Morpho-Functional Machines

We define morpho-functional machines as machines which are adaptive not only by means of their neural substrate, but by being able to change their morphology as they perform a task in the real world. Changes in agents can occur at several time scales: short-term, ontogenetic, and phylogenetic. All of these need to be taken into account in designing morpho-functional machines. Studying natural systems, building robots with many degrees of freedom, and artificial evolution (in particular artificial ontogeny) are methods to gain insights into the field. Because the field is only in its infancy, there are no generally accepted methodologies and a multiplicity of methods must be employed to make progress.
Morpho-functional machines are adaptive intelligent devices designed to achieve their tasks not only by some adaptive control scheme, but by changing their morphology. Depending on the demands of the task and the current situation,
morpho-functional machines change their shapes accordingly. The term "morphofunctional machines" was coined by Hara and his colleagues (e.g. Kawai and Hara,
1998; Hara and Pfeifer, 2000) and is explored in detail in Hara and Kikuchi (this volume). One of the central concepts underlying morpho-functional machines is embodiment, the hallmark of New Artificial Intelligence (Brooks, 2000; Pfeifer and Scheier, 1999).
Early approaches to understanding intelligence have abstracted from physical properties - i.e. from the embodiment - of individual organisms. The generally accepted assumption was that behavior can be studied at the level of algorithms,
which is why for many years the major tool of AI researchers has been the computer (this has become known as classical AI). As researchers started to build robots they realized that the hardest issues in the study of intelligence involve perception and action in the real world.

Rodney Brooks of the MIT AI Lab, who was among the first to recognize the importance of system-environment interaction for intelligence, started a new research field called "behavior-based robotics" (or "embodied AI", or "New AI") (e.g. Brooks, 1991). The concept of embodiment became an important focus of research in AI, psychology, and what is normally subsumed under the label "cognitive science". By embodiment we mean that agents are realized as physical systems and can thus exhibit behavior that can be observed in the real world. The physical characteristics of sensory and motor systems as well as their morphology,
i.e. their form and structure, are important aspects of embodiment. In biology morphology has been a central research topic for a long time (e.g. 0'Arcy Thompson, 1942). More recently, with the emergence of the field of molecular embryology, the interest in how form comes about in natural agents, has started to explode. Gerald Edelman (1988) in his seminal book "Topobiology" recognized the importance of studying ontogenetic development and morphogenesis. In the cognitive sciences, especially in AI, cognitive psychology, and in neurobiology,
morphogensis has been largely neglected which implies that an essential explanatory component is missing. For example, if we want to understand the function of the neural substrate, the brain, it is not sufficient to look at the neural substrate itself. It has to be known how it is embedded in the physical agent and what the properties, i.e. the morphology and the physical characteristics of the sensory and the motor systems are to which the neural network is connected. In addition, the task environment - the ecological niche and the tasks - have to be taken into account. Through embodiment it is determined what signals the neural system has to process in the first place. During the course of evolution, neural systems have started to exploit embodiment in ingenious ways.

Morpho-functional Machines:  The New Species  Designing Embodied Intelligence by F. Hara, R. Pfeifer (Eds.)