Premium Interviews

Starcraft AI from Build Orders to Unit Micro-Management
Premium Interview January 14th, 2014

Starcraft AI from Build Orders to Unit Micro-Management

This interview with Dave Churchill covers his winning entry in the Starcraft Competition 2013, known as UAlberta Bot. In particular, you'll hear about a real-time solver that can determine near-optimal build orders given specific goals. You'll also dizcover about different ways to do large scale unit battle management of units (from UCT to greedy search), and the efficient simulator called SparCraft required to ...

Applying Monte-Carlo Tree Search (MCTS) to The Octagon Theory
Premium Interview January 8th, 2014

Applying Monte-Carlo Tree Search (MCTS) to The Octagon Theory

Monte-Carlo Tree Search is a promising technique that is revolutionizing board game AI. In this interview, find out how MCTS can be applied to The Octagon Theory, a mobile game which combines challenges similar to Othello or Go. What's necessary to build a competitive AI with probabilistic search techniques? How do you take into account opponent models into the process?

AI in FORCED: Building Enemies for Tactical Arenas
Premium Interview December 12th, 2013

AI in FORCED: Building Enemies for Tactical Arenas

How does a small team of two programmers embrace change and build an AI that's fun to play against? In this interview with Kenneth Harder from BetaDwarf, you'll find out how the AI was iteratively made to be more challenging yet easy to understand by the player. You'll also learn about the techniques used to implement third-person tactical action AI with tight constraints in Unity.

Planning Domains and Compiling HTN to C++
Premium Interview December 5th, 2013

Planning Domains and Compiling HTN to C++

In this masterclass with Alexander Shafranov, winner of the Capture The Flag competition with an HTN planner, you'll learn about plan compilation to C++ using a JSHOP2-inspired architecture. Find out how to reduce time spent planning for team and individual coordination using domain specific compiled code rather than an interpreted solution.

If You Can Build a Behavior Tree, You Can Dodge a Ball
Premium Interview November 14th, 2013

If You Can Build a Behavior Tree, You Can Dodge a Ball

In this interview with Emil "AngryAnt" Johansen, you'll learn about applying behavior trees to a problem reminiscent of RTS games, in particular a dodgeball scenario. Emil will share insights from the creation of those trees and the recent release of his Behave 2 tool for Unity.