• Back

Blog

QOMPLX Intelligence: Grid Based Map Extraction with StarCraft

This is the fourth in a multi-part QOMPLX Intelligence series that examines how QOMPLX uses the real-time strategy (RTS) game StarCraft as a training ground for research to support advanced reinforcement learning and effective decision making using machine learning and artificial intelligence. Download a PDF copy of  Part 4: Grid-Based Map & Feature Extraction. We have also provided links to the previous installments of this series below.


Like other researchers, QOMPLX's team is using Blizzard’s StarCraft real time strategy (RTS) game for experimentation and concept validation in areas like reinforcement learning, multi-agent reinforcement learning techniques, and select experiences in applied research supporting these improved decision-making goals.

In recent posts we've been digging deep into our team's ongoing research and development of machine learning and artificial intelligence, leveraging various aspects of StarCraft. For example in our last installment, we discussed approaches to  pathfinding in partially observable environments like StarCraft's "fog of war." Previous discussions have explored topics like strategy selection and build order selection.

Mind the Terrain!

In our latest installment in this series, we take a look at how terrain features in StarCraft can affect pathfinding challenges.

Our previous discussion talked generally about pathfinding in StarCraft and the impact of different learning algorithms on path finding and efficiency. But StarCraft is a game that operates in three dimensions. Map terrain is common including features such as rivers, impassible tiles and neutral buildings and units. Depending on their destination, ground units have to find either a way around imposing terrain, or over it (say, via a ramp or by flying).

3_mobile_units
Knowing the position and the range information helps a pathfinding routine discover the best position from which to fire on an opponent. 

In short, terrain is a critical factor in pathfinding within the game. In our research, in fact, QOMPLX found that terrain features, especially altitudes, are valuable in path finding exercises.  Regardless of how the pathfinding algorithms proceed, capturing such features is the necessary first step.

In this installment of our QOMPLX Intelligence series, we discuss how we facilitated pathfinding in StarCraft by constructing a grid system to capture map features effectively.

Use the download button below or the following link to download and read Part 4: Grid-Based Map & Feature Extraction in PDF format.


Additional Reading

Part 1: Build Order Selection in StarCraft

Part 2: Opponent Strategy Identification Using StarCraft

Part 3: StarCraft: Pathfinding in the Fog