Achieving Environmental Tolerance through the Initiation and Exploitation of External Information Gunnar Tufte The Norwegian University of Science and Technology Department of Computer and Information Science Sem Selandsvei 7-9, 7491 Trondheim, Norway gunnart@idi.ntnu.no Pauline C. Haddow The Norwegian University of Science and Technology Department of Computer and Information Science Sem Selandsvei 7-9, 7491 Trondheim, Norway pauline@idi.ntnu.no Abstract— It is often believed that biological organisms have an inherent tolerance to environmental changes. This is a seductive concept if transferred to artificial organisms. An experimental approach is taken to investigate if tolerance to the environmental is an inherent property of developing organisms. The environment used is an external environment which the phenotype have to develop and survive in. As such, it is the phenotype that needs to adapt to its surrounding. The results show that tolerance is a property that can be sought by evolution rather than an inherent property of the organism, i.e. genomes that have explicitly been exposed to environmental changes. Further, to exploit this property a definition of what environmental changes mean in an artificial development setting is needed. The article classifies the roles of environment in artificial development models and suggests an external environment wherein the organism develops and function. I. I NTRODUCTION In a direct mapping i.e. one-to-one mapping, there is a single phenotype, defined by the genotype. This implies that each element in the phenotype is represented explicitly in the genotype. Introducing mechanisms, inspired by development, into evolutionary computation is motivated by a need to overcome limitations in such direct mapping approaches, e.g. scaling [1] and/or to mimic developmental properties in artificial systems e.g. replicating and regenerating systems [2]. A developmental mapping is an example of an indirect mapping. In biological development, an initial unit — a cell, holds the complete building plan (DNA) for an organism. It is important to note that this plan is generative — it describes how to build the system, not what the system will look like. Similarly in a developmental mapping, the artificial organism starts out as a single cell where the genome provides the cell’s DNA. The processing of the genome may be based on gene regulation [3]. Each development step, or stage in the mapping, produces a candidate phenotype which continues to emerge during development. Gene regulation implies that different parts of the genome are expressed in different cells at different times in the emerging phenotype. An interesting feature of developed biological organisms is that they can, not only, tolerate certain changes due to external environmental influences but that such influences may, in fact, affect their structure or functionality so as to create a robust organism. In an artificial setting, this tolerance may be regarded as an emerging tolerance to external fluctuations. In the work of Miller [4] and [5] robustness appeared to be a shadow effect [6] or emergent property of evolution and development as it was not specified as a target behaviour in the fitness function. However, is robustness really an inherent property of artificially developed organisms or something that may be sought? The focus of this paper is to investigate how external influences affect the emerging phenotype, in what way these influences are tolerated and how such a tolerance may be exploited in artificial systems including a developmental approach. The article is laid out as follows: Section II classifies the roles of environment in artificial development models. The cellular developmental model is presented in Section III. Section IV describes the evaluation used herein. The en- vironmental properties of the sytem is given in Section V. Experimental results are presented in Section VI. Finally, Section ?? concludes the work. II. BACKGROUND As stated, an important feature of natural development is that the developing organism develops within an en- vironment. However, what is the environment? The word “environment” may be said to have a different meaning in biological development than in artificial development. This, in itself, is not a problem, but within the field we need a clear definition such that there is consistency in the use of the term to aid understanding of the characteristics of the different development models under investigation. The first level at which the word environment is used is within the cell itself, referring to the intra-cell environment that the DNA resides in — see [7]. This is also referred to as the cell’s metabolism — see [8], [7] which, for the purposes of clearly defining environment seems a more appropriate term. The second level of environment, found in most models is the neighbour environment referring to the inter-cell communication. Cells may communicate their protein levels as in [7] or chemical levels and cell type ,as in [9], [8]; or cell types alone as in [10] . A further process of cell communication is to allow chemicals to diffuse as in [9], [8], [11]. Neighbours may consist of a 2D von Neumann neighbourhood, as in [10], [8]; a 3D von Neumann