONY ZUROVEC: And this was one of the big things with the Squadron 42 Vertical Slice where one of the largest issues was the animation integration…


TONY ZUROVEC: …into the system. As opposed to whether you were, under the hood, doing all the appropriate AI logic. The AI logic, a lot of it was…

FRANCESCO ROCCUCCI: It’s…yeah, a good point…

TONY ZUROVEC: …operational.

FRANCESCO ROCCUCCI: …because the Subsumption logic, It’s already, we have in our game, our builds, we already use the activities. We already use sub-activities. We already have secondary sub-activities surrounding people say “Hi” to you or greet you, seeing from far, having different activities…

TONY ZUROVEC: Right but we were having problems with the heads weren’t orientating correctly and…

FRANCESCO ROCCUCCI: But, exactly. If you showed something like that…

TONY ZUROVEC: …to any user you show that…


TONY ZUROVEC: …It looks much more primitive than that it really is. And so, that was one of the reasons why, up until the very end, we were like “Well how quickly can we tie up some of the most egregious animations issues?”

FRANCESCO ROCCUCCI: Because you want to make sure something that can be very impressive, it is. Because otherwise if you ruin the look of something that is quite complex and quite advanced, with a bad look, then it’s hard to remove that type of first impression. And it’s not smart and it’s like…I don’t think it gives the right credit to the system…


FRANCESCO ROCCUCCI: …And the developers that worked on it.

TONY ZUROVEC: …We want the quality of the animation to match the ambition…


TONY ZUROVEC: …Of the actual logic…


TONY ZUROVEC: …And that’s still going to require some effort.

3.0 NPC AI

TONY ZUROVEC: 3.0 I think we’re going to aim for a pretty diverse set of things. Most obviously would just be starting to make the landing zones start to feel alive, so that you encounter a variety of different characters all pursuing their own set of interests. This will include everything from the prototypical bartender and bar patrons, tourists, vandals, businessmen…


TONY ZUROVEC: …Yeah, various shopkeepers, muggers, criminal elements. Things of that sort. And then some of the more interesting stuff in terms of the gameplay potential, I would say, would be the combat aspects.

FRANCESCO ROCCUCCI: Because the combat is, of course, is what I worked on and it’s something I’m really happy to work on and I want to push it even further. That is going to be interesting because we try to make this first iteration on it where there is already coordination between the guys that fight. People that can advance while somebody else covers the advancement from a cover spot or if no cover spot are allowed they will just…I think the main thing with the combat that I’m trying to push is the fact that they understand the environment. And you are sure, and you understand, that what they are doing makes sense for the type of environment which they are in. And that is going to be what I think is cool in this scenario because maybe you’ll find some pirates on the side you want to get a good deal but then they turn on you. And then the fight makes interesting, the situations.

TONY ZUROVEC: And this will play into several things so that when you encounter…Like right now the only ships, as you’re just generally wandering around stuff, you tend to encounter is other players…


TONY ZUROVEC: …And that should be dramatically different in 3.0. Not only different in terms of you encountering them but in terms of what you can do with them. If you disable…If you’re transporting cargo and you basically get attacked by a pirate and he disables your engines, then it’s entirely possible that he’s going to try to board your ship.

Or vice versa if you’re a mercenary, if you’re a pirate and you’re trying to take out a freighter then part of what you’re going to have to do is, after you disable the ship is board it and those characters will be able to fight…


TONY ZUROVEC: …Fight back within the confines of their ship. And this is, again, just going to open a world of possibilities in terms of making the world, space, etc. feel much more alive.

FRANCESCO ROCCUCCI: Kind of unique as well. Because I think it creates these kinds of scenarios that are not very easy to find somewhere else. It’s going to be a very interesting to see how the players will play.

In the Studio

CHRIS ROBERTS: Thanks guys! Hopefully that gives you a better sense of the progress we’re making and gives you an idea of how awesome the final version can be. As more and more of the features like the AI and the mission system, locomotion for player and AI come online and are implemented into the game we’re able to unlock even more opportunities to push the game even further and create that first-person universe that Star Citizen is going to be.

SANDI GARDINER: And we know that AI has been one of our most asked about topics from the community so we decided to open the conversation to the community and let Tony and Francesco field some of your questions about how it works.

Fan Q&A with Tony Zurovec & Francesco Roccucci

For NPC ships, are those under the control of a single AI or will it be multiple AI working together? What are the challenges of this?

TONY ZUROVEC: Multi-crew ships will operate similarly to the hierarchy. The chain of command that you’d see in a typical civilian or military vessel in that you’ll have a captain and the captain will tell the navigator where he wants to go. And the captain will basically tell the engineer what to prioritize: whether he wants/needs the weapon systems back online first or the shield generators or the engines. And he’ll dispatch the chief medical officer to deal with medical problems in this area or that area.

So, it’s at a very high level. I’ve said before we don’t want to be “Sim Citizen”. We don’t want to have too much micromanagement in terms of control. That’s not the type of game we are. But the concept of being able to just generally prioritize things and then to have your lieutenant characters address those.

And this segues nicely into how players are going to be integrated into the equation. If you’ve got a captain and you’ve got a navigator and you’ve got a chief medical officer and you’ve got a chief engineer and you’ve got a chief weapons officer, and all this other stuff and then suddenly the player wants to, one of your player friends wants to deal with weapons or deal with engineering, then you’ll have a very easy means of factoring in how the other characters are going to interface with them.

And what you’re going to wind up getting out of this is a nice concise system to where you can customize as much or as little as you want. You can recruit NPC characters to do as much of the work, that you and your friends don’t want to do, as is necessary to accomplish the job.

FRANCESCO ROCCUCCI: Yeah, gets you moving away from one behavior for one ship. So, it’s like each seat will be controlled by the people, just one, to work on that seat. So, I think there would be the crew feeling working on the spaceship as Tony already said. I think that is the only way to give a bit more realism in the sense like even if you board a ship and the pilot just tries to steal to bring the ship somewhere and you kill him then you don’t want the spaceship has this external behavior still like drives automatically, controls the turrets anyway. You just want to see that something is there.

Another thing we also try is to make these type of AI modules that you can buy and assign to specific seats. So, you don’t have an actual crew member sometime, but you want to have a small computer at that controls, turret for example. So, some stuff, maybe the medical officer, cannot be substituted by computer, but the turret can have an automatic controller.

And, if you just sit down then you can immediately take control over things, turn on and off. And, I think, as a player, you are also able to customize a bit how the ships automatically respond to things. Can be cool because you can have a combination of, that you can have that already makes a huge difference between one ship and another.

What advantages/disadvantages would there be in running an AI-crew versus an all-human crew?

TONY ZUROVEC: The advantages of AI crew versus human crew? I wouldn’t say there are definitive advantages. I would say the devil’s always in the details. What you will have as a player is the ability to recruit NPC characters to fulfill specific jobs on your ship. But when you basically go to recruit them, that’s where the details start to matter. How much are you willing to pay? And that’s going to determine the level of quality, the level of professionalism, that you’re going to get from that member.

If you want a turret gunner that’s effective then you’re going to have to pay a higher monthly salary. You’re going to have a higher running cost, and so what you’ll see a lot of players…And the exact same scenario exists within players. Some will be more adept at a given occupation, a given skill, others will be less capable, and so what I think you’re going to wind up seeing in a lot of cases is players.
You may have a ship that’s ideally crewed by 6 or 8 or 10 people, and the player’s party will take on those roles that they’re best at and then they’ll recruit NPCs to whatever skill level they can actually afford to fill in the remainder.

FRANCESCO ROCCUCCI: Also about the interest, right? Because maybe some players have interest on piloting, some others have good interest in shooting. And then you can just use the AI to fit the roles you don’t like too much. But on the other hand, the AI won’t rage quit on you, right? You won’t have like a player disconnecting, so…

TONY ZUROVEC: Yeah, but they…

FRANCESCO ROCCUCCI: …You have an an advantage with AI?

TONY ZUROVEC: Well and that’s true. Whereas most of the things that players will do will depend upon their actual capabilities, or their ship capabilities, or their equipment’s capabilities to expose things. But how well you can fire, for example on another player, your accuracy, that’s determined by your physical dexterity, and it’s influenced by the weapon which may have kickback or a wider area of effect in terms of its damage. And NPCs will have a variety of characteristics that will model how well they deal with these different things that the player must handle themselves.

FRANCESCO ROCCUCCI: Yeah, the skills will be a huge part of deciding, like, this NPC is good at doing specific type of jobs. But maybe you have two that are kind of good at the same job, and you still need to fill another role and you put the one. They say, “Well I still need the shield management,” and even if none of them is good but maybe after a certain amount of missions. Well, he starts to get used to that, and he starts to improve his skills as well. You know you can decide, “I will pay a bit less, but I risk a bit more in the beginning at trying to make simple missions so they get used to these things.” I think that is also like interesting to see how the player will use their strategy to their hiring and the other NPC they have with how they can improve and learn.

TONY ZUROVEC: Yeah, it becomes another resource to manage. Another way by which players can differentiate their style of gameplay…


TONY ZUROVEC: …Versus another. Well, it’s like do you go with three engineers of lower quality or one stellar performer?


TONY ZUROVEC: So, there’s going to be a multitude of ways for you to customize the game-play experience via the choices you make in these different areas.

What do you anticipate to be the largest ship capable of operating effectively with an AI crew?

TONY ZUROVEC: There’s no real limitation just given the distribution of responsibilities. It would just be more along the lines of the larger ships will require larger crews to operate effectively and therefore their running costs on average will tend to be higher than ships that can be operated with a much smaller crew. Obviously, you can modulate that to some degree with the quality of the crew you’re hiring…


TONY ZUROVEC: …And things of that sort, but there’s no hard-coded limit and there’s surely no reason why there would be, given how these tasks and responsibilities that they’re going to be executing are going to be ready.

FRANCESCO ROCCUCCI: I think that is coming from the fact that basically within the individual behavior are controlled by a captain or something like that, but then each of these NPCs which do its job. So, there is no real limit on the number of people that execute their job if you know you have like five, six to control five different turrets. Well, you know, that would be just like five different people. Their coordination, of course, depends on the number of targets they have. Who they can target, but then it doesn’t matter like if there are three or five or fifteen, because the code would just give for that would try and distribute the targets, but so exact you know…

TONY ZUROVEC: But as you go to upper-echelon of ships you are going to probably need more experience…


TONY ZUROVEC: …Better commanders, because they are effectively going to be in control of more, of a larger number of crew members and we would probably…


TONY ZUROVEC: …Just model it such that if you basically get a lower…A lower quality chief engineer starts to buckle under the pressure and he starts to make bad decisions…

FRANCESCO ROCCUCCI: …Yeah. Assigns wrong people…

TONY ZUROVEC: …Once he starts getting beyond three or four or five guys that he is controlling. The exact same thing you’d see in the real world.

What kind of relationships do you expect us to develop with AI NPCs? Will an NPC react differently to me based on prior actions? If I ruin a relationship with an NPC, will there be ways to repair it afterward

TONY ZUROVEC: It’s always been our intention to allow the player to form long lasting relationships with all the characters in the game. What this means specifically is that you will develop what we call reputation in a number of different fields. Are you known for your acts of piracy? Are you known for your humanitarian missions? You’re basically very medically oriented and you are rescuing people, healing people, solving medical crisis, you know. Do you…Are you well known for how effective a transporter you are in terms of commodity goods and things of that sort? You have a very high success rate in terms of getting it done?

So, there will be those reputational aspects that characters will be able to query to determine how much they like you, dislike you, whether they want to offer you the more advanced missions or just the starter missions, what they are willing to pay you to do.

There will also be what I call many to one relationships, where it’s not just based upon your piracy rating or your transportation rating. It’s based specifically on your one to one relationship. How do they feel about you? Did you take a specific mission from them? That’s probably going to be weighted much more significantly than stuff that they heard that you did for other characters in the universe.

They will also be able to take into account your…The organizations to which you’re a part. Do they like that organization? Do they hate that organization? Are they neutral to that organization?

So, these things together are all basically variables that the characters, within their conversation logic can factor into how they respond to you. Then we’ll wind up having these things gradually, you know, return to a baseline if you don’t keep doing it. Other words if you were exceptionally good at transporting merchandise, very high success rate. If you haven’t done that in a day, in a week, in a month then people will no longer be talking about that behind the scenes, therefore you expect that reputational aspect to gradually deteriorate over time and you’re going to have to keep doing that type of thing and keep doing it well if you want to maintain your standing with the community, in that regard.

FRANCESCO ROCCUCCI: Yeah, so for us, for me especially now on the AI team, what we try to do is I test these functionalities to make sure that the AI can create as much content as they can. Once we have these things, the relationship variables or attributes, it is better to give these to the designers and then the designers can use any of these. Can be the reputation, can be this, you know, what they feel about you and they can just create all the logic that they want.

Plus, we’re kind of in a unique position because the game can constantly be updated so these characters can also get new responses. You know, based on the state and all the result is new content that we can create with time that makes the relation always more smooth. And we can go deeper into details and mentioned other stuff that you’ve done, mentioned events that have happened. If I have a good relation with you maybe I gossip about something with you and you know, we can create this and always have more and more and more.

TONY ZUROVEC: Yeah and there’s the stuff that matters in terms of the missions they give you whether they’ll talk to you at all and that stuff. Then there’s also just the more cosmetic stuff, where they tell you jokes or they’re light-hearted or they’re basically bantering on about their history because they feel comfortable with you because you’ve done three missions with them within the last week.

Were they frowning at you as you go? Were they repulsed by you? Were they disgusted by you? All that type of stuff. So, I would you say you’ve got different elements coming together which is not just how they explicitly react to you but their more subtle behavioral characteristics when you’re in proximity.

In conclusion, what are you most looking forward to?

TONY ZUROVEC: For the future AI, I think that clearly going to head more towards the algorithmic generation of content. That does not mean by any stretch of the imagination…We’ve said from the very beginning we don’t want to be pure algorithmic. We have no interest in being purely algorithmic, we don’t think that makes for a good gameplay experience. It just makes for good taglines in terms of, ‘hey look, we’re giving you an arbitrarily large amount of…This,’ but it’s all bland and uninteresting. There’s nothing really handcrafted to draw you into it.

And so what we’ve been pursuing for years now, is basically the ability to create all these handcrafted pieces and then to allow the designers to build on top of those. To link those pieces together, to create very quickly what seems like incredibly customized missions but they’re able to do it in much more rapid fashion than what otherwise would be the case.

So, I think what we’re going to wind up having is the ability to algorithmically create more variations on those things. You’re not going to notice it in terms of quality of the content, it’s just going to look like you went from 12 designers dedicated to this to 24, 36 as you basically refine all those algorithms and start to get your base library functions and stuff all entirely operational.

FRANCESCO ROCCUCCI: Yeah, the thing I think we can talk about as well is the thing I would always like to push forward, is the kind of realism. If you want to start to follow one guy, for example, and we start with something like the guy will leave the planet but maybe eventually this guy change planets. And he’ll be able to understand that there are connections between jumping into space ship, then take the spaceship to fly, quantum jump, and then go to this other planet and then go down and search for another part.

You can just follow it, and then you can be always be more realistic. From a tech perspective, we’ll have some optimization. What you’re trading, stuff like this that we want to have as quick as possible. So, I think there are like these two parallel work that we are doing both on the content production tools to create always more and give more feeling and all the technical problems that we are trying to solve like using objects while you use another object, you know sitting on a spaceship and while you are sitting you interact with all the objects and each these interactions triggers something else.

I think these of things that we are going to eventually have…It’s the details. It’s an amount of details that you will eventually have based off all the iteration. So, first iteration will have, even on the combat side, first iteration will have a sort of coordination, but then these other iterations will always have more and more coordination using more type of weapons, being smart on which type of weapon is best in this specific environment or in a specific scenario. I think this is just going to give always more realistic feeling.

TONY ZUROVEC: Yeah because another long term objective and this is not within the next 12 months it’s more of a 12-24 month window type of thing is basically support for what we refer to as the persistent dynamic NPCs and what we mean by that is, of course there will be specific characters at specific areas within even the Persistent Universe, certainly within Squadron 42, but we’ll have those characters even within the Persistent Universe, and then there will be another class of characters that are just created as you wander around the world.

Say you’re basically transporting a load of commodities from one planet to a refinery and you encounter a pirate and that pirate winds up inflicting significant damage and you wind up having to call for help and once your help arrives you basically high tail it and get out of there.

What we want to eventually to be able to do is to keep the essence of what that character was, that pirate, what was his name? What does he look like? Was he talking to you on the radio? What voice were we using for him? What was his rank? What was the source of the interaction that you had with him? What’s your history? He basically got the better of you in a battle. And then we want to be able to use that as an input into this library of mission content such that next time we’re looking for a pirate to be utilized on one of these mission archetypes, we have the possibility to insert him as the specific character and he then can refer to that prior history.

And so, what you wind up having is your own little custom version of the world where these are characters we didn’t specifically go out and hand create and give a history to, but you’re encountering them, you’re interacting with them, made them real, made them persistent for you.

They’re also persistent for any party members that you’ve got, it’s like everybody can see them, but those characters that you’ve interacted with in a certain way become available for us to utilize when we need an equivalent character as you encounter lots of different scenarios within the game world.

Final Thoughts

FRANCESCO ROCCUCCI: Well I think that other things I also look forward is when the wildlife on the planets, that’s going to be interesting because it’s something I worked on a little bit before. In Crysis we had some little wildlife and there’s always this big difference between the smart wildlife and the more simplistic wildlife. If you have a frog, well you don’t expect the frog to be super smart. You can create this kind of, the CryEngine word: Boids, animals that can move around and they react to your presence or your proximity much more.

Then you have other things, like pets or even just bigger animals like a deer or something you know that they’re smarter and they need to do something with you, they’re not as complex as a human, but they’re complex and to give the actual feeling of planet, we’ll have to tackle these things and it’s going to be very interesting.

TONY ZUROVEC: Well, but it’s just like with the cities to where you want to make this world that you’re travelling through look as alive and as interesting as possible because if…Like right now, I’m sure bothers us all is the just the fact that we’ve got a great big world, but we’re not yet able to effectively populate it with a lot of interesting material for the player to see, and we’ve talked in the past about how we’ve got 17 missions in the Persistent Universe that you can go and accomplish right now. But those were implemented in a completely different fashion than what we’ve been aiming to eventually be able to do. But these systems are such that you need to have, maybe not the entirety, but a significant majority of the base functionality operational before you have anything. And so, what makes me most excited about 2017 is just the fact that we should start to get this stuff, that we’ve been talking about on the content generation side, into the game and this is where, all of a sudden, the ability for players to start to do a variety of different occupations and for them to just fly around, orbit around a moon, content we didn’t explicitly put there will be able to fill in, and then there will be interesting stories and things to do all littered throughout the entire system.

So, I think that once we get 3.0, it’s supposed to be the initial iteration of pushing this stuff out there, and then all the successive releases are just going to be refining, refining, basically taking advantage of this system we’ve put in place.

I think that Star Citizen in 3.X iterations is going to finally start to look like a real finished game to where you can just go in and lose yourself for hours and hours at a time and that’s not because you’re marveling at just the technological ambition of the project, it’s because there’s enough gameplay content in there to keep you interested.

FRANCESCO ROCCUCCI: For me, after where we have that, we can see how the player plays with the game. The most interesting thing for me working on games is the fact that you produce something that people play with and they interact with, it’s not that you just give it and it’s not that only, it’s how you use it. It’s how you play that, and the game plays with you so I think at that point you can also start to tweak things and say, “Ah this is how people like to do, and the way they like to play”, and then you can adjust things and give more content here and I think that is going to be very interesting.

TONY ZUROVEC: I can foresee us using spectator mode quite a bit to see how players are interacting with the initial set of mission concepts.

FRANCESCO ROCCUCCI: Stalking Mode. [Laughter]


SANDI GARDINER: Always great to hear more about how NPC crews are going to work. It sounds like it’s been a huge undertaking to develop a system this complex.

CHRIS ROBERTS: Yeah, no it’s a huge undertaking and we’re still hard at work at it, especially Francesco and Tony. We want our NPCs to have complex behaviors you would expect. So, the work Tony, Francesco, and the rest of the AI team have been doing is vital to our plan moving forward in creating this immersive Star Citizen, and Squadron 42 first person universe.

So, once we have the foundation in place we’ll be able to tie it into the mission manager and the mission system itself and create some emergent and immersive scenarios either by scripted design on the designer’s side or through systemic or procedural generation that’s in response to player or AI behaviors. So, I think it’s going to provide some cool and unique gameplay.

SANDI GARDINER: Cool, well that brings us to the end of this episode of Around the Verse, and as always we’d like to thank our subscribers for contributing to the creation of all our in-depth behind the scenes concept.

CHRIS ROBERTS: Yup, thank you guys very much. I thank you for having patience for us being here and doing all of this and a huge thanks to all our backers who are supporting Star Citizen’s development. You guys are a big part of what makes the project unique and special and we could not do it without you guys so thank you very, very much.

SANDI GARDINER: We could not, and if you would like to hang out with more of the team make sure to tune in tomorrow at 10AM Pacific for the latest Star Citizen Happy Hour stream to watch some live gameplay and discussion.

CHRIS ROBERTS: Yes. I’m not quite sure which one of our devs will be there, but they’ll be there and it’s kind of a cool opportunity to have an informal Q&A and discussion, but until then.

SANDI GARDINER: We will see you.

BOTH: Around the Verse.

About the Author:

Leave a Reply