An optimisation-based robot platform for the generation of action paintings

Abstract

We present the robot JacksonBot which is capable to produce paintings inspired by the action painting style of Jackson Pollock. A dynamically moving robot arm splashes colour from a container at the end effector on the canvas. The paintings produced by this platform rely on a combination of the algorithmic generation of robot arm motions with random effects of the splashing colour. The robot can be considered as a complex and powerful tool to generate artworks programmed by a user. Desired end effector motions can be prescribed either by mathematical functions, by point sequences or by data glove motions. We have evaluated the effect of different shapes of input motions on the resulting painting. To compute the robot joint trajectories necessary to move along a desired end effector path, we use an optimal control-based approach to solve the inverse kinematics problem.