A major impediment of cellular automata (CA) stems from the difficulty of utilizing their complex behavior to perform useful computations. Recent studies by [ Packard, 1988, Mitchell et al., 1994b ] have shown that CAs can be evolved to perform a computational task. In this paper non-uniform CAs are studied, where each cell may contain a different rule, in contrast to the original, uniform model. We describe experiments in which non-uniform CAs are evolved to perform the computational task using a local, co-evolutionary algorithm. For radius r = 3 we attain peak performance values of 0:92 comparable to those obtained for uniform CAs (0:93 \Gamma 0:95). This is notable considering the huge search spaces involved, much larger than the uniform case. Smaller radius CAs (previously unstudied in this context) attain performance values of 0:93 \Gamma 0:94. For r = 1 this is considerably higher than the maximal possible uniform CA performance of 0:83, suggesting that nonuniformity reduces con...

Self-reproducing, cellular automata-based systems developed to date broadly fall under two categories; the first consists of machines which are capable of performing elaborate tasks, yet are too complex to simulate, while the second consists of extremely simple machines which can be entirely implemented, yet lack any additional functionality aside from self-reproduction. In this paper we present a self-reproducing system which is completely realizable, while capable of executing any desired program, thereby exhibiting universal computation. Our starting point is a simple self-reproducing loop structure onto which we "attach" an executable program (Turing machine) along with its data. The three parts of our system (loop, program, data) are all reproduced, after which the program is run on the given data. The system reported in this paper has been simulated in its entirety; thus, we attain a viable, self-reproducing machine with programmable capabilities. 1 Introduction The study of art...

Some of the major outstanding problems in biology are related to issues of emergence and evolution. These include: (1) how do populations of organisms traverse their adaptive landscapes? (2) what is the relation between adaptedness and fitness? (3) the formation of multi-cellular organisms from basic units or cells. In this paper we study these issues using a model which is both general and simple. The system, derived from the CA (cellular automata) model, consists of a two-dimensional grid of interacting organisms which may evolve over time. We first present designed multi-cellular organisms which display several interesting behaviors including: reproduction, growth, mobility. We then turn our attention to evolution in various environments, including: an environment in which competition for space occurs, an IPD (Iterated Prisoner's Dilemma) environment, an environment of spatial niches, and an environment of temporal niches. One of the advantages of AL models is the opportunities they...

We succeeded for the first time in constructing evolutionary systems on a simple 9-state 5-neighbor cellular automata (CA) space by utilizing Langton's self-reproducing loop. CA are deterministic dynamical systems capable of representing extremely complex nonlinear phenomena, where time, space and states of sites are all discrete. Many CA models of self-reproductive behavior of theoretical organisms have so far been energetically studied, but the evolutionary process of organisms driven by variation and natural selection has never been realized on CA space yet. In this dissertation, we added three improvements into Langton's loop, i.e., to realize a kind of death by introducing a new dissolving state `8' into the set of states of the CA, to enhance the adaptability (a degree of the variety of situations in which the structures in the CA space can operate regularly) of the selfreproductive mechanism described by the state-transition rules of the CA, and to modify the initial structure o...

We have previously shown that non-uniform cellular automata (CA) can be evolved to perform computational tasks, using the cellular programming algorithm. In this paper we focus on two novel issues, namely the evolution of asynchronous CAs, and the scalability of our evolved systems. We find that asynchrony presents a more difficult case for evolution though good CAs can still be attained. We describe an empirically-derived scaling procedure by which successful CAs of any size may be obtained from a particular evolved system. Our motivation for this study stems in part by our desire to attain realistic systems that are more amenable to implementation as `evolving ware', evolware. 1 Introduction Cellular automata (CA) are dynamical systems in which space and time are discrete. A cellular automaton consists of an array of cells, each of which can be in one of a finite number of possible states, updated synchronously in discrete time steps according to a local, identical interaction rule...

this paper non-uniform cellular automata (CA) are studied, presenting the cellular programming algorithm for co-evolving such CAs to perform computations. The algorithm is applied to the evolution of random number generators; our results suggest that evolved generators are at least as good as previously described CAs, with notable advantages arising from the existence of a "tunable" algorithm for obtaining random number generators.

A new self-replicating cellular automata (CA) model is proposed as a latest effort toward the realization of an artificial evolutionary system on CA where structural complexity of self-replicators can increase in some cases. I utilize the idea of `shape encoding' proposed by Morita and Imai (Morita & Imai 1996b) and make the state-transition rules of the model allow organisms to transmit genetic information to others when colliding against each other. Simulations with random initial configuration demonstrate that it is possible that the average length of organisms and the average frequency of brancing per organism both increase, with decreasing self-replication fidelity, and saturate at some constant level. The saturation is caused in part by the fixation of place and shape of organisms onto particular sites. This implies the necessity of introducing some fluidity of site arrangements into the model for further development of evolutionary models using CA-like artificial medi...

Recent studies have shown that non-uniform cellular automata (CA), where cellular rules need not necessarily be identical, can be co-evolved to perform computational tasks. This paper extends these studies by generalizing on a second aspect of CAs, namely their standard, homogeneous connectivity. We study non-standard architectures, where each cell has a small, identical number of connections, yet not necessarily from its most immediate neighboring cells. We show that such architectures are computationally more efficient than standard architectures in solving global tasks, and also provide the reasoning for this. It is shown that one can successfully evolve non-standard architectures through a two-level evolutionary process, in which the cellular rules evolve concomitantly with the cellular connections. Specifically, studying the global density task, we identify the average cellular distance as a prime architectural parameter determining cellular automata performance. We carry out a ...

We propose to extend cellular automata for defining a new class of machines, that are a super-class of them and also of Turing machines. On the basis of a kind of fusion between Turing machines and cellular automata, a new family of parallel machines called bio-machines is described. A bio-machine is able to synthesize some biological-like phenomena including evolution and lifetime individual adaptation of simple organisms, such as unicellulars (free-living and colonial ones) and simple multicellulars. 1 Introduction Life is extremely complex and, as a consequence, biology is an extremely difficult science. One of the reasons of that complexity is that the "basis" of life, its underlying physics, is very rich. Our aim is therefore to offer an Artificial Life model that simplifies this basis to a minimum set of fundamental physical laws supporting certain biological-like processes and allowing to study a synthetic approximation of life. These laws include material and energetic const...

. Cellular automata (CA) are dynamical systems in which space and time are discrete, where each cell obeys the same rule and has a finite number of states. In this paper we study non-uniform CA, i.e. with non-uniform local interaction rules. Our focal point is the issue of universal computation, which has been proven for uniform automata using complicated designs embedded in cellular space. The computation-universal system presented here is simpler than previous ones, and is embedded in the minimal possible two-dimensional cellular space, namely 2-state, 5-neighbor (which is insufficient for universal computation in the uniform model). The space studied is quasi-uniform, meaning that a small number of rules is used (our final design consists of just two rules which is minimal), distributed such that most of the grid contains one rule except for an infinitely small region which contains the others. We maintain that such automata provide us with a simple, general model for studying Artif...