Parkour: How to Improve Freedom of Movement in First-Person Games in 20 Simple Steps

[Music] my name is bartosz Coulomb also known as guava I was a little grammar responsible for player mechanics and Dead Island and dying like first of all please forgive me for I have sinned I lied to you I was supposed to talk about 20 steps of improving movement in fpp games but when I actually wrote the content of this speech it was so long that we would end up talking to the end of the conference and we must have time for other amazing lectures so instead I would like to focus on three major topics crucial player mechanics implemented into dying light parkour problems and solutions that we've encountered during their productions and additional details that helped us improve the experience I hope this lecture will help you build an extent freedom of movement in your game

so first of all if we want our players to be able to climb he need to know what he can climb onto so how we describe a climb a ballad with its start and an forward vector this info is enough for him to start climbing at first we had a meeting and spoke about placement detection of those ledges and our first approach was to try put manually disinformation into our geometry our level designers wanted to have control over the places that player could go to but it turned out that we ended up with 50,000 ledges placed on a single map and we still needed more working with so many objects is horrible and impossible to manage we also our memory started to shrink because we initially started developing dying light on an old generation console and memory space in them was [ __ ] low so we decided to

take another approach detect ledges in runtime now how we did it we do a number of forward traces to the tackle and obstacles in front of the play for for those who heat we validate space above and below them to see if the player body will fit by simply measuring the distance of the traces for those who are valid we then do a vertical trace to detect the exact spot of climb being led those three steps has been repeated to the other side of the player because our game is 3d so we and also our ledge definition has a start and an end then we can pair those points with some angle restrictions horizontal average than vertical and create our post plausible ledges and then we can sort them by distance and look angles to the player

and add another space validation with traces y with traces instead of collision collision queries because our physics engine did not support getting results from physics queries in current frame and getting the results from next one would be too late and all that turned out to be the right way to go climbing changed from frustration into pure joy of overcoming all obstacles in players way but this new freedom created two major problems so beware level designers and level artists now have to take care of all places that player could go to and I mean all places rooftop backyards etc everything had to play and look good this was pretty scary at the beginning but eventually they got so excited about the new gameplay possibilities that they ended up loft

working on it they also told me that it actually opened their eyes on true meaning of open-world environment but this is a story for another lecture and I don't feel competent of explaining it the second problem was the performance thankfully we have super intelligent and resourceful programmers that managed to come up with some magical optimizations that included batching geometry around the players and something we call carpet of traces technically we knew that we have a special case of traces with same direction and length so we optimize our tracing algorithm to take that into account and perform less operation on budget geometry around the plate the players I don't know the exact details I'm not so good at math but it did the trick now that we know how to detect a ledge in front of the player we can move to another important thing if you look at

this ledge from above you will notice that there is no space for the player body so player won't grab that ledge but this is invisible from player point of view so we had to include new climbing state hang when we were designing and implementing our first approach we didn't forget about it but we wanted our movement to be fast and have smallest amount of stops as possible later life verified our vision and we had to include that into our ledge detection algorithm so how we did it we took the results from forward traces from previous stage we add a different restriction above and below above like to 10 centimeters for player hand but below it was more restricted 1 because of the player hanging body and pairing and space validation was done in similar fashion to previous one but if the player is hanging and he has some space on his left or right side we now

had to add additional moves to make him feel comfortable with climbing hence we've added side and optionally but now we have to detect places for shimmy and that require additional modifications to the algorithm but I don't want to bore you with a lot of algorithm description so I leave this topic on a lecture on another occasion maybe next year if you like this one now I want to talk about something very simple jumping this is so old mechanic that all players should know how to do it right well as it turns out jumping in fpp isn't as intuitive as jumping it flat platformers the first player perspective works against precision and precision is crucial part of our game is that a major problem well we had test cases when large part of the players would jump too soon or too late all of the time so our

obstacles end up being too easy or too hard for large part of the players to fix that problem we have to give the players additional help this help included three simple steps that we call jump assist first step was to delay jump and wait for a better moment to jump we define better as there is something in front of us like a wall or a Ledge or players just started to fall is not touching the ground that helps the players that jump too soon at full speed that gave us additional 1 meter of jump range the second step was allowing the player to perform a jump after he fall for a brief moment with this one we had to be careful because soon it look really bad in coop and that helped the players that jump too late and third

step if the player is jumping over an obstacle but doesn't really touch it we would still allow him to perform that job technically it's a jump in air but players feel more natural like they are extending their like and bouncing of that object again you have to be very careful because it works on a short range like 20 centimeters from the object those three steps result most of our problems and even a beginner had no problem with jumping so frustration was gone ok the end of part 1 now we can move to problems and solutions that we've encountered during the production animation and freedom of movement when we have detected the ledge in front of the player we will choose his starting position from four positions on straight climbable ledge one position on not climbable that there

is a box that you cannot climb above and to four positions to per site on steep climb Bible ledges this one is tricky so bear with me after choosing correct position after choosing correct animation player position is linearity set with a starting position of the animation with special state we call begin in this state player is reaching his hands and legs to grab the ledge this tail duration is scaled up or down according to player speed before the climb this is due to the fact that we want to avoid changing speed as much as possible so the movement is more fluid in this state we are also forcing player loop direction to blend with forward direction of the climbing ledge when this state is finished we have set the player in the exact same position and

orientation as in the animation so all we have to do right now is synchronize the rest of the movement with the info from the animation BAM simple right here is a sample video of the first implementation some of you may notice that player character doesn't quite fit into the tank like universe

and it looked okay but played horrible why because when we initiated the climb we would actually play a cinematic that takes away control of player hands and no control feels really bad so what we do we started to do a shorter climbing sequence and it pushed us in the right direction but the problem was still there then we added a special state to the animation and when player was still in the middle of climb the camera was still moving up but we allowed the player to walk it felt a lot better but the beginning of the climbing had this issue and we couldn't resolve that in similar manner but what we could do is to free player head from restrictions and allow to look around while climbing

that did the trick even when player body was locked in climbing there was always some degree of control in player hands with that we could rest for a while until new problem showed up something I call straight line out of hell when you see player history position from above you will notice that his line wanted line of movement the red line I hope it's visible it's offset a little bit to the left after the clap this change depends on an angle that he approaches the left the bigger the difference from forward of the climbed Inglot the bigger the changes in theory this shouldn't be a big deal but as it turns out my beloved colleagues from work are very creative they have created some weird-ass obstacles that requires extra precision

from player control and that precision was not possible in current implementation thankfully solution was simple allowed the player to climb on his current forward vector instead of the one from the ledge that fixed one problem but created two new ones the red ellipse represent presents the movement from the animation as you may notice it ends before the climb Inglot this is due to the fact that the animation has been prepared facing directly the forward of climbing laps so players started to fall before the climb being glad we fix that by slightly restricting the angles of climb so ant position would then differ more than player capsule radius and allow physics to climb on higher stair hype midair the second problem was the hands they were grabbing the wrong places one was going into geometry the other

one was grabbing the air the first one looked easy render player hands at the end of the frame it looks easy but as we use deferred rendering in our engine it would be very performance heavy so what we did we've cheated we scaled down the hands of the player with camera projection matrix and it did the trick have a look how it looks from fvp and from the outside we found a missing link a proof that people evolved from t-rex the second hand used some trick compute each position in screen space and use hand I K to set the hand in 3d space it can be too far from the original position as the hand would stretch unnaturally what we loosened then the animations a bit and we got like 40 centimeters of freedom that did the trick

last but not least I would like to talk about problems that we had with jumping it's important for me because we had many discussions about it and we've resolved our problem in some maybe controversial way first question is why we change the regulary used jump button from X to R 1 our game has very complex geometry so we figure out that we would allow the players to have full control of the player even in air so air control the factor in air was set to 100% because we wanted the players to move as fluid as possible through our geometry but if the player wants to press the X button he will move his tongue dump to from right stick to press it that means he's losing control of player for a brief moment the pacing in dying

light is sometimes unforgiving so even our short lack of control is bad so in the end we moved the jump button to r1 as it grants more control over the players and fixed the problem the second complain about jump button was to hold or not to hold it imagine this situation you have two ledges in front of the player one in the window and one above him as they have described while searching the ledge we define which one is better by comparing the distances and look angle but the distance in this case is equal because we measure it in 2d so how we choose between them by slightly looking above one of them the problem was no matter how I tweaked this algorithm to implement that slightly it never worked perfectly for some people it did and for some people it didn't what really helped is adding the restriction to hold the button to grab the ledge you click you jump you grab

the button you grab the ledge it took some time time to teach the player how it works but eventually remove all of the mention problems at some point we were also experimenting with something similar to Assassin's Creed play style when holding the trigger would switch player character into freerunning state in this state player would jump and grab ledges automatically and as a prototype it worked amazingly good but it took away all player skill and satisfaction of performing a perfect time jump don't get me wrong I love Assassin's Creed franchise but Assassin's Creed for me is all about the joonie the world and the set the setting and we wanted our game to be more about second to second gameplay so action in every aspect of game was crucial ability to make mistakes to master the movement

was important this is something I noticed when I play old prince of persia sense of time versus prince of persia 2008 the first one had all moves jumping while running monkey bars etc preset to different buttons so it felt very acrobatic to overcome obstacles harder but again more satisfying in Prince of Persia 2008 all of those mentioned moves was moved to one button so the difficulty went down and I feel the freerunning state took that simplification even further but this is due to the fact that Assassin's Creed and Prince of Persia are different games and we figure out we shouldn't go that way either the end of Part 2 small break okay now I want to talk about something slightly different vomiting

this only happens to me when I party too hard it never happens to me after gaming so imagine my confusion when one day my colleague came to me and told me this story dude I was showing our game on game event to one of the reporters from polygon suddenly she turns pale runs out of the booth and vomits he was in short later that day he even wrote him an email and tell him that she felt bad for the rest of the day that was a disaster we couldn't let that happen you don't wanna report a run of out of your booth you don't want the players to feel sick when they are playing your game unfortunately we didn't have anyone with that symptoms in our company yet well this was like 2014 so no vr stuff and motion sickness wasn't that common thankfully we've read some articles about motion sickness and prepared a built with

abilities of turning and tweaking things that could have influenced a motion sickness big help came from specialists in that manner because we have requested a report in which a lot of test was conducted I hope not on students only and they figure out what caused motion sickness in our title it turned out that frequent and often cover a target movement caused motion sickness the left one shows before the corrections the right one was after the corrections in the animations camera moving backward when player body was moving forward that created an awkward feeling players had so we had to remember to never go with camera back when moving when performing a while fast-forward moves the highlighted area shows that on the left side there is a slight bump backward and

we eliminated that in the animation and that helped us third thing is player not being able to focus on constant element on the screen adding a simple dot there fix that it's even invisible for you but your eyes automatically rely on that constant element and amortizes all of the movement so it really helps lasting but not last but not least is the distance between the TV and the player when player is sitting too close to the camera his brain can stand even the slightest movement of the camera while we couldn't feel that at player homes we could definitely fix that on get next game event Techland booth became larger and we made sure that the distance between the player and the TV is large the result no one felt motion sickness after playing deal on events even the journalist that pointed out this problem

so it was a massive relief this task was very frustrating for me as I was fixing problems that I didn't quite understand and I couldn't see really see the results of my work but eventually we got very little negative comments about it so I guess we did a good job we're running out of time so I will now give a list of things that we've implemented also it will explain why we did it what I won't tell how you have to do it for yourself when you see parkour videos you will notice that freedom and fluidity of movements come to the runners very naturally this is due to the fact that they have very good and well-trained knowledge of their bodies that helps when they are running on narrow walkways when they are jumping on knee height obstacles or move their shoulders to fit into narrow ways that kind of special

awareness it's impossible to simulate in FTP games player often lose obstacles from the rear sight due to limited fov or input slightly wrong direction and falls from Heights or bumps into us to compensate that problems we created a couple of systems that help the player balance if the player is walking on narrow object or near the ledge of the building we would slightly correct his input so he wouldn't fall avoid when player is running we would add site speed to his movement to avoid make him fall into walls and does make him run much smoother use you if you look you see will a little bump to the right left this is artificial player inputs only forward direction and there is no collision like we prevent collision from happening so there won't be any slowdown and out of step if the player is has a small obstacle in front of him like 40 centimeters we would automatically climb

bond it because those obstacles when he is closed them are invisible for the players so it feels very natural to walk on them so far I was going I was talking about moveset about things that help the player move up or forward but professional runners can do a lot of different stuff to go down to help the player move down we've added several systems that allowed us to build the true image of free runner with added jumping into dumpsters and braking falls landing on cars jumping into water and swim slide down ladders and pipes using a zipline and enjoy the view or

even break Falls by rolling those mechanism had to be used by level designers to create fun and variate environment and as a wrap up where to go from here well sky is the limit one could add turning on shimmy so parkour puzzles would be even more interesting outside wall run with some clever level designed to support them at environmental awareness so player would run when running through near obstacles would touch walls and other things sky's the limit don't give up and you will succeed at the end I would like to thank all of the people that work on creating dang right all of them deserve a round of applause so give it up for them [Applause] you