K10! Login

New User     |     Forgot Password     |     Resend Activation Mail

Game AI Contest

Tired of writing and re-using the same path planning algorithm in all game AI competitions? Here is something different.

Last year, we conducted the Game AI competition with the idea of applying AI techniques to game AI. But the recently concluded CIG produced this



Developing AI for Super Mario was one of the competitions at last year's CIG and the video above shows the winning entry. It was developed using the A* Path Planning algorithm. Ring a bell! So this time, we have the same challenge, except that we encourage participants to stick to path planning algorithms and variations thereof to develop AI for the Lunar Lander Game

The Story

The ISRO decides to train their astronauts for their missions due to the repeated failures of space missions worldwide. ISRO chief, Rabdul Ralam has been asked to handle the training. He feels that landing on the moon is a great problem and decides to train all aspiring robonauts on the landing techniques. The Indian government gripped by recession, fearing further economic downturn, issues strict orders to cut down the expenses of the training. So the chief decides to give VIRTUAL training to the robonauts. He hired ME to generate different landscapes resembling the lunar environment. Incidentally, no robonaut has landed on the Moon before. So will you take up the challenge?


The Game


Your goal is to program this robonaut to safely land on the surface of the VIRTUAL moon. 



Jan 17 : Indiscriminate reduction in fuel. Corrected
Jan 19 : Bad communication with C++. Corrected
Please download the new version. Click 
this to download the API

Click this to download the required documentation


  • Game AI contest is a team event.
  • Maximum  team size is 3.
  • Register your team here.
  • Mail your code in zipped format along with necessary documentation on how to run your code to, ai [at] kurukshetra [dot] org [dot] in, before January 20, 23:59 hrs
  • Judging will be objective. The program which clears the most number of levels will be adjudged the winner. In case of a tie, we'll look at the code. Obviously, the simpler code wins. 
  • As a benchmark, last year's winning entry cleared five levels.

Q. I do not know Java, can I still participate in the event?

A. You do not need any knowledge of Java. Your code can be in C/C++ or any other programming language. For C++ programmers not acquainted with network programming, sample codes are provided. Further, you may use the forum or email us if you need any help.

Q. I do not know networks programming, is it needed?
A. You do not need to know networks programming at all. Use the C++ and Java code provided. We would be more than happy to help you with the network programming part for other languages. You can just use the send and receive procedures we have provide.
Q. OK. I read the problem statement, where do I start?
A. Once you've gone through the documentation, start by experimenting with the sample player, we have provided, in the llplayer folder.
Q. Why have you provided the code? Couldn't you have just given us a jar file and be done with it?
A. We believe that treating the game as a black box while developing AI for the same is a purposeless pursuit. This way you are free to nose around the game as you like and understand it thoroughly (only if you feel like it of course) before you develop the AI.
Q. Is this a real game?
A. Of course, It was one of the earliest games ever built. Play it online here. See, it even has its own wiki page
Feel free to post other queries in the forums

Sanjeev S

For any doubts or queries, mail to ai [at] kurukshetra [dot] org [dot] in

We strongly recommend the forums to get any doubts clarified. Access the AiGameDev Forum here 


The winner of the Game AI Contest is Srikanth Doss,  BE CSE Final year, from Coimbatore Institute of Technology

And the runner up is Sudharssun, BE ECE Third year, from SSN College of Engineering, Chennai

Congratulations Guys! We'll get you your prizes soon