Vehicle Feel Masterclass: Balancing Arcade Accessibility with Simulation Depth

[Music] so hello everyone my name is Matthew Harris from criterion games and welcome to vehicle film master class balancing arcade accessibility with simulation depth and if you think that's a lot of words to felt on one slide then you're gonna love the rest of this presentation during this talk I'm going to discuss proteins vehicle design philosophy and how we build vehicle handling that is both intuitive but deep and master ball this talk isn't a set of hard and fast rules to follow to me the perfect vehicle game as all games are different but it's more sharing about processes and guidelines that we follow throughout development with a view to both inform and inspire hopefully you'll find some useful learnings whether you're new to vehicle design or who just shipped your seventh Forza title so before I get stuck into the presentation I want to briefly talk about criterion games as history originally set up in 1993 as criterion software by turn developed the render wear engine and started making games in order to show it off early games such as trick style that aren't incredibly popular burnout franchise and then on to Need for Speed hot pursuit

and Most Wanted over a decade of working after over a decade of working on racing games we started to lend our vehicle expertise to dice where we worked on vehicles in battlefield 1 x-wing VR mission and most recently stars battlefront 2 as for myself I joined criterion 5 years ago I need 50 verse Need for Speed Most Wanted and went on this selection of titles a criterion we encourage cross discipline skills as much as possible and so of my career accrue over my career criterion as a programmer I took special interest in game feel particularly in the design of vehicle handling this proved incredibly useful as understanding how to tune in design great vehicle handling as a programmer meant that I could work closely with our designers to deliver great game feel and that's why I'm a programmer standing here giving a design talk ok so what I'm actually going to talk about today as you might guess from my background this is give me a technical design talk and it's aimed at taking you through the key parts of criterions vehicle design philosophy along the way I'm going to talk about the key differences between arcade and simulation games and how each type bounces accessibility with mastery and

depth then I'll move on to the simulation itself and discuss our physicality first approach to designing vehicles and how to identify where the play in our vehicles comes from after setting up a physical simulation we had accessibility to the input layer on our cysts which are critical to providing accessibility without reducing depth once the vehicle is controllable and satisfying we need to feedback the vehicles handling by the camera I'll talk through some key concepts we use here throughout the majority of the talk I'll be using motorsport racing as the context as it provides a great variety of examples however in the final section I'm going to show how a design philosophy can apply to a very different type of vehicle the starfighters in styles about a front to one important thing to know here is that the whole talk is focused around the handling side of game field things like audio VFX sensor speed and force feedback are very important to game feel but you could produce higher presentations on those so I won't be talking about those today let's start with a discussion on arcade versus simulation when thinking about the term arcade I don't mean a game genre like

top-down shooters such a geometry Wars I'm talking about a particular set of game design decisions that lead to a particular arcade feel games that have an arcade e-field tend to focus on intuitive controls that are readily apparent to the player make compromises on realism for the sake of fun and haven't often have a multitude of game mechanics that the players trying to balance against each other simulation games on the other hand I trying to replicate a real-world experience and so tend to have a control system that's about immersiveness first and aims for a high degree of realism and believability from his gameplay and handling simulation games tend to have fewer game mechanics than arcade games as they tend to focus their gameplay around controlling a realistic vehicle rather than one with weapons and boost different design decisions can lead to a more arcade or simulation feel for your game so in the next slide I'm going to try putting a selection of multiple games on a sliding scale between Arcana simulation let's start the simulation end of the spectrum with games such as for the horizon for regular Forza and iRacing these games are focused towards delivering a realistic simulation of motorsport with a gameplay entirely around racing

lines cornering and drafting they tend to have most of their assists off by default but will often give the user the option to turn on accessibility options such as help with braking or steering simulation games tend to end up with quite high learning curve but a very rewarding for players who put in the time to master them as we move towards arcade-style games we have a subcategory in the middle of sim cade which aware hybrid games such as Forza horizon and Need for Speed they've these examples aim for a Hollywood realism tone where things are a bit more over the top but still within the realms of real-life pleasure doing exaggerated drifts and massive jumps putting cars from real manufacturers they're often additional challenges added on top of racing in Simchat games such as extreme horror extreme weather in Forza horizon or weapons and cop chases in Need for Speed finally we get to the arcade star racers which tend to go for a heavily stylized or even a cartoonish tone our case star races tend to have a large variety of game mechanics such as drifting boosting stunts and weapons in

which the players have to learn in addition to racing for this reason the racing gameplay tends to be quite accessible for new players but still provides plenty of depth for players who really want to get the most out of the game so in comparing Arkady's simulation the arcade style races have a variety of game for mechanics on top of cornering and speed management such as weapons stunts and boosting simulation races have fewer mechanics for the player to focus on but have a simulation focused towards realism and their aim is to make cornering and racing as satisfying as possible arcade is generally more accessible and fun whereas simulation is hard to get into but offers a more serious rate and more serious take on racing competition one key thing that is consistent across all these games is that the player has two input controls that are satisfying cadence the actions per minute for simulation games and arcade games are actually pretty comparable for a simulation game the player focuses all their attention on their racing line trying to tap break and turn at just the right intervals to nail that corner whereas in arcade game player that the player divides their attention between

drifting around corners but also boosting and firing weapons although some people might see arcade style races as casual and simple in fact they have a wider variety of game challenges for the player to focus on than in simulation games so when deciding where on this scale you're aiming for game to be you should consider a number of factors what sort of players are you targeting your game at are they serious sports fans the demand realism or your players more interested in fantasy fulfillment how serious and realistic is your game tone if you're planning on having lots of mechanics such as boost in weapons then you're likely for more on the arcade side of the spectrum balance the number of mechanics and complexity of the simulation against your assists in order to achieve a sensible rate of actions per minute how much effort do you understand your physical simulation if you can't put much time into physical simulation but you're trying to build a realistic motorsport simulator then you might end up with boring and shallow gameplay do you want your game to be accessible by lots of players or you

focused on an experienced target market if you're building an arcade or simce Dreiser you'll likely need some assists to help keep the input frequency at satisfying levels once again these aren't hard and fast rules but you should have a strong idea of what you want to see on the spectrum don't try and have it all otherwise you risk building an overly demanding and complex game a criterion our approach has been to combine what we see is the best of both worlds between our caving simulation we start with physicality and build a realistic physically based simulation then we build assists on top of that which makes the simulation more accessible and allows the player to focus on more things at once our designers them very carefully balanced physicality accessibility in order to get a balanced experience that hits our aim of easy to learn but hard to master criterions use this type of approach since never speed hot pursuit and that's worked well for us ever since ok so the next part of the vehicle design philosophy that I'm going to talk about is the physical simulation of the vehicle

so we see the physical simulation as a foundation allows us to build a game with deep mastery and rewarding challenge in this section I'm going to start by talking about how the context of our game affects the elements we might want to simulate and then go on to compare different physical simulations of car dynamics but using approaches that mimic real-world physical effects we ideally want to pick parts of the simulation that will have played to the handling giving the player multiple things to balance and learn in order to achieve their goal before starting work on your physical simulation it's important to consider your game context this is because we see the context as a serve requirements on your physical simulation think about if your environment is wide open or enclosed how might this affect that the desired behavior of a vehicle also consider what the players goal is when driving the vehicle are they racing performing stunts or they just trying to get from A to B if getting from A to B needs to be a challenge then maybe your game is about complex traversal here's an

example of a racing sim Forza Motorsport the vehicle is on a wide road and will encounter long curved turns and the player wants to win the race through precision driving through corners and drafting behind other races on the other side of the racing spectrum we've got something like burnout which is sent mainly on urban roads with some tight turns and some jumps navigate the player could be racing against other vehicles or could be doing stunts to win a score attack mode there still have focused on cornering here but also on boosting and stunts destinies Sparrow is a great example of a transport vehicle it's primarily designed to go across open terrain but the player will often pilot it down narrow corridors so it needs to deal with both the main player intention here is to get to their target without colliding and slowing down too much but it's lower pressure than the previous examples in uncharted the Jeep provides some light puzzle mechanics when trying to navigate our rookie mud slides and the challenge to the players in picking an effective route of the mudslide while trying to keep the tires in a state of grip each context has its own

requirements from the vehicle if we change the context we should re-evaluate the vehicle and see if it's still suits it one example of this would be the sparrow in destiny which I really enjoyed driving as a transport vehicle but when they introduced Sparrow racing I didn't feel like the lent itself to racing very well as a time special event this was fine if they wanted to build more of the game around racing I'd hope that they would retune the vehicle to fit that context during production it's also easy to forget about updating a vehicle if the context shifts during development particularly if that vehicle is shipped before we build the speeder bike for Star Wars Battlefront 1 and because it shipped in that game we made the assumption that it wouldn't take too much more work for styles Battlefront 2 however the new games map topography coupled with some tech debt ended up meaning we had to rework it entirely for the sequel vehicle and environment are complementary to each other and changing one may require changing the other so for this next part of the talk I'm gonna use motorsport racing as a context although my examples could apply to all racing games let's think of this is

somewhere on the simce deepest part of the spectrum such as Forza horizon or Need for Speed there's an expectation from this context for realism so we need a simulation view based off of real physical effects and allow the player to play with grip and acceleration there needs to be enough depth to the simulation for each vehicle to accelerate and turn differently so our simulation needs to be detailed enough to support this accurately there's an added benefit that if we use a realistic physical simulation we should be able to use real-world tuning data such as engine or tyre metrics as a starting point for tuning our vehicles let's start with a trivially simple model in order to illustrate why detailed physics models are needed this is so basic the very few games would be served well by this physical model uses a simple frictionless rigidbody with a single force and torque apply to the center of mass to move it the result is vehicle motion that is flat unrealistic and not enjoyable or interesting to drive it's very abstract in the way the forces are simulated so we should take some inspiration from real-world vehicles in order to improve it in the following slides I'm going to

go through how a more complex physical simulation could improve the physicality of our vehicle I'll start by looking at a list of real-world systems for motorsport vehicles and then I'll pick a couple of those to go into in more depth we'll look at the powertrain and also tire friction finally I'll conclude this section with some footage from the Student Need for Speed Most Wanted in order to show off physicality in a shift game so here a handful of real-world systems that you might want to simulate to add more physicality and play to our vehicle this is an exhaustive list but it should give you an idea the sort of physical effects you might want to consider in the following slides I'm going to go through engine transmission and tire friction the engine and transmission are key parts of the powertrain some when I refer to them collectively as a powertrain system let's look at an element that we would like for our complex physical model the powertrain the purpose of the powertrain is to transmit energy from the engine into the wheels in this case the engine provides energy to the clutch which passes it

through to the transmission and then to the differential and wheels our first glance you might think that we could just tell the wheels to start tearing at a set speed and we get the same result however the powertrain has an important quality it is a feedback loop the engine provides talk to the wheels however the wheels will resist movement not just because the wheel itself has inertia but because the wheel is in contact with the ground and needs to overcome the inertia of the whole vehicle before it starts to rotate an optimal speed so we get resistance feedback through the whole powertrain this means that some of the energy is used to overcome its inertia affecting the amount of energy to use there's actually used to drive the vehicle forward so if we decide to model these parts of a car we should look into each piece and see what we might want to put into our simulation for example your engine simulation could need author will talk at different operating rpm as well as the inertia of the engine itself and supporting that feedback mechanism that we just talked about your clutch might want differ at different operating efficiencies at different rpm your transmission could simulate different numerous different gear setups and how

long it takes to change between them finally your tires should apply force in a realistic manner via tire friction which I'll go into later so we've gone away and implemented an improved powertrain so let's look at what that actually gets us let's compare the simple model with the new complex powertrain version using speed over time graphs on the Left we have the simple model which has a very uniform acceleration profile comparing that to the more complex powertrain we overall see a more curved acceleration profile indicating that we get more and more acceleration at the beginning of takeoff than we do and when they're in top speed you'll also know that there are small bumps in the graph which represent the gear changes due to the simulation of the transmission we have multiple gears that are more effective at providing acceleration at the lower of their working speed range and then starts to tail off as the vehicle speed approaches the next gears working range even just looking at a vehicle driving in a straight line the simulation is added much more complexity and interesting physicality to than this simple model next up let's have a look at another system tire friction so looking at this

diagram we have the wheel hub with hubcaps illustration in the centre and the rubber tire around the outside the section of the bottom of the tire is slightly flattened another load of the vehicle and is called the contact patch the blue lines represent the tread and wall of the tire and we'll show how it deforms as the wheel moves forward the green dot is a specific point on the wheel hub and the red point is the corresponding point where the tire touches the ground via the contact patch now as the wheel hub rotates clockwise if we look at the green point on the hub its moved a small amount and this means that the wall and tread of the tire stretches between the green and red points the elastic force of the rubber tire now strives to spring back into his rest position moving the red point on the tire surface towards the left even though remains in contact with the ground this applies an equal force back on to the wheel hub which pushes the whole vehicle forward slightly this happens continually until the vehicle is traveling in a spew which matches the rotational velocity of the tire now we understand the physical behavior were trying to simulate we need a simulation model the dynamics of how

much force is applied under what conditions is incredibly complex and there are many models that can used to simulate it in this case I'm going to use the Pacheco model in the Pacheco model the formula takes the rotation speed of the wheel and relative velocity of the road the slip angle is the angle between the direction of the ground velocity and the direction of the wheel hub the Pacheco formula produces two forces the longitudinal force which drives the vehicle forward and the lateral force which turns the vehicle during acceleration the road speed is lower than the wheel speed so we get longitudinal slip which maps to a particular force in this graph the right-hand side is when the vehicle is accelerating and the left-hand side is when it's breaking this red graph is the lateral force applied under a particular slip angle so represents the turning force applied during cornering what's interesting about these graphs is that each of them has a peak which is where the most effective acceleration or cornering behavior is found this is what gives us the limit the effect of playing with the limits of grip this means that the player has to try and hit these Peaks in order to be as effective as possible during the race this adds a lot

of depth of mastery but let's look at what this actually results in so let's have a look at how the physical simulation ends up looking in the final game pay close attention to the way the chassis the vehicle rocks back and forth during gear changes due to the simulation of the drivetrain what other body rolls to the side during corners due to the simulation of tire friction and suspension the depth of the simulation shows through and gives the player precise control and physical play to master [Music] [Music] so now that we have our physical simulation sorted and we've got an accurate deep simulation of the desired vehicle we should look at input and assists in order to give the game that feel we want and get the balance between arcade and simulation that we're aiming

for in this section I'm going to talk through drifting which in case you didn't know is whether we forced the rear wheels of a tire to break traction so the vehicle drift sideways in Ottoman maintain momentum what's in a turn drifting is something is possible but tricky to achieve in real life so in criterion games we apply several assists to help players achieve a long satisfying drift so I'll use that as an example in the following slides let's start by thinking about the input layer this is effectively the control mapping and any modifications you might do to those control inputs these control inputs are then sent across to the simulation where it turns the wheels or drives the engine in the simulation however the important point to take away here is that you have assists that might work on either side for example the input layer level assist might be to detect a particular button combo to trigger drift mode and a simulation level assist might be to apply increased physical damping whilst in this drift mode so we know where we might power assists but what we want to achieve by adding some once you've got your input layer communicating with a simulation you'll have something playable and you

can then review it and see what needs to change you might find the IMP and complexity is too high and you might want to reduce it it could also be to reduce the simulation feel and move towards more of the arcade side of the spectrum you might want to take existing forces or effects in the simulation and emphasize them such as making drifts last longer a detailed simulation can also bring unwanted side effects such as extreme slowdown on uneven terrain or even collision hitching this might be something you want to target with assists it could also be something which conveys the game state back to the player in a more intuitive or obvious manner so the learning curve is easier even though most of these examples of your intended default experience you could also use the Syst as accessibility options to make your game playable by a wider range of players now that we decided what we want to achieve through our assists let's look at some of the different types I divide assists up into five categories player intention driver physical real world and back assists I'm going to talk through

each of these in the next few slides these aren't hard templates for more thought exercise that might help you think what parts of the game to target when we were building the drift mechanics in Need for Speed we took our simulation which have realistic drifting behavior and used assists to make it more stable easier to sustain and maintain more of your speed than in real life to get this to feel great use multiple types of a assists to achieve this I use drifting as an example when talking through these categories the first type of assist is the player intention assist this is an assist of the input level which looked which looks at the players input and tries to work out whether the players intention is different from the normal mapping it can look at the current simulation to provide context what the players intention might be an example is using a brake tap to enter drift if you're driving if you're driving straight ahead and tap the brake then you probably want to slow down if you're dry if you're turning heavily when you to have the brake you probably want to enter drift in order to turn more aggressively next up is the notion of driver assists which can simulate the experience and muscle memory of a

professional driver an example of how this is used in drifts is in bouncing your talk when you're drifting in a real car the tires are trying to twist themselves to match the direction the vehicle is drifting and the driver is able to feel this through the steering wheel and adjust accordingly to extend the drift on a console pad the analog stick doesn't really lend itself to this so we emulate this behavior by Reese entering the steering input so that zero input equates to keep drifting at this angle rather than meaning straighten up the wheels this is absolutely key for a controllable drift and gives the player the level of control and ace driver could achieve we also apply physical assists which are forces that we applied directly to the vehicle's physics simulation we take existing qualities of the simulation that we want to feel hyper real and emphasize what's there we want to look at the forces applied by the situation that are hindering the effect we want and counteract them for example if you're in a deep drift a component of the tire friction is opposing motion and slowing the vehicle down we want our drifts to maintain speed round corners so we calculate the magnitude of the frictional force

slowing us down and reapply that force in the direction of the vehicles velocity in order to minimize the speed loss or allow our desires to tune how much speed loss they desire real-world assists are where we might simulate part of a vehicle which is available in real life and we can manipulate it to achieve the desired effect for example we might want an anti-lock braking system to allow players to brake effectively without how to carefully tap the triggers when we start drifting we can turn off ABS to allow the back wheels to spin and lose traction entering the drift quickly when we detect that the player is trying to exit the drift we can then turn ABS back on in order to allow fast acceleration out of the drift finally we've got feedback assists which don't actually affect the handling but our assists which help the player understand the simulation better and show you why to help guide the player wherever possible we feedback the simulation from real-world sources like the strain of the engine or the squeal of the tires for feedback assists are specifically things you wouldn't see in real life an example of this is the Forza braking line which looks at the upcoming turn

and highlights the line read if your tires are gonna lose traction and you ought to slow down seeing as assist can influence both the controls and the physical simulation getting the feel of them right is just as important as tuning a physical simulation itself depending on how you engineer and shooting them you could end up disguising all that great simulation work so here are a few tips to bring out the physicality when writing assists firstly try to tune them to be as subtle as possible while still achieving the desired effect it's often easy to have very strong stabilization forces which help the player effectively a wipe over all your great simulation work rather than switching your behavior on or off based on a binary condition try to fade them in or out over some parameter even if that parameter is time this means there is less obvious to the player they're being assisted and add some nuance to it think about where the optimal assist strength is and allow the player to gradually learn how to find the perfect balance giving it depth and mastering the game should feed this back to the player and not to allow them to tell if they're doing better or not also as a programmer try to make sure you add

in plenty of helpful debug visualizations in order to help your designers understand the systems but being able to tell what by being able to see what physical effect their parameter changes have and tune them as well as they possibly can if possible set up auto tester to your vehicles and they're expected top speeds accelerations etc as this can help catch unexpected changes or your physical facts affecting vehicle physical assists affecting vehicle performance we had an issue on Most Wanted where a vehicle artist adjusted the size of a wheel and then this affected the top speed of a vehicle but thankfully this was caught by an auto test and fixed quickly so we've got our physical we got physics and our assists working nicely together now we need to feed this back to the player to affect it feed this back to the player effectively through camera behavior I wanted to start a section off with a phrasal user criterion a lot camera is handling it doesn't matter how great your simulation and assists are if your camera doesn't feed it back to the player well a stiffle loose camera can

massively change the feel of a vehicle so should be considered par for handling we have the same designers that tuned the handling parameters on a vehicle also tune the camera parameters because they're so intrinsically linked there are two fundamentally different approaches you can take one stand to think about vehicle cameras one is a camera LED system where the player input turns the camera and then there's a layer that interprets the camera direction and passes that through to the vehicle controls alternatively a vehicle LED system is where the player input controls the vehicle directly and the camera follows the vehicle as a separate system camera LED systems usually match the shooter input paradigm with aim and pitch on and pitch on the right stick generally allows the player to point the camera accurately or something.but at the loss of handling feel and precision this can be good for games where the primary vehicle gameplay is aiming and shooting a weapon and the driving component is secondary an example of this type of camera is the monk v tank in battlefield 1 this video shows me rotating the camera separately to the vehicle using the right stick and

holding the left stick forward in order to move towards where the cameras pointing there's an intelligent layer that decides how to move the tracks of the tank in order to make it point towards the camera rather than direct control one problem with a layer like this is that it can add some ambiguity which means you can get some unexpected motion particularly on trying to reverse a vehicle with camera LED handling vehicle LED camera are usually the default option for racing games and usually match the driving in for paradigm where steering is on the left stick and the camera follows the Vickers motion this gives a more direct control feel and can allow greater driving precision and camera feedback but makes it harder for the player to quickly move the camera to look behind them these types of cameras are often accompanied by a look behind camera button in order to compensate for this here's an example of one of the transport vehicles the criterion worked on it for battlefield one the armored car which is we decided to use a vehicle LED camera for best game feel a criterion we almost always use vehicle though cameras due to the precise handling feel we require from our games I'm going to consider it

continue this section by looking at vehicle at cameras in more detail so a good camera should feedback important game state to the player such as camera angle while drifting you will respond to player needs ensuring that the cameras focused on what the player needs to see and does an abstract line of sight the camera should sell physicality by using the physical simulation to complement the motion of the vehicle so let's look at a rigid camera which is locked behind the vehicle and points directly down the front of the vehicle this is a camera with no behaviour and this is simplest possible starting point the vehicles position is effectively locked on the screen which means that an either physicality from the simulation is shown through the background of the scene rolling and pitching as the vehicle drives which can be quite disorienting it makes it difficult to focus on game objectives to improve this we should start by

conveying the motion of the vehicle by moving it in relation to the frame of the screen if the vehicle is in a deep right turn would expect that the camera would lag behind slightly and the vehicle will shift to the left of the screen in order to achieve this we use a sprung camera system whereby we set a target point for the camera to tend towards but allow some motion in all directions with a force pulling the camera to the target to a target position proportionally to how far away it is from the target we can tune this power access to land for certain behaviors to convey differently on screen for example we might want lots of sideways motion when turning but less forward and backward motion when accelerating and decelerating in conjunction with pulling the camera towards a target we also point the camera to look ahead target which is fixed in front of the vehicle and shows the player where the vehicle is heading it's also sometimes worth considering using velocity extrapolation on your look ahead target in order to mate your camera look into turns and keep the players focused near the center of the screen when turning let's take a look at what that looks like so know through walk throughout all these videos the

physical simulation and handling are identical all we're changing is the camera so as you can see the vehicle is much more dynamic on screen and actually see the size of it as well as the motion of the vehicle when traveling over bumpy terrain there are three key parameters we use when tuning this type of camera the most important one is convergence which is how far we move the camera towards the target each sim frame the higher the value the tighter the camera is to the back of the vehicle then we have Mon momentum which accelerates the camera in the direction it's already traveling and damping which is used to stop the camera overshooting the target and resulting in oscillations in order to demonstrate the different types of feel that different converters of parameters can get you here's an example of strong convergence [Music] notice it is very similar to the rigid

camera example vehicle isn't moving too much on screen it doesn't feel very physical switching to a weak convergence we see a lot more motion in the frame the vehicle feels dynamic and aligned however because the convergence is so weak the vehicle appears to be pointing off-screen during large turns this can make it tricky to control as you can't see what you're going in this situation we could either tighten up the camera or add velocity extrapolation to our look ahead target so that the camera looks towards the vehicles direction more effectively the camera also needs to sell the physicality of the vehicle and feedback what the simulation is doing a good example of this is the landing camera when the vehicle comes off a large jump and then hits the ground so hard that the suspension bottoms out we wanted to convey this with a large camera movement that doesn't obscure the road ahead for long but still sells that impact we use some camera shake when we land Plus this is a good example of how old momentum and damping forces are fed the camera to give it an exaggerated

behavior when landing so if we don't do anything special to the camera this front camera still gives you some feedback this isn't too bad and you could probably ship this but let's go the extra mile and try to improve it let's add some camera shake on landing when we detect that all four wheels are on the ground after traveling downwards quickly this camera shake that I'm about to show you is 30 Hertz oscillation in pitch and roll in north 25 seconds this helps a little bit but we can do better [Music] let's reduce the damping of our sprung camera so the momentum term carries it further downwards than the target giving us a much better sense of the impact this should make our impact a much more impressive exciting and physical

so how does AI battlefield one work stack up to the camera requirements we set ourselves earlier well we feed the simulation back to the player by using the camera convergence to affect the vehicles position on screen allowing the player to tell the turning radius of the vehicle the camera responds to player needs by looking into turns letting the player see where they're going we sell physicality by giving special attention to the landing camera making the vehicle feel extra heavy and solid here's another look at the armored car illustrating all those previous points but also gives you a chance to hear the audio treatment and give the V answer show them and see the VFX which adds even more on top of our handling work [Music] [Music]

so in this final section I'm going to show how we apply these principles when making star fighters in Star Wars Battlefront 2 when we started looking at what we could bring to Styles by the front two we decided on the overall direction of wanting players to feel like an ace pilot and give them that fantasy when controlling the star fighters after playing some battlefront 1 we felt we could help improve the handling significantly we wanted this to go alongside a game mode there was much larger scale that in that game this is the original Fighter Squadron mode from battle for one where players for Danang swings and TIE fighters above the surface of planets the main shooting mechanic here was to hold left trigger to lock onto anything within a wide code in the middle of the screen a right trigger to fire shots

have them be auto aimed as long as you had a lock on you may also tap Baylor's help in this video that bar from one used a camera led handling level where the camera moves first and then the vehicle catches up [Music] [Music] so when we started to look at how we could improve over Battlefront one we looked at player feedback players really enjoy getting to play out their Star Wars fantasies plus the gameplay was very accessible to a wide audience however there's a general feeling of lack of depth around the auditorium shooting that it was difficult to control and the staff fighters felt too much like toys on a string on players really wanted manual roll control so we took each of those points and make sure we focus on fixing them with aiming close to that of a shooter handling the felt precise and a camera that showed off of physicality the vehicles and of course roll control this is a great

chance for us to apply the vehicle design philosophy when designing our star fighters here's some gameplay where we ended up with on battlefront twos star fighter assault mode [Music] [Music] so as for reviews we've got a lot of good responses around the stuff I do what we did with critics highlighting the handling and shooting being a huge improvement over the first game so now we know how it turned out let's see how we used each section the vehicle design philosophy during development when

looking at balance between Arcada simulation we looked at our target market which consisted of both fans who are casual gamers and wanted an accessible experience but also hardcore players who expected something more like a flight sim to live out their pilot fantasies we ended up deciding to balance this by having intuitive pitching your controls that you might find in an arcade style game but give players full roll control and allow players to execute their own evasive maneuvers rather than having them triggered as cutscenes on the d-pad but like in battlefront 1 we also decided to go with a shooter input paradigm with flight and aiming on the right stick as it be more intuitive to players coming from a shooter background and also match the rest of the game however we know that anyone expecting a flight sim style control so that would find this quite jarring so we made sure there was an option to remap the analog sticks fully when thinking about how we were gonna simulate our starfighters we look to the real world analog of fighter planes which were the original inspiration for Star Wars is dogfighting scenes which is why in Star Wars x-wings Bank into turns like planes and crashing Star Destroyers fall downwards thinking of fighter planes

wheeler two air foils which is the thing that keeps planes in the air but also gives secondary effects such as altitude loss when turning we decided this didn't really match with what we expected for a Starfighter so we went with a simple anti-graft simulation instead we did however when the banked turns from a fighter plane so we simulated this with a horizon steering setup where the controls remained a lot to the horizon but the Starfighter banks on-screen we used a simple engine thrust force instead of a more complex prop or jet engine as we didn't feel like we really needed the play of airflow in space where there isn't any air when in a heavy turn the aerodynamic drag of a plane can affect how aggressive it can turn we ended up doing something similar with an air brake that allowed a lot of experienced players to pull the air brake when turning to get a temporarily reduced radius even though there's no physical justification for it it ended up fitting right and it was something that we investigated in a prototype and play tested it until we were convinced we wanted it in production in order to get

the feeling of bank turns we use horizon steering which is quite a simple concept where we have two sim two separate frames represent an orientation in practice this could be a matrix or a quaternion the control frame stays flat when the player yours which means that it points in the direction of flight following the pitch of the Starfighter but allows the yaw of this ship to be expressed in the direction it appears to the player on screen when the player turns left and right on the stick the control frame stays flat against the horizon but the vehicles body frame is rolled into the tone giving the feeling of banked hands we wanted looking at the video on the right you can see the Starfighter banking with a camera staying flat allowing the player to intuitively control the fighter direction if the Starfighter was pitched up then the staff ID would rotate on this newly oriented horizon so if we were pointing directly upwards instead of turning left or right we would end up pointing downwards in the local pitch horizon rather than fading to the world's horizon so let's move on to the camera initially we wanted to we initially wanted to work out whether we should use a camera LED or vehicle LED

system we initially thought that a camera LED system would allow for the precise aiming that we wanted however this was the system that battlefront one used and I didn't have the sort of precision that we'd hoped for using a vehicle LED camera instead would allow us greater flight precision which is one of our initial goals and improved physicality for that reason we opted for a vehicle LED camera which is a sprung camera in a similar style to what we did on battlefield 1 y-wing so thinking about the approaches we learn about in the camera section I've got a vehicle LED camera who looks into the direction of travel and allows the Starfighter to move significantly but predictably on screen [Music] one problem we encountered while authoring camera shake on battlefront 2 was it was really hard to aim while the camera was shaking this seems pretty obvious because you're aiming down the camera so of course shaking it makes

things harder however we wanted to try and find a way to convey the same camera shake feeling whilst maintaining aim precision conventional camera shake works by rotating the camera around its center point pitch Ural rolling by X degrees which means that if your focus point is in the distance it moves a lot on screen to come by this we wrote an orbit camera shake system which would shake the camera by moving it laterally as well as rotating it keeping the camera pointed at a particular point far away towards the horizon there's men of the things in the foreground such as the vehicle would move around on screen but distant targets we were aiming at remains stable in this video keep an eye on the pillar in the distance notice of the pillar crosshair and vehicle all shaking around significantly when we move over to the orbit camera shaking only the vehicle is shaky I mean only the player feels that they're being buffeted around but can still aim effectively it's easy to imagine blending these two effects together to allow our designers to tune just how much shape they want for near and far objects but in our case we opted for

distant aim stability due to our games focus on shooting seeing as we have a strong sense of up in our levels we found that players would end up upside down quite frequently even if they weren't using the roll control in order to help alleviate this we added some auto rolling which would kick in when the sticks are centered the idea being that return to center is a player intention to fly straight so that's always accompany by rolling to the horizon players won't find it too obtrusive it's also a smaller more subtle force when you're almost level again trying to be as unobtrusive as possible we also have a small 20 degree window when the star fighter is inverted where the auto roll never kicks in this is to allow players to fly upside down when they intend to all other angles tend back to the horizon it's also worth mentioning that we anticipated more hardcore players of our audience would prefer to have total control over orientation so we added a toggle to turn auto roll off in the settings after getting the starfighters to handle quite nicely we looked at shooting we

found they was incredibly hard to hit a moving target by pointing your ship towards another as an illustration let's think about aiming towards this Thai interceptor firstly our maximum targeting range was around 800 meters which meant that ships could end up quite small on screen they were also traveling it up to 120 meters per second and also the lasers and star wars have a travel time so you need to lead your target these factors combine to make unassisted aiming incredibly difficult for our players so we needed to add some assists one of the main assists we added was the auto aim cone which is a cone in front of the player which we targeted the gun barrels to fire the laser shots more towards the correct target position in theory it's quite similar to the battlefront one approach except that we gradually fade the effect in and out rather than switching it on and off like in battle from one in the case of this diagram we've got the player aiming ahead of the target and we have a predicted position of the target which is the yellow dot which is where the ship will be by the time a player's shot gets to it we show a reticle and screen in this position so the player knows

where to aim this is a type of feedback assist because it's providing extra simulation data to the player fire you I if this predictive position is inside the aim cone then we shift the gun barrels towards this perfect predicted position however we fail it gradually based on how far into the cone the predicted position is this means that if the players aim is quite far from the target they only get a little bit of help but as the crosshair gets closer and closer to the target reticle they get more and more of an assist until they're inside the reticle which guarantees our here because it's fully assisted this makes the effect subtle enough that players don't notice the shots diverging from their crosshair but they hit a lot more hits than they would otherwise it also gives a nice feedback loop of getting gradually more and more hits as your name improves rather than the old approach that was either hit or miss the aim cone helped a lot but the feel of putting your crosshair onto the predictive reticle still felt quite fiddly players were constantly overshooting making a lot of correctional inputs the main reason for this is that the targets are constantly moving so instead of the player trying to aim at a particular static point

they're actually targeting an ideal angular velocity as well an orientation north to keep up with a moving target we decided to reduce this complexity by introducing sticky targeting this was inspired by one of the amos sisters scene in destiny where they pull the cursor along with the enemy when the enemy strafe is left or right effectively we want to drag the cursor along with the target and to do so we take the linear velocity of the target in world space and then transform it into the players and local space angular velocity this target angular velocity can then be mapped into the input layer we can effectively work out how close the player was to achieving this ideal input and then fade the value towards this ideal in a similar way to how we did it with the cut with the aim cone logic we use the aim cone to fade the effect in and out and we only apply us to get outing when the player has nonzero right stick input which means that we only help them aim when they're trying to aim the senator working really well as we show in the next demonstration video in this first part

Steve targeting spot you can see that the player is making a lot of dirty directional movements to try and aim at the enemy his results are the most kills only happening when the enemy is traveling in a straight line or makes for an easier target this next video shows the contrast when we turn sticky toffee after maximum the aim point is super stable non stop tiger but he'll be able to feel the curse is being dragged along with the timer as in this case where the cursor calls the tiger plate for total flight path exactly then this last part shows the tuning ended up shipping live where the player where the players much reduce correctional input but still subtle enough that the player doesn't notice that they're being assisted so overall we ended up with a hybrid vehicle shooter experience that we're really proud of we use a simplified physical model but with the control options for highly skilled players we use the vehicle LED camera to allow precise handling and then layer on assists to help our shooting feel precise and satisfying so we're pretty

much at the end of the presentation now so I'll just summarize some of the key points we discussed today games should balance simulation depth with arcade accessibility in order to achieve intuitive controls with challenge and mastery build your vehicle for the context is intended for and retune if the context changes think about how devious emulation should go look at real world vehicles for inspiration and decide on what parts would add to the depth to the player experience build assists to make the game more accessible bounce the amount and strength of the assists depending on how many game mechanics the players to interact with ensuring the input frequency isn't too high consider camera led and vehicle LED cameras allow the physicality of the vehicle to be conveyed through the camera and give the player the feedback they need and that's it thank you very much [Music]

who any questions hello bring doc thank you very much thanks I've been working on console vehicle systems for a very long time but recently been asked to help out with PC and I was wondering if you had any input or thoughts about digital inputs that are more prevalent on PC controls versus analog sticks and joysticks yes so the question was thank you the question was about digital inputs versus analog inputs and console versus PC so there are a couple of a couple of big things there so firstly if you're using keyboard to input then you kind of you've only got on and off inputs so one of the things that we do for both console and PC is we have steer rates so it's kind of fading in and out our input in terms of you know just turn to maximum and put you're kind of more fade in and out we use that for our console input consult had inputs but we also use it for our PC input in terms of if you

hold down a then they will gradually fade towards that which helps a little bit one of the things that we did for stars Battlefront 2 is that we had a and D would control the role and one of the things you can do on a pad is because you have an analog input you can effectively balance the rolling against the auto roll system so if your system is trying to like level you then you can kind of balance it like quarter percent left input to kind of keep you locked at a particular angle which is totally impossible to do with just AMD so we had a separate kind of input mechanism on PC which is if you were holding down AMD at the same time then we use that new concept to say just lock wherever I am so you could kind of say a two here and then up both of them to this lock in place so we tried to come up with new mechanisms that will suit the PC better than on pad high with sim keyed racing games that you were talking about how do you approach handling and assist when

you're tuning vehicles to be understeery versus over Siri okay so thank you the question was about understeer versus oversteer I think it depends on what you're trying to get what you're trying to get out of your key experience so in terms of you if you want your vehicle to feel quite heavy it's often nice to have a slightly under steering vehicle where you're kind of even if you're staring at maximum you might not be able to make that corner and then you encourage your players to brake turn into the corner drift round so I think a lot of most wanted traveling was a little bit understeery but then it encourages you to like really kick that back out and get into a nice big drift whereas if you want if you don't really want your game to be about drifting and you want to be very much like a little bit more along the lines of like f0 or something works pretty much I want complete control over the left stick and I don't need to be drifting at all it's just maximum

responsiveness then you probably have something that's way more oversteer Ian's gonna like kick into turns really aggressively so you have full control over there that's kinda how I would do it in the demo when you had the landing camera in bottom out and hit the ground on the landing was it colliding elastically with the ground or was just zeroing at its velocity and then the spring force was putting it back so the question was about suspension bottoming out so we have on battlefield one we had for like full regular sort of raycast suspension wheels and then we had a rigid body as well so when you completely bottomed out we detect when it goes above a certain threshold and the suspension of like it's more than 80% compressed and we'd count that as being bottoming out oftentimes we would find that the the actual rigid body would collide and we'd need to try and smooth it as much as possible sometimes you get a little bit

of collision hitching but in general we want to try and keep that maximum 20% of the suspension like very very rigid in terms of the suspension force that it applies back so you still get the effect of you've pretty much bottomed out the suspension but you still remain in that 20% to try and keep it floating off the ground effect stopping the rigidbody from actually colliding so yes we did get some collision hitting and hitting the ground we try to minimize it through a very small amount of very very strong suspension at the top of the suspension raycast does the camera itself hit the ground so it's not so technically with the stuff that I showed in the presentation it's possible future tuned the camera so that it would like lurch below so the momentum would keep it going so it would keep going down so if you had a massive momentum term it could potentially go all the way down below the ground however even if you did tune it like that we have another separate system which is a capsule collision system that basically says hey if you're

gonna collide with the ground kind of push yourself out and resolve it in a way that won't actually go through the terrain so we have another system on top of that to protect against those sorts of issues for the audience's you talked about on the starfighters did you do anything different for projectile convergence and something like the x-wing versus the TIE fighter where there's projectiles are spawning from different points and colliding basic coming from different angles the TIE fighters are very straight for the x-wings are coming from the outer wings hmm yeah really good question so this is a bow you're not actually shooting from where you think you're shooting from you're shooting from you know like gum barrel to the edges of the wings it's more complex than just being far away from the center of the vehicle it's also far away from the center of the camera so one of the most difficult vehicles for this was slave 1 which is like a massive vehicle and the gun barrels are right at the bottom and we needed the camera to be up here so if you've got

the camera way up here and the gun barrels down here you'd be trying to aim along where the cameras looking and the camera the the gun barrels would effectively be pointing up so what we found was there's this really weird behavior where if you were trying to aim at something that was quite close your gun brows will be pointing like up 40 degrees or something so you would be like I want to fire here but then all the all your missed shots are actually going up in the screen and looks really weird and that's actually a very difficult problem to solve because you either say I want to shoot down the camera center or I want to shoot accurately from where the gun barrels actually are so what we did was we did some we kind of fudged it a little bit by looking at what your target is in terms of I'm shooting towards this guy over here so I'm going to try and fade it so that my shots will will always hit him and try and make it so that if he's really really close to me in terms of like if he's right up in

the camera and the vertical deviation from where the gun barrel is actually pointing would be very aggressive and would give me that effective shots going off into the sky then limited to a particular set angle all right and did you also have to account I'm reusing that an accommodation what's the prediction point so that if the player was trying to aim in front of a moving Atari that's moving laterally that it would have to predict in front of them where the shots could converge - yeah so we did all of our calculations based on the predicted position so we know where you're trying to aim and we would aim towards that point so we'd use that to fit it in there now just one more actually sort of a comment and suggestion for people if you're working on racing games or vehicle dynamic systems I'm not sure about the UK but in the United States we have a sport called Auto cross that people do in parking lots and you set up cone tracks and stuff and I would encourage

you to take a look and see if there's you know that sort of an activity in whatever city you live in because driving a real vehicle at the edge you develop a different intuition for the way a lot of these systems work and the vehicle dynamics programmer that I worked with he and I would go out on weekends and participate in you know championships and stuff and it it was great fun but it also really gave us a better you know fundamental just intuition and understanding of the way the vehicles work when they're really at the edge of the performance envelope which most people never experience I mean even if you're go karting or you're doing other stuff there's different you know ways of white transfer and when you actually lose control what happens when you spin out and those sorts of things so just suggesting if people are working on those systems see if there's like a SCCA or Porsche Club or a BMW Club or anything in your area and you know participate in those sorts of things know yeah definitely like actually getting real experiences there because they're super valuable I think whenever we were trying to do a new type of vehicle we would try and actually go and drive something similar unfortunately we

can't get World War 1 vehicles so we could actually drive any of those but it's it's super valuable and also it works the other way around as well like I found that playing a lot of racing games throughout my life made me really good at go-karting I mean not that good but you know for the future turkey bottom I want to go with shooting the guns from not from the camera yes did you do anything to handle where there is an object between the gun and the target but that same object is not between the camera and the target so we didn't do too much oh so sorry this was about if there's a object between the gun and where the target actually is it's kind of obstructing we didn't do too much of that mainly because assets you like the situation of the game is mainly aiming it like a Starfighter which is over here and I'm kind of going through a tunnel maybe is the most geometry we tend to get in in between things we didn't find that players had that a lot in in in battlefront - I'd see a few if you were

doing a first-person shooter that obviously can just like position the camera right up again at the position the gun right up against the wall and you get that situation a lot in which case I think you'd probably something along the lines of doing ray casts from the gun and being like well if that's obstructed then fire the bullet here instead of where I think it should go from the cameras point of view and decide which is most apropos to flickable thanks cool if there's any more questions or anyone wants to catch up afterwards grab a business card or something fun then come grab me afterwards thank you very much