Evolution of Robotic Behaviour Using Gene Expression Programming

Abstract

The main objective in automatic robot controller development is to devise mechanisms\nwhereby robot controllers can be developed with less reliance on human developers. One\nsuch mechanism is the use of evolutionary algorithms (EAs) to automatically develop\nrobot controllers and occasionally, robot morphology. This area of research is referred\nto as evolutionary robotics (ER). Through the use of evolutionary techniques such as\ngenetic algorithms (GAs) and genetic programming (GP), ER has shown to be a promising\napproach through which robust robot controllers can be developed. \nThe standard ER techniques use monolithic evolution to evolve robot behaviour: monolithic\nevolution involves the use of one chromosome to code for an entire target behaviour.\nIn complex problems, monolithic evolution has been shown to suffer from bootstrap problems;\nthat is, a lack of improvement in fitness due to randomness in the solution set\n[103, 105, 100, 90]. Thus, approaches to dividing the tasks, such that the main behaviours\nemerge from the interaction of these simple tasks with the robot environment\nhave been devised. These techniques include the subsumption architecture in behaviour\nbased robotics, incremental learning and more recently the layered learning approach\n[55, 103, 56, 105, 136, 95]. These new techniques enable ER to develop complex controllers\nfor autonomous robot. Work presented in this thesis extends the field of evolutionary robotics by introducing Gene\nExpression Programming (GEP) to the ER field. GEP is a newly developed evolutionary\nalgorithm akin to GA and GP, which has shown great promise in optimisation problems.\nThe presented research shows through experimentation that the unique formulation of\nGEP genes is sufficient for robot controller representation and development. The obtained\nresults show that GEP is a plausible technique for ER problems. Additionally, it is shown\nthat controllers evolved using GEP algorithm are able to adapt when introduced to new\nenvironments.\nFurther, the capabilities of GEP chromosomes to code for more than one gene have been\nutilised to show that GEP can be used to evolve manually sub-divided robot behaviours.\nAdditionally, this thesis extends the GEP algorithm by proposing two new evolutionary\ntechniques named multigenic GEP with Linker Evolution (mgGEP-LE) and multigenic\nGEP with a Regulator Gene (mgGEP-RG). The results obtained from the proposed algorithms\nshow that the new techniques can be used to automatically evolve modularity\nin robot behaviour. This ability to automate the process of behaviour sub-division and\noptimisation in a modular chromosome is unique to the GEP formulations discussed, and\nis an important advance in the development of machines that are able to evolve stratified\nbehavioural architectures with little human intervention.