We plan to develop and implement an AI for the skeleton mob in minecraft that will use a bow and a sword to efficiently kill friendly mobs such as sheeps and cows and hostile mobs such as kreeper and zombie. As an addition to our main idea, we plan to combine our AI with the AI of another group who is creating a defending AI, to see who is going to win, theirs AI defender or our AI offender.__ The first option: as an input, we will give a map grid to our AI which will represent where the mob is located. The second option: as an input we wiil use a first view camera, analyze an image and find the mob that we want to kill. As an output, our skeleton will move towards that mob (our target), shooting at it and when coming close enough hitting the mob with a sword.
AI algorithm: we are not sure what algorithm we are going to use but for now we decided to use reinforcement learning with neural function approximator. However, we might change our selection.
Quantitative evaluation: firstly, we will evaluate how the skeleton is shooting the arrows (e.g. if our skeleton is able to hit the static target from different distances). Then we want to be sure that our skeleton is able to hit the moving target. Then we will test our skeleton with the hostile mobs such as kreepers or zombies. Then we will test our AI with several mobs. We will also record the time it takes to kill one or more friendly mobs. Obviously, less time it takes to kill a friendly/hostile mob is better.__ Qualitative evaluation: count how many hostile mobs we will be able to kill before our skeleton die. We will gradually add more hostile mobs and see what is the amount of mobs our skeelton will be able to kill. Ideally, our skeleton should be able to kill at least 5 hostile mobs and then die.
Appointment: 01:00pm - Tuesday, May 2, 2017. DBH 4204