In Genetic Programming, introns play at least two substantial roles: (1) A structural protection role, allowing the population to preserve highly-fit building blocks; and (2) A global protection role, enabling an individual to protect itself almost entirely against the destructive effect of crossover. We introduce Explicitly Defined Introns into Genetic Programming. Our results suggest that the introduction of Explicitly Defined Introns can improve fitness, generalization, and CPU time. Further, Explicitly Defined Introns partially replace the role of Implicit Introns ( that is, introns that emerge from crossover and mutation without being explicitly defined as such). Finally, Explicitly Defined Introns and Implicit Introns appear, in some situations, to work in tandem to produce better training, fitness and generalization than occurs without Explicitly Defined Introns. 1 Introduction Introns are an important part of the genomes of eucaryotic cells. In some genes, up to 70 %...

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.

The importance of digital data compression in the future media arena cannot be overestimated. A novel approach to data compression is built on Genetic Programming. This technique has been referred to as "programmatic compression". In this paper we apply a variant of programmatic compression to the compression of bitmap images and sampled digital sound. The work presented here constitutes the first successful result of genetic programming applied to compression of real full size images and sound. A compiling genetic programming system is used for efficiency reasons. Different related issues are discussed, such as the handling of very large fitness case sets. 1 Introduction Programmatic compression is a very general form of compression. The basic idea behind this technique is that any system, which evolves programs or algorithms for generating data, can be viewed as a data compression system. The data that should be compressed are presented to the Genetic Programming system as fitness ...

We have evaluated the use of Genetic Programming to directly control a miniature robot. The goal of the GP-system was to evolve real-time obstacle avoiding behaviour from sensorial data. The evolved programs are used in a sense-think-act context. We employed a novel technique to enable real time learning with a real robot. The technique uses a probabilistic sampling of the environment where each individual is tested on a new real-time fitness case in a tournament selection procedure. The fitness has a pain and a pleasure part. The negative part of fitness, the pain, is simply the sum of the proximity sensor values. In order to keep the robot from standing still or gyrating, it has a pleasure componentton fitness. It gets pleasure from going straight and fast. The evolved algorithm shows robust performance even if the robot is lifted and placed in a completely different environment or if obstacles are moved around.

. Ordinarily, Genetic Programming uses little or no mutation. Crossover is the predominant operator. This study tests the effect of a very aggressive use of the mutation operator on the generalization performance of our Compiling Genetic Programming System ('CPGS'). We ran our tests on two benchmark classification problems on very sparse training sets. In all, we performed 240 complete runs of population 3000 for each of the problems, varying mutation rate between 5% and 80%. We found that increasing the mutation rate can significantly improve the generalization capabilities of GP. The mechanism by which mutation affects the generalization capability of GP is not entirely clear. What is clear is that changing the balance between mutation and crossover effects the course of GP training substantially --- for example, increasing mutation greatly extends the number of generations for which the GP system can train before the population converges. 1 Introduction Evolutionary A...

Stack-based genetic programming is an alternative to Koza-style tree-based genetic programming that generates linear programs that are executed on a virtual machine using a FORTH-style operand stack instead of tree-based function calls. A stack-based genetic programming system was extended to include the ability to generate programs containing automatically defined functions. Experiments were run to test the system using Koza's lawnmower problem. The stack-based system using automatically defined functions was able to successfully solve the lawnmower problem. Solutions sizes using automatically defined functions were comparable to those reported by Koza for the tree-based system. Solution sizes without using automatically defined functions were much larger in the stack-based system. The stack-based system both with and without automatically defined functions required significantly more search than was performed by the tree-based system. The efficiency and average structural co...

A computer language is a very general form of representing and specifying an autonomous agent's behavior. The task of planning feasible actions could then simply be reduced to an instance of automatic programming. We have evaluated the use of an evolutionary technique for automatic programming called Genetic Programming (GP) to directly control a miniature robot. To our knowledge, this is the first attempt to control a real robot with a GP based learning method. Two schemes are presented. The objective of the GP system in our first approach is to evolve real-time obstacle avoiding behavior. This technique enables real-time learning with a real robot using genetic programming. It has, however, the drawback that the learning time is limited by the response dynamics of the environment. To overcome this problems we have devised a second method, learning from past experiences which are stored in memory. This new system allows a speed-up of the algorithm by a factor of more than 2000. Obstac...

We have used an automatic programming method called Genetic Programming (GP) for control of a miniature robot. Our earlier work on real-time learning suffered from the drawback of the learning time being limited by the response dynamics of the robot's environment. In order to overcome this problem we have devised a new technique which allows learning from past experiences that are stored in memory. The new method shows its advantage when perfect behavior emerges in experiments quickly and reliably. It is tested on two control tasks, obstacle avoiding and wall following behavior, both in simulation and on the real robot platform Khepera.

One of the most general forms of representing and specifying behavior is by using a computer language. We have evaluated the use of the evolutionary technique of Genetic Programming (GP) to directly control a miniature robot. The goal of the GP-system was to evolve real-time obstacle avoiding behavior from sensorial. The evolved programs are used in a sense-think-act context. We employed a novel technique to enable real time learning with a real robot using genetic programming. To our knowledge, this is the first use of GP with a real robot. The method uses a probabilistic sampling of the environment where each individual is tested on a new real-time fitness case in a tournament selection procedure. The robots behavior is evolved without any knowledge of the task except for the feed-back from a fitness function. The fitness has a pain and a pleasure part. The negative part of fitness, the pain, is simply the sum of the proximity sensor values. In order to keep the robot fr...

In recent years different variants of genetic programming (GP) have emerged all following the basic idea of GP, the automatic evolution of computer programs. Today, three basic forms of representation for genetic programs are used, namely tree, graph and linear structures. We introduce a multi-representation system, SYSGP, that allows researchers to experiment with different representations with only a minimum implementation overhead. The system further offers the possibility to combine modules of different representation forms into one genetic program. SYSGP has been implemented as a C++ library using templates that operate with a generic data type.