Computer Creativity in the Automatic Design of Robots

Abstract

This article demonstrates the possibility that robotic systems can automatically design robots with complex morphologies and tightly adapted control systems at a low cost. These automatic designs are inspired by nature and achieved through an artificial coevolutionary process to adapt the bodies and brains of artificial life-forms simultaneously through interaction with a simulated reality. Through the use of rapid manufacturing, these evolved designs can be transferred from virtual to true reality. The artificial evolution process embedded in realistic physical simulation can create simple designs often recognizable from the history of biology or engineering. This paper provides a brief review of three generations of these robots, from automatically designed LEGO structures, through the GOLEM project of electromechanical systems based on "truss" structures, to new modular designs that make use of a generative, DNA-like representation.

Traditionally, robots have been designed on a hardware-first, software-last basis: Mechanical and electrical engineers design complex articulated bodies with state-of-the-art sensors and actuators and multiple degrees of freedom; the next task is simply to "write the software." But robot designers have drastically underestimated the difficulty in writing control software and this, in addition to the high costs of designing and building the hardware, has led robot development to a stasis [1]. Modern commercial robots perform only simple and highly repetitive manufacturing tasks with very little decision-making, if any, by the onboard software [2]. The central issue addressed by our work is the capability to design robots of more complex structure and more onboard intelligence, yet at a lower cost. We suggest that this can be achieved only when robot design and construction are fully automatic.

In nature, the equivalent to fully automatic design software is the process of evolution. The body and brain of a creature are tightly coupled, the fruit of a long series of small mutual adaptations—like the chicken and the egg, neither one was designed first. There is never a situation in which the hardware has no software, or where a growth or mutation beyond the adaptive ability of the brain survives. Autonomous robots, like living creatures, require a highly sophisticated correspondence between brain, body and environment. Using natural systems as inspiration, we coevolve both the brain and the body, simultaneously and continuously, from a simple controllable mechanism to one of sufficient complexity for a particular specialized task [3]. We then create a representation scheme that allows for the hierarchical construction of more complex components from previously defined ones for the evolution of modular designs. Although we were not the first to propose brain/body coevolution [4]-[6], we are the first to have gone from computer simulation to reality.

The results of our work are predictive of a future in which humans are engaged in more complex and artistic forms of creativity, while the lower-level details of design and interactions between components are managed by computers. In this paper, we show that computer software, properly situated, can create functional forms and, further, that there are representational techniques that enable these systems to begin to scale to more complex forms in which evolved components are replicated and reused in

Fig. 1.

Photographs of the FAD LEGO bridge (cantilever) and crane (triangle).

Photos: © Pablo Funes and Jordan Pollack

Fig. 2.

(a) A tetrahedral mechanism produces hinge-like motion and advances by pushing the central bar against the floor. (b) Bipedalism: the left and right limbs are advanced in alternating thrusts. (c) Moving the two articulated components produces crab-like sideways motion. (d) While the upper two limbs push, the central body is retracted and vice versa. (e) This simple mechanism uses the top bar to delicately shift balance from side to side, shifting the friction point to either side as it creates oscillatory motion and advances. (f) This mechanism has an elevated body, from which it pushes an actuator down directly onto the floor to create a ratcheting motion. Redundant...