A new AI challenge for collaborative AI using Microsoft's Project Malmo platform has just started on Crowd AI, a platform for open data science challenges built, and currently maintained, by the Digital Epidemiology Lab at EPFL in Switzerland.

Project Malmo (which is built on the popular multiplayer game Minecraft), is an AI research tool for investigating how to train intelligent agents to collaborate. It originated in 2016 as an open-ended platform to advance the state of the art in AI research, especially reinforcement learning in a complex world.

As Nikos Vaggalis reported last year, when the first Malmo Collaborative AI Challenge was launched.

It all started with Microsoft researchers trying to make a Minecraft character climb a virtual hill. This might not sound a big deal, but the difference was that this AI agent tried to overcome its hurdle through learning and interacting with its environment, not by being programmed to do so.

Last year's inaugural challenge, which was designed to push the state of the art of collaborative AI for reinforcement learning, asked teams of PhD and masters students to solve a Minecraft game using collaborative agents. The game concerned, Catch the Pig, is based on an extension of Stag Hunt, a classic example modeling the trade-offs between choosing to collaborate and going solo.

More than 80 teams comprising postgraduate students from 26 countries entered the challenge. The winning entries have been discussed in a Microsoft Reseach blog post and one of the winning teams, from Nanyang Technological University, published an AAAI paper on their approach combined a generalized agent type hypothesis framework with a novel Q-learning approach to learn effective policies against each type of the collaborating agents.

Now Microsoft is partnering with Queen Mary University of London and CrowdAI to co-host a second competition, MARLO 2018, or to give it its full name, "Learning to Play: The Multi-Agent Reinforcement Learning in MalmO Competition". 

As before the challenge tasks, ranging from competitive Build Battles to collaborative Pig Chases, for this contest and sample code are available on GitHub. The GitHub README.rst explains the relationship between Marlo and Malmo: 

The Malmo platform provides an API which enables access to actions, observations (i.e. location, surroundings, video frames, game statistics) and other general data that Minecraft provides. Marlo, on the other hand, is a wrapper for Malmo that provides a higher level API and more standardized RL-friendly environment for scientific study.

It also states that the framework is an extension to OpenAI's Gym framework which has become an industry standard and a familiar platform since we first encountered it in 2016.

The Qualify Round for this contest began on July 27, 2018, and ends on October 7, 2018 and already over 50 teams have signed up. To join in you have to be enrolled in postgraduate studies and be 18 or over. First need to sign up at CrowdAI and then head over to the contest repo on GitHub and fork the starter kit there. At the end of this round 32 teams will be selected to be invited to the final tournament for a live competition to be held at AIIDE’18, the 14th AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment at the University of Alberta in Edmonton,  Canada on November 14, 2018. The 32 invited teams will be assigned to eight groups to determine a ranking, and the top teams in each progress to the next round, the quarter finals. The process repeats for the semi-finals and the grand final.

At the end of the contest, the top seven teams will be awarded a Travel Grant with a maximum value of $2,500 USD to join a relevant academic conference or workshop. Additionally, the top team will be awarded a second MARLO Travel Grant with a maximum value of $2,500 USD to join the Applied Machine Learning Days 2019. Three winning teams will be awarded Microsoft Azure Sponsorship with a maximum value of $10,000 USD for the 1st place, $5,000 USD for the 2nd place and $3,000 USD for the 3rd place.

Last year's contest proved the concept behind this contest and in moving from Malmo to Marlo even more advances can be expected in the field of collaborative AI. If you are a postgraduate student and AI is your area of research this could be both fun and worthwhile.

More Information 

Presenting the winners of the Project Malmo Collaborative AI Challenge

Project Malmo

Project Malmo on GitHub

HogRider: Champion Agent of Microsoft Malmo Collaborative AI Challenge

Related Articles

The Malmo Challenge - Collaborative AI

Malmo AI Available On Github

The AIX Minecraft Project Makes Thinking Software Possible

OpenAI Gym Gives Reinforcement Learning A Work Out

To be informed about new articles on I Programmer, sign up for our weekly newsletter, subscribe to the RSS feed and follow us on Twitter, Facebook or Linkedin.

Comments

Make a Comment or View Existing Comments Using Disqus

or email your comment to: comments@i-programmer.info