The system has a camera. Our software stream images from web cam to a computer and processes image by the help of image processing techniques. After image processing, the arm determines which physical coordinate of an item that human inserted.
By term AI which I have used is not the reaction of arm for some type of specific inputs. So as you mentioned before a short set of basic rules can be adequate for making an interaction. In our system, we collate some useful information by the help of camera. We process this information and create a searchable game tree which consists of all possible game board states. At this point the AI itself takes place. Searching through all possible board states, deciding which move is better or worse and providing an output which can be called best move is the actual thing AI does. So in software litrature there are so many accepted decision making AI algorithms. We are using one of them. So it is impossible to win againts it because mathematically it is impossible, it is proofed.
As a gripper we have used a programmable electro magnet which is specially designed for this project. Behaviour of gripper can be controlled with our software.