Games and Simulations

Machine Learning for Games

ML for Strategy Games

Applications of Machine Learning techniques (incl. Reinforcement Learning and Inverse RL) to implement a human-level AI player of real-time strategy games. Development of open-source tools that make this easier for research community.

• Docker Images for StarCraft & SC Bots

Opponent Modeling for RTS

Opponent Modeling for RTS

Opponent Modeling solutions combining machine learning methods with expert knowledge to infer and predict opponent's actions. Released as an open-source library for StarCraft: Brood War bots.

• Github Repository

GG Bot

GG Bot

A bot for StarCraft: Brood War that serves as a testbed for AI methods we currently deal with during our research (e.g. Inverse Reinforcement Learning, Deep RL, MAS or Genetic Algorithms).

• Github Repository

SSCAIT: Student StarCraft AI Tournament

SSCAIT: StarCraft AI Tournament

We're organizing a 24/7 educational event, serving as a challenging competitive environment for hundreds of AI students and researchers. Participants submit bots programmed in C++ or Java to play 1v1 StarCraft matches.

• SSCAIT Website
• YouTube Channel
• Facebook Group

Mobility Models

Urban Population Modeling

We build and maintain a large-scale, data-driven, multi-agent model of urban population. Compared to traditional approaches, the high level of detail provided by data-driven agent-based models enables answering a wider range of what-if questions, including the impact of infrastructure developments, adoption of new mobility and transport policies or changes in mobility services available.

DRT Simulations

Simulation of Demand Responsive Transport

We study algorithms for demand-responsive transport schemes using our agent-based simulation and benchmarking testbed. It can simulate and evaluate a wide variety of centralized and distributed coordination or resource allocation mechanisms and is avaliable for public as an open source project.

• Testbed on Wikipedia
• Github Repository

Population Tracking

Automated Collection of Mobility Data

We develop software systems for automated collection, annotation and processing of data about the movement and activities of urban population. Collected information is meant to replace surveys and questionnaires as the input for data-driven mobility models.

Urban Algorithmics

Analytics, Visualization and Reporting

Based on our models, simulations and collected data, we prepare extensible analyses, reports and visualizations for a variety of public sector institutions and private companies. For this purpose, we have founded a spin-off company called Urban Algorithmics.

• Urban Algorithmics website