Michal Joachimczak

Evolution of the morphology and patterning of artificial embryos: scaling the tricolour problem to the third dimension

We present a model of three-dimensional artificial embryogenesis in which a multicellular embryo develops controlled by a continuous regulatory network encoded in a linear genome. Development takes place in a continuous space, with spherical cells of variable size, and is controlled by simulated physics. We apply a genetic algorithm to the problem of the simultaneous evolution of morphology and patterning into colour stripes and demonstrate how the system achieves the task by exploiting physical forces and using self-generated morphogen gradients. We observe a high degree of robustness to damage in evolved individuals and explore the limits of the system using more complex variations of the problem. We find that the system remains highly evolvable despite the increased complexity of three-dimensional space and the flexible coding of the genome requiring from evolution to invent all necessary morphogens and transcription factors.

Co-evolution of morphology and control of soft-bodied multicellular animats

We present a platform that allows for co-evolution of development and motion control of soft-bodied, multicellular animats in a 2-dimensional fluid-like environment. Artificial gene regulatory networks (GRNs) with real-valued expression levels control cell division and differentiation in multicellular embryos. Embryos develop in a simulated physics environment and are converted into animat structures by connecting neighboring cells with elastic springs. The springs connecting outer cells form the external envelope which is subject to fluid drag. Both the developmental program and motion control are encoded indirectly in a single linear genome, which consists of regulatory regions and regions that code for regulatory products (some of which act as morphogens). We applied a genetic algorithm to co-evolve morphology and control using a fitness measure whose value depends on distance traveled during the evaluation phase. We obtained various emergent morphologies and types of locomotion, some of them showing the use of appendages.

Evolution of the driving styles of anticipatory agent remotely operating a scaled model of racing car

We present an approach for automated evolutionary design of driving agent, able to remotely operate a scale model of racing car running in a fastest possible way. The agents actions are conveyed to the car via standard radio control transmitter. The agent perceives the environment from a live video feedback of an overhead camera. In order to cope with the inherent video feed latency, which renders even the straightforward tasks of following simple routes unsolvable, we implement an anticipatory modeling - the agent considers its current actions based on anticipated intrinsic (rather than currently available, outdated) state of the car and its surrounding. The driving style (i.e. the driving line combined with the speed at which the car travels along this line) is first evolved offline on a software simulator of the car and then adapted online to the real world. Experimental results demonstrate that on long runs the agent-operated car is only marginally slower than a human-operated one, while the consistence of lap times posted by the evolved driving style of the agent is better than that of a human. This work can be viewed as a step towards the development of a framework for automated design of the controllers of remotely operated vehicles capable to find an optimal solution to various tasks in different traffic situations and road conditions.

Brainless Bodies: Controlling the Development and Behavior of Multicellular Animats by Gene Regulation and Diffusive Signals

We present a model of parallel co-evolution of development and motion control in soft-bodied, multicellular animats without neural networks. Development is guided by an artificial gene regulatory network (GRN), with real-valued expression levels, contained in every cell. Embryos develop within a simulated physics environment and are converted into animat structures by connecting neighboring cells through elastic springs. Outer cells, which form the external envelope, are affected by drag forces in a fluid-like environment. Both the developmental program and locomotion controller are encoded into a single genomic sequence, which consists of regulatory regions and genes expressed into transcription factors and morphogens. We apply a genetic algorithm to evolve individuals able to swim in the simulated fluid, where the fitness depends on distance traveled during the evaluation phase. We obtain various emergent morphologies and types of locomotion, some of them showing the use of rudimentary appendages. An analysis of the selected evolved controllers is provided.

Evolution of robustness to damage in artificial 3-dimensional development.

GReaNs is an Artificial Life platform we have built to investigate the general principles that guide evolution of multicellular development and evolution of artificial gene regulatory networks. The embryos develop in GReaNs in a continuous 3-dimensional (3D) space with simple physics. The developmental trajectories are indirectly encoded in linear genomes. The genomes are not limited in size and determine the topology of gene regulatory networks that are not limited in the number of nodes. The expression of the genes is continuous and can be modified by adding environmental noise. In this paper we evolved development of structures with a specific shape (an ellipsoid) and asymmetrical pattering (a 3D pattern inspired by the French flag problem), and investigated emergence of the robustness to damage in development and the emergence of the robustness to noise. Our results indicate that both types of robustness are related, and that including noise during evolution promotes higher robustness to damage. Interestingly, we have observed that some evolved gene regulatory networks rely on noise for proper behaviour.

Evolution of driving agent, remotely operating a scale model of a car with obstacle avoidance capabilities

We present an approach for evolutionary design of an agent, remotely operating a scale model of a car running in a fastest possible way. The agent perceives the environment from a video camera and conveys its actions to the car via standard radio control transmitter. In order to cope with the video feed latency we propose an anticipatory modeling in which the agent considers its current actions based on the anticipated intrinsic (rather than currently available, outdated) state of the car and its surrounding. The agent is first evolved on software models of the car and tracks, and then adapted to the real world. During the adaptation, the lap times improve steadily to the values close to the values obtained from the evolution on the models. An evolutionary optimization of the avoidance of a small obstacle results in lap times that are virtually the same as the best lap times achieved on the same track without obstacles. Presented work can be viewed as a step towards developing a racing game in which the human competes against a computer, both operating scale models of racing cars.

ATR artificial brain project: 2004 progress report

This article presents the key assumptions and current status of the ATR Artificial Brain Project being undertaken to create Volitron, a device equipped with circuitry that enables the emergence of thought. Such thought would be recognized from Volitrons specific communication behaviors. The project consists of three complementary themes: psychodynamic architecture, brain-specific evolvable hardware, and the management of brain-building. The psychodynamic architecture is designed to develop automatically, driven by “pleasure” coming from discharges of tension gathered in special tension-accumulating devices. Tension-discharging patterns come first of all from the robots interactions with its care giver/provider. For the dedicated hardware, we developed qcellular-automata (qCA), in which groups of uniform logic primitives (q-cells) serve as spike-train-processing units, as well as pulsed para-neural networks (PPNN) that can be evolved, using fuzzified signals and a genetic algorithm combined with hill climbing, and converted into qCA. The psychodynamic ideas were tested using three robots: Neko, equipped with a pleasure-driven associator, Miao, equipped with MemeStorms (a special working memory in which conflicting ideas fight for access to the long-term memory and actuators), and Miao+, whose brain is equipped with a growing neural network.

Using the Genetic Regulatory Evolving Artificial Networks (GReaNs) Platform for Signal Processing, Animat Control, and Artificial Multicellular Development

Building a system that allows for pattern formation and morphogenesis is a first step towards a biologically-inspired developmental-evolutionary approach to generate complex neural networks. In this chapter we present one such system, for Genetic Regulatory evolving artificial Networks (GReaNs). We review the results of previous experiments in which we investigated the evolvability of the encoding used in GReaNs in problems which involved: (i) controlling development of multicellular 2-dimensional (2D) soft-bodied animats; (ii) controlling development of 3-dimensional (3D) multicellular artificial bodies with asymmetrical shapes and patterning; (iii) directed movement of unicellular animats in 2D; and (iv) processing signals at the level of single cells. We also report a recent introduction of spiking neuron models in GReaNs. We then present a road map towards using this system for evolution and development of neural networks.

Controlling development and chemotaxis of soft-bodied multicellular animats with the same gene regulatory network

The ability to actively forage for resources is one of the defining properties of animals, and can be seen as a form of minimal cognition. In this paper we model soft-bodied robots, or “animats”, which are able to swim in a simulated twodimensional fluid environment toward food particles emitting a diffusive chemical signal. Both the multicellular development and behaviour of the animats are controlled by a gene regulatory network (GRN), which is encoded in a linear genome. Coupled with the simulated physics, the activity of the GRN affects cell divisions and cell movements during development, as well as the expansion and contraction of filaments connecting the cells in the swimming adult body. The global motion that emerges from the dynamics of the animat relies on the spring-like filaments and drag forces created by the fluid. Our study shows that it is possible to evolve the animat’s genome (through mutations, duplications and deletions) to achieve directional motion in this environment. It also suggests that a “minimally cognitive” behaviour of this kind can emerge without a central control or nervous system.

Evolving Networks Processing Signals with a Mixed Paradigm, Inspired by Gene Regulatory Networks and Spiking Neurons

In this paper we extend our artificial life platform, called GReaNs (for Genetic Regulatory evolving artificial Networks) to allow evolution of spiking neural networks performing simple computational tasks. GReaNs has been previously used to model evolution of gene regulatory networks for processing signals, and also for controlling the behaviour of unicellular animats and the development of multicellular structures in two and three dimensions. The connectivity of the regulatory network in GReaNs is encoded in a linear genome. No explicit restrictions are set for the size of the genome or the size of the network. In our previous work, the way the nodes in the regulatory network worked was inspired by biological transcriptional units. In the extension presented here we modify the equations governing the behaviour of the units so that they describe spiking neurons: either leaky integrate and fire neurons with a fixed threshold or adaptive-exponential integrate and fire neurons. As a proof-of-principle, we report the evolution of spiking networks that match desired spiking patterns.