GDC 2014

GDC 2014 Session Scheduler

View, browse, and sort the ever-growing list of GDC sessions by pass type, track, summit, format, and day. With GDC Session Scheduler you can build your schedule in advance, and access it during the show via export or with the GDC mobile app, Please note that adding sessions to your schedule does not guarantee you a seat in that session. Sessions do fill up, so please arrive early to sessions that you would like to attend.

Start building your GDC agenda — create your linked account for the GDC Session Scheduler and the GDC Mobile App now!

Register now for GDC 2014!

Please note: Registration accounts are separate from Session Scheduler. Even if you've registered, please create a Session Scheduler account below.

Export Schedule:

The Last of Us: Human Enemy AI

Travis McIntosh  |  Lead Programmer, Naughty Dog, Inc.

Location: Room 2005, West Hall

Format: Session
Track: ProgrammingDesign
Vault Recording: Video

This talk covers the technical and design decisions made to produce the human enemy AI as seen in The Last of Us. It covers the systems that formed the core of the AI decision making, as well as a higher-level look at the behaviors that formed the building blocks of enemy combat and stealth. Navigation, real-time level analysis, combat, search and stealth are all covered in detail, as well as mistakes and future improvements. Finally, the performance characteristics of the AI and various optimizations are discussed in detail. Consider this talk a description, then a postmortem of the human enemy AI.


Attendees will take away knowledge of how The Last of Us AI made high-level decisions in regard to combat, searching and stealth, as well as an understanding of the function and implementation of the systems that supported those decisions.

Intended Audience

This talk is intended for programmers and designers with an interest in combat AI. Although some technical knowledge is helpful (particularly knowledge of AI pathfinding techniques), anyone with a cursory knowledge of high-level AI should be able to understand the majority of the talk.