The plotter in the picture below was designed to draw on uneven surfaces. It is a prototype which will be part of a system for marking cracks on the surface of combustion chambers of turbines (as used in large commercial aircraft; shown below). The robot can reach into the combustion chamber without the need of dismantling it or adjusting it to the features of the combustion chamber. The unevenness of the surfaces to be marked is compensated by two hinges allowing the arms to fold up.
A program running on a PC controls the robot in real-time. Currently, a camera system, through which the robot can be controlled via a mouse(-pointer) on a computer screen is developed.
The spring-and-strings-only robot tries to do away with conventional bar-type and other stiff linkages. Inside the spring are three strings, which are reeled in by stepper motors. This allows to retract the spring as well as to bend it into a chosen direction. The yellow "stick" in the middle is a laser pointer, which is used to draw on a wall.
Experiments show that the whole mechanism is essentially working. It has 2 major disadvantages:
Currently, the principle of this robot is used to create a crawling robot (using 2 strings and two servos).
The strange looking device is intended for course to demonstrate how a genetic algorithm can be used to learn to control a rather complicated mechanical system. In the tower-like part on the left, a hammer suspended in the middle of a 5-bar linkage controlled by 2 servos. A third servo pulls the hammer up and then lets it fall so it knocks a ball towards the targets on the right.
On a PC, a genetic algorithm is used to optimise the position, orientation, and the height the hammer falls such that the balls falls at a given position into the drain at the right and then rolls back to the hammer. The video (12MB) shows one cycle of this. The genetic algorithm (a very text-book like implementation) is able to learn in 5 to 15 generations with 4 individuals to bring the ball on target.