Posts Tagged ‘rts games’

11 Jan

Transfer Learning in Real-Time Strategy Games Using Hybrid CBR/RL

Posted by cognitivecomputing in Game AI, Learning. Tagged: , , , , .

The goal of transfer learning is to use the knowledge acquired in a set of source tasks to improve performance in a related but previously unseen target task. In this paper, we present a multi-layered architecture named CAse-Based Reinforcement Learner (CARL). It uses a novel combination of Case-Based Reasoning (CBR) and Reinforcement Learning (RL) to achieve transfer while playing against the Game AI across a variety of scenarios in MadRTS, a commercial Real-Time Strategy game. Our experiments demonstrate that CARL not only performs well on individual tasks but also exhibits significant performance gains when allowed to transfer knowledge from previous tasks.

Read the paper:

Transfer Learning in Real-Time Strategy Games Using Hybrid CBR/RL

by Manu Sharma, Michael Holmes, Juan Santamaria, Arya Irani, Charles Isbell, Ashwin Ram

International Joint Conference on Artificial Intelligence (IJCAI-07), Hyderabad, India, January 2007
www.cc.gatech.edu/faculty/ashwin/papers/er-07-01.pdf

15 May

Plan Recognition in Large-Scale Multi-Agent Tactical Domains

Posted by cognitivecomputing in Learning. Tagged: , .

This research addresses the task of representing and recognizing events in a tactical domain from large-scale spatio-temporal data under conditions of limited observability and high noise with real-time response constraints.  These assumptions differ from those traditionally made in  plan recognition and produce a problem that combines aspects of plan recognition, pattern recognition and object tracking. This research provides evidence that parsimonious qualitative representations used to represent pair-wise interactions among agents can be combined to identify large-scale group behaviors that form the basis of increasingly complex patterns of activity.

A comprehensive software application was constructed to demonstrate the claims of the thesis by evaluating performance on a real-world problem involving the recognition of a tactical maneuver in actual US Army training battles.  Evaluations were conducted and performance evaluated by both novices and active military subject matter experts.

Plan Recognition in Large-Scale Multi-Agent Tactical Domains

by Mark Devaney

PhD Thesis, College of Computing, Georgia Institute of Technology, Atlanta, GA, 2003
www.cc.gatech.edu/faculty/ashwin/papers/git-cc-03-01.ps.zip
Newer Entries »