We present a novel evolutionary approach to robotic control of a real robot based on genetic programming (GP). Our approach uses genetic programming techniques that manipulate machine code to evolve control programs for robots. This variant of GP has several advantages over a conventional GP system, such as higher speed, lower memory requirements and better real time properties. Previous attempts to apply GP in robotics use simulations to evaluate control programs and have difficulties with learning tasks involving a real robot. We present an on-line control method that is evaluated in two different physical environments and applied to two tasks using the Khepera robot platform: obstacle avoidance and object following. The results show fast learning and good generalization.

Most automated reasoning systems relies on human knowledge or heuristics to guide the reasoning or search for proofs. We have evaluated the use of a powerful general search algorithm to search in the space of mathematical proofs. In our approach automated reasoning is seen as an instance of automated programming where the proof is seen as a program (of functions corresponding to rules of inference) that transforms a statement into an axiom. Genetic programming is a technique for automated programming that evolves programs with a genetic algorithm. We show that such a system can be used to evolve mathematical proofs in complex domains i.e. arithmetics and program verification. The system is not restricted to evaluations of classical two-valued logic but can be used with for instance Kleene's three valued logic in order to detect paradoxes that can occur in real life reasoning applications. email:nordin@ls11.informatik.uni-dortmund.de y email: banzhaf@ls11.informatik.uni-dortmund.de ...

Genetic Programming (GP) is a machine learning technique that writes computer programs, automatically. Although individual researchers used GP techniques in the 1960’s and 1970’s, GP emerged as a distinct discipline in 1992. 63 Since that time, over one thousand academic studies have been published in the field and, in 1998, commercial GP software – Discipulus – reached the market. 20 Discipulus is an extremely fast,