Perception driven robotic assembly based on ecological approach

The key difficulty in robotic assembly is not the control problem but the problem of perception. In this paper, from the viewpoint of the affordance theory, an action-perception cycle model is employed for realizing a perception driven assembly robot. The proposed robot has neither a set of rational action rules nor a map of the environment. However, the robot achieves assembling through the interaction with its surroundings. Instead of action rules, the robot uses a kind of oscillator, which is named action pattern generator (APG), to take active actions. In order to improve the performance of the assembly robot, two types of memory, namely, working memory and episodic memory, are also introduced into the brain of the robot. Simulations show that the assembly robot mates two parts rapidly even though it does not know their initial arrangements.