A DAPTIVE I NTELLIGENT AGENT IN REAL - TIME STRATEGY GAMES An Introduction
P ROJECT M EMBERS Omar Enayet Amr Saqr Ahmed Atta Abdelrahman Al-Ogail Dr. Mostafa Aref Dr. Ibrahim Fathy
A GENDA Problem Definition Theoretical Areas of Problem. Project Domain. Specific Problem Definition. Motivations. Objectives. Project Background Survey & Approach. Domain Platform. AI Engine Architecture. Expected Deliverables. Development Tools. Project Time Plan. Web Resources & References.
T HEORETICAL A REAS Learning Make the machine learn.
T HEORETICAL A REAS Planning Plan then re-plan according to new givens.
T HEORETICAL A REAS Knowledge Sharing Let everyone know instantly what you knew through experience.
T HEORETICAL A REAS - S UMMARY Learning Planning Knowledge Sharing
P ROJECT D OMAIN RTS Games Real-Time Strategy Games.
P ROBLEM D EFINITION Predictability Computer Opponent actions easily predicted.
P ROBLEM D EFINITION Non-Adaptability Computer Opponent doesn’t adapt to changes in human actions.
P ROBLEM D EFINITION Static Scripts Computer AI relies on static scripting techniques.
P ROBLEM D EFINITION Experience Loss The Absence of sharing experience costs a lot.
P ROBLEM D EFINITION - S UMMARY Predictability Non-Adaptability Static Scripting Experience Loss
M OTIVATIONS Interested In Machine Learning
M OTIVATIONS Interested In RTS Games
M OTIVATIONS Meets Our Career Ambitions as AI Programmers
W HY THIS D OMAIN Rich Environment. Severe Time Constraints – Real-Time AI – Many Objects – Imperfect Information – Micro-Actions
W HY THIS D OMAIN Active Research Theses and Papers are from 2003 to 2009.
W HY THIS D OMAIN Wars Simulation Research in this field contributes to the modern warfare Research. Half of The USA’s Army will be robots in the coming years.
O BJECTIVES Adaptive A.I. Making the Computer Opponent adapt to changes like human do.
O BJECTIVES Mobile Experience Making Sharing Experience Possible Among Machines
S URVEY Recent Papers We Collected more than 30 papers concerning this field, different in their way of approaching the problem and the techniques used to solve the problem. Examples of them are above ^ Adaptive Reinforcement Learning Agents in RTS Games Case-based planning and execution for real-time strategy games. Transfer Learning in Real-Time Strategy Games Using Hybrid CBR/RL
S URVEY Asking Experts Alex Champandard, Eric Kok and more. Alex ChampandardEric Kok
S URVEY Reference Book The Book “AI Game Engine Programming” talks about the drawbacks in learning and planning in RTS Games.
A PPROACHES - T ECHNIQUES Reinforcement Learning A Sub-Science of Machine Learning.
A PPROACHES - T ECHNIQUES Case-Based Planning Planning using Case-based reasoning.
A PPROACHES - T ECHNIQUES BDI Agents Tech. Beliefs-Desires-Intentions Agents.
T ECHNIQUES - S UMMARY Reinforcement Learning Case-Based Planning BDI-Agent Tech.
A PPROACHES - L ANGUAGES C++ The Main Language our Open Source Game is coded with.
A PPROACHES - L ANGUAGES 2APL An Agent-Oriented Language.
A PPROACHES - L ANGUAGES LUA A Scripting Language widely used in Video-Games.
P LATFORM – N OT C HOSEN ORTS An Open-Source RTS Game
P LATFORM – N OT C HOSEN Wargus An Open-Source RTS Game based in Stratagus Game Engine
P LATFORM – T HE C HOSEN O NE BosWars An Open-Source RTS Game based in Stratagus Game Engine
P ROJECT T IME P LAN
G AME E NGINE AI Engine AI Engine That’s our guy
AI E NGINE A RCHITECTURE
E XPECTED D ELIVERABLES Enhanced AI Engine An AI Engine which makes the computer behavior in the game as human as possible.
E XPECTED D ELIVERABLES Experimental Results Comparison of the results of the enhanced AI Engine with ordinary static AI.
Visual Studio 2008 Professional Edition 2APL Environment C++ Libraries : Boost, Guichan..etc. D EVELOPMENT T OOLS
Project Blog : SVN Repository : W EB R ESOURCES
Book : AI Game Engine Programming Book : Artificial Intelligence for Games The Most Important Papers/Theses : Thanks R EFERENCES Eric Kok - Adaptive Reinforcement Learning Agents in RTS Games –– Master Thesis – University of Utrecht Santi Onta˜n´on, Kinshuk Mishra, Neha Sugandh, and Ashwin Ram. Case- based planning and execution for real-time strategy games. In Proceedings of ICCBR Manu Sharma, Michael Holmes, Juan Carlos Santamaria, Arya Irani, Charles Lee Isbell Jr., Ashwin Ram: Transfer Learning in Real-Time Strategy Games Using Hybrid CBR/RL. IJCAI 2007:
W EB R EFERENCES Will-be-Half-Robots-Half-Human-by html usa-robot-future-american-army-videogame- soldiers-machine