Showing results for 
Search instead for 
Did you mean: 

PLAYSTATION EXPERIENCE, DLC AND AI

Between the Call of Duty World League qualifiers, the announcement of our first DLC and a panel at PlayStation Experience, it was a busy weekend...
By John_Rafacz on December 8, 2015
John_Rafacz
Level 3
‎08-12-2015 03:20 PM
‎08-12-2015 03:20 PM

It was a busy weekend.

Hopefully, you were able to catch the Call of Duty World League action where the members of this year’s Pro Division were solidified. (Stay tuned for a re-cap.)

We were also pretty busy at PlayStation Experience in San Francisco – and it all started with the announcement of our first DLC map pack for Black Ops 3: Awakening. Revealed during Saturday’s keynote address, it was announced that Awakening would arrive first on PlayStation 4 and feature four new Multiplayer maps: Skyjacked, Rise, Splash and Gauntlet, as well as the next chapter in the Black Ops 3 Zombies saga, Der Eisendrachen. Check out a full recap at the Activision Blog.

There were also several members of the team that were on hand to share with you some of thought process about what went into making Call of Duty: Black Ops 3 with a panel we did on Unlocking the Secrets of AI. A multidiciplinary group representing the team were:

  • Gary Stelmack, Senior Designer
  • Sumeet Jakatdar, Senior Software Engineer
  • Yanick Lebel, Lead Animator, Animation
  • Craig Houston, Lead Writer

Moderated by Activision’s Scott Lowe (@ScottLowe / @Activision), the team dove into the design decisions and implications associated with the studio’s decision to create a four-player co-op game. A complete VOD can be located HERE.

In the meantime, here is a bit of insight, directly from the team…

BLACK OPS 3: Secrets of AI

In Black Ops 3 (BO3), we set out to create larger, open combat spaces where players will get to choose how to engage with enemies. To populate these bigger spaces, AI design was required to have more variety of enemies with unique behaviors with defined roles to play in combat. We needed new AI Architecture to define their individual characteristics and behaviors.

This Architecture is called as “Archetype Architecture.”

Archetype System.png

An Archetype can be thought of as a container of available behaviors, behavior tree and animation states. Multiple “Variants” can be created using an Archetype by choosing between available behaviors of an Archetype. A Variant defines additional attributes such as weapon, character and combat attributes such as engagement distances which add flavor to the combat. In addition, Variant also specifies the animation asset table. Given that the Archetype definition does not contain any animation references, multiple Variants can be created which look and animate differently.

Designers can easily create multiple Variants of a given Archetype based on their needs. Like in the figure above, a Shotgun Variant of Human Archetype is choosing all the behaviors from except flanking. This Variant also includes movement and engagement attributes which are tweaked such that he will try to remain in closer range of the enemy.

Here is the screenshot with few Archetypes working together on in a battle in Cairo. In situations like these, players fight against humans, bipedal robots, aerial robots, big tanks and many of their Variants. When equipped with Cyber core abilities, players can hack and control some of these Variants anytime they want. Player can choose to break robots into multiple pieces before killing them. We try to incorporate unique reactions to player’s actions, which we hope will make these engagements more interesting.

Screen Shot 2015-12-07 at 1.55.54 PM.png

How it all works?

Unified AI

Previously, different versions of AI Architectures existed within the engine for different gamemodes to suit the individual needs of that mode. Take the Zombies Archetype for example, which was designed such that it will work in a networked environment specifically for cooperative Zombies experience.

Going forward, we needed an AI Architecture which will be agnostic to the game modes so that a given Archetype can be used anywhere design needs. This means we had to overhaul multiple aspects of AI Architecture. The BO2 AI Architecture managed behaviors and animations in the same place and was written considering a smaller number enemy types in mind. This needed to change so that we can create more Archetypes without writing new code.

In BO3, we broke down the behaviors and animations in two separate modules; Behavior Module and Animation Module.

AI Architecture.png

The Behavior Module is responsible for selecting the appropriate behavior at a given time and it is purely Server side. It communicates with the Animation Module by choosing a high level animation state. The Animation Module is then responsible for manage transitions and managing animation layers. The animation module is also networked to keep the Clients in sync with the Server.

At a given time, a Variant can have multiple animations playing on top of each other. The Animation Module manages animation states and their logic. An animation state definition is a container of individual animation layers. There is always one full body animation layer active at a time, but more than one additive layers can be present. Both, the Client and Server are responsible for updating these layers using the same animation logic and animation state. This means that we can add many more animations without sending more information across to the Client.

There can be simpler, “Client only” layers that can be active as well. A good example is the Facial Additive Layer.

Design and Animation

The new Architecture gives flexibility to the designers to use Archetypes in game modes without restrictions, helping them create a richer experience. The G.I.U. Robot kill streak is a good example in Multiplayer. By and large it is just a Variant of the Robot Archetype used in Campaign mode.

Because of the separate Animation Module our animators can author and plug in animations independently and create Variants. For example, the male soldiers and female soldiers are Variants of the Human Archetype but behave slightly differently and use completely independent animation assets.

For the first time in the Black Ops Series, we could create fully animated vehicles using the same AI Architecture. Take the A.S.P.-C for example in Cairo. It is a futuristic tank with multiple turrets which can maneuver the environment using the parametric animation blends.

In fact, on BO3, every AI animation is authored completely from scratch, and as mentioned earlier, some of our Archetypes have thousands of animations.

Combat Spaces, Player Movement and AI

There are more open and wide combat spaces in BO3 than any prior games in the Black Ops series. The combat design process is evolved such that designers choose Archetypes that they want to use and only control their high level attributes. Designers can iterate faster and try to tune the experience for every type of the player and style of approach.

BO3 features a fully revamped player movement Architecture allowing them to double jump and perform wall runs. One of the goals was to cope with the player’s enhanced mobility. The result was complete overhaul of the AI navigation Architecture. This Architecture has full support 3D navigation and underwater Pathfinding as well.

One of the goals of BO3 is to create a diverse combat experience. Bigger open spaces, AI Archetypes and Cyber Core abilities are the few of the many pillars of that experience.

We hope you have fun when you play BO3. There is more to talk about and more to explore.

Keep playing and stay tuned!

- Treyarch Development Team

NOTE: Main Photo Credit - Sony / PlayStation Experience