J. J. Collins

Grammatical Evolution: Evolving Programs for an Arbitrary Language

We describe a Genetic Algorithm that can evolve complete programs. Using a variable length linear genome to govern how a Backus Naur Form grammar definition is mapped to a program, expressions and programs of arbitrary complexity may be evolved. Other automatic programming methods are described, before our system, Grammatical Evolution, is applied to a symbolic regression problem.

Evolving Market Index Trading Rules Using Grammatical Evolution

This study examines the potential of an evolutionary automatic programming methodology to uncover a series of useful technical trading rules for the UK FTSE 100 stock index. Index values for the period 26/4/1984 to 4/12/1997 are used to train and test the model. The preliminary findings indicate that the methodology has much potential, outperforming the benchmark strategy adopted.

Occupancy grid mapping: An empirical evaluation

In this paper a quantitative analysis of robotic mapping utilising the fields dominant paradigm, the occupancy grid, is presented. The aim of this work is to determine which approach to the robotic mapping problem imbues a mobile robot with the greatest ability to create an accurate representation of its operating environment. We accomplish this by analysing the performance of several established mapping techniques using identical test data. Through evaluating the maps generated by these paradigms using an extensible benchmarking suite that our group has developed we outline which paradigm yields the greatest representational ability.

Polygenic Inheritance - A Haploid Scheme that Can Outperform Diploidy

Nonstationary function optimisation has proved a difficult area for Genetic Algorithms. Standard haploid populations find it difficult to track a moving target and tend to converge to a local optimum that appears early in a run. While it is generally accepted that various approaches involving diploidy can cope better with these kinds of problems, none of these have gained wide acceptance in the GA community. We survey a number of diploid GAs and outline some possible reasons why they have failed to gain wide acceptance, before describing a new haploid system which uses Polygenic Inheritance. Polygenic inheritance differs from most implementations of GAs in that several genes contribute to each phenotypic trait. A nonstationary function optmisation problem from the literature is described, and it is shown how various represenation scheme affect the performance of GAs on this problem.

Toward a benchmarking framework for research into bio-inspired hardware-software artefacts

In this paper, we suggest that one of the more crucial tasks currently facing researchers into the field of autonomous mobile robotics is the provision of a common task, or set of tasks, as a means of evaluating different approaches to robot design and architecture, and the generation of a common set of experimental frameworks to facilitate these different approaches. This paper stars with a brief introduction to the field, and behavior-based control in particular. We then discuss the issue of animal versus robot behavior, and focus on simulated experimentation versus embodied robotics. Finally, we move to the feasibility of evaluating and benchmarking different architectures, with the aim of producing mobile robots of continuously higher utility, with specific reference to our current four-layered robot control architecture.

Non-stationary function optimization using polygenic inheritance

Non-stationary function optimization has proved a difficult area for Genetic Algorithms. Standard haploid populations find it difficult to track a moving target, and tend to converge on a local optimum that appears early in a run. It is generally accepted that diploid GAs can cope with these problems because they have a genetic memory, that is, genes that may be required in the future are maintained in the current population. This paper describes a haploid GA that appears to have this property, through the use of Polygenic Inheritance. Polygenic inheritance differs from most implementations of GAs in that several genes contribute to each phenotypic trait. Two non-stationary function optimization problems from the literature are described, and a number of comparisons performed. We show that Polygenic inheritance enjoys all the advantages normally associated with diploid structures, with none of the usual costs, such as complex crossover mechanisms, huge mutation rates or ambiguity in the mapping process.

Evaluating techniques for resolving redundant information and specularity in occupancy grids

In this paper we consider the effect that techniques designed to deal with the problems of redundant information and erroneous sensory data have on the results of robotic mapping. We accomplish this by evaluating several configurations of these techniques using identical test data. Through evaluating the results of these experiments using an extensible benchmarking suite, that our group has developed, we outline which technique yields the greatest environmental representational gain.

Attribute grammar genetic programming algorithm for automatic code parallelization

A method is presented for evolving individuals that use an Attribute Grammar (AG) in a generative way. AGs are considerably more flexible and powerful than the closed, context free grammars normally employed by GP. Rather than evolving derivation trees as in most approaches, we employ a two step process that first generates a vector of real numbers using standard GP, before using the vector to produce a parse tree. As the parse tree is being produced, the choices in the grammar depend on the attributes being input to the current node of the parse tree. The motivation is automatic parallelization or the discovery of a re-factoring of a sequential code or equivalent parallel code that satisfies certain performance gains when implemented on a target parallel computing platform such as a multicore processor. An illustrative and a computed example demonstrate this methodology.

Artificial life and embodied robotics: current issues and future challenges

In this article we explore some of the issues currently facing researchers in the interface between the twin fields of Artificial Life and Robotics, and the challenges and potential synergy of these two areas in the creation of future robotic life forms. There are three strands of research which we feel will be of key importance in the possible development of future embodied artificial life forms. These are the areas of evolutionary robotics and evolutionary humanoid robotics in particular, probabilistic robotics for deliberation, and robot benchmarking with associated metrics and standards. We briefly explore each of these areas in turn, focusing on our current research in each field and what we see as the potential issues and challenges for the future.

Filtering reading sets for improved occupancy grid mapping

The performance of an autonomous mobile robot in acquiring a meaningful spatial model of its operating environment depends greatly on the accuracy of its perceptual capabilities. A key challenge in the field arises from handling sensor measurements and in particular the handling of measurement noise. The current prevailing approach used when attempting to deal with the problem of noisy readings is to tune the sensory model to explicitly accommodate this sensory noise. An alternative approach, however, is to consider the issue of noisy readings as a separate problem and address accordingly. In this paper such an approach is outlined. Experimental evaluation illustrate the associated improvement in map quality across a number of established paradigms.