Neurorobotics:From Vision to Action 62.4 The Role of Mirror Systems 1467 or direct-current (DC)motors with spindles and added Gomi [62.911).By this regularization of the complexity elastic components are used does not affect the control of the skeletomuscular system,the complexity of the for- approach at the cerebellar level,but rather at the motor ward model stored in the cerebellum is correspondingly control level (cf.the spinal cord level).Of key impor- reduced.The whole picture therefore seems to be that tance,however,are the resulting dynamical properties the cerebellum,controlling a piecewise-linear skeleto- of the system,which are of course influenced by its muscular system,incorporates a forward model thereof actuators. to cope with delays in the peripheral nervous system. Passive flexibility at the joints,which is a key feature Consequently,although the applicability of cerebellar of muscle systems,is essential for reasons of safety,sta-systems to highly nonlinear dynamics control of tra- bility during contact with the environment,and storage ditional robots is questionable,the use of cerebellar of kinetic energy.As mentioned before,however,bio-systems as forward models appears to be useful in the logical systems are immensely complex,requiring large control of more complex and flexible robotic systems. groups of muscles for comparatively simple movements. The control challenge posed by the currently emerg- A reason for this complexity is the resulting nearly lin-ing generation of robots employing antagonistic motor ear behavior,which has been noted for,e.g.,muscle control therefore opens a new wealth of applications of activation with respect to joint stiffness (see Osu and cerebellar systems. 62.4 The Role of Mirror Systems Mirror neurons were first discovered in the brain of latti [62.100]and Arbib [62.1011).Indeed,it has been the macaque monkey -neurons that fire both when suggested that mirror neurons underlie the motor the- the monkey exhibits a particular grasping action,and ory of speech perception of Liberman et al.[62.102]. when the monkey observes another(monkey or human) which holds that speech perception rests on the ability perform a similar grasp (see Rizzolatti et al.[62.92] to recognize the motor acts that produce speech sounds. and Gallese et al.[62.93]).Since then,human stud- Section 62.4.1 reviews basic neurophysiological ies have revealed a mirror system for grasping in the data on mirror neurons in the macaque,and presents human brain-a mirror system for a class X of ac-both the Fagg-Arbib-Rizzolatti-Sakata(FARS)model tions being a set of regions that are active both when of canonical neurons (unlike mirror neurons,these are the human performs some action from X and when active when the monkey executes an action but not when the human observes someone else performing an ac- he observes it)and the mirror neuron system (MNS) tion from X (see,e.g.,Grafion et al.[62.94],Rizzolatti model of mirror neurons and their supporting brain re- et al.[62.95],and Fadiga et al.[62.961).Until recently gions.Section 62.4.2 then uses Bayes's rule to offer we had no single-neuron studies of humans proving the a new,probabilistic view of the mirror system's role in reasonable hypothesis that the human mirror system for action recognition,and demonstrates the operation of the grasping contains mirror neurons for specific actions. new model in the context of studies with two robots.Fi- However,data from neurosurgery are now becoming nally,Sect.62.4.3 briefly shifts the emphasis ofour study available (M.Iacoboni,personal communication).In of mirror neurons to imitation,which in fact is the area any case,most models of the human mirror system that has most captured the imagination of roboticists. Part for grasping assume that it contains circuitry analogous to the mirror neuron circuitry of the macaque brain. 62.4.1 Mirror Neurons and the Recognition However,the current consensus is that monkeys have of Hand Movements little or no ability for imitation (but see Voelkl and Huber [62.97]for a very simple form of imitation-like Area F5 in the premotor cortex of the macaque contains, behavior in marmoset monkeys):great apes have the among others,neurons which fire when the monkey ex- ability to master certain skills after extended bouts of ecutes a specific manual action,e.g.,one neuron might observation (Byrne [62.98]),whereas the human mir- fire when the monkey performs a precision pinch,an- ror system plays a key role in our capability for much other when it executes a power grasp.(In discussing richer forms of imitation (see lacoboni et al.[62.99]), neurorobotics,it seems unnecessary to explain in any pantomime,and even language (see Arbib and Rizzo- detail the areas like F5,AIP,and STS described hereNeurorobotics: From Vision to Action 62.4 The Role of Mirror Systems 1467 or direct-current (DC) motors with spindles and added elastic components are used does not affect the control approach at the cerebellar level, but rather at the motor control level (cf. the spinal cord level). Of key impor￾tance, however, are the resulting dynamical properties of the system, which are of course influenced by its actuators. Passive flexibility at the joints, which is a key feature of muscle systems, is essential for reasons of safety, sta￾bility during contact with the environment, and storage of kinetic energy. As mentioned before, however, bio￾logical systems are immensely complex, requiring large groups of muscles for comparatively simple movements. A reason for this complexity is the resulting nearly lin￾ear behavior, which has been noted for, e.g., muscle activation with respect to joint stiffness (see Osu and Gomi [62.91]). By this regularization of the complexity of the skeletomuscular system, the complexity of the for￾ward model stored in the cerebellum is correspondingly reduced. The whole picture therefore seems to be that the cerebellum, controlling a piecewise-linear skeleto￾muscular system, incorporates a forward model thereof to cope with delays in the peripheral nervous system. Consequently, although the applicability of cerebellar systems to highly nonlinear dynamics control of tra￾ditional robots is questionable, the use of cerebellar systems as forward models appears to be useful in the control of more complex and flexible robotic systems. The control challenge posed by the currently emerg￾ing generation of robots employing antagonistic motor control therefore opens a new wealth of applications of cerebellar systems. 62.4 The Role of Mirror Systems Mirror neurons were first discovered in the brain of the macaque monkey – neurons that fire both when the monkey exhibits a particular grasping action, and when the monkey observes another (monkey or human) perform a similar grasp (see Rizzolatti et al. [62.92] and Gallese et al. [62.93]). Since then, human stud￾ies have revealed a mirror system for grasping in the human brain – a mirror system for a class X of ac￾tions being a set of regions that are active both when the human performs some action from X and when the human observes someone else performing an ac￾tion from X (see, e.g., Grafton et al. [62.94], Rizzolatti et al. [62.95], and Fadiga et al. [62.96]). Until recently we had no single-neuron studies of humans proving the reasonable hypothesis that the human mirror system for grasping contains mirror neurons for specific actions. However, data from neurosurgery are now becoming available (M. Iacoboni, personal communication). In any case, most models of the human mirror system for grasping assume that it contains circuitry analogous to the mirror neuron circuitry of the macaque brain. However, the current consensus is that monkeys have little or no ability for imitation (but see Voelkl and Huber [62.97] for a very simple form of imitation-like behavior in marmoset monkeys); great apes have the ability to master certain skills after extended bouts of observation (Byrne [62.98]), whereas the human mir￾ror system plays a key role in our capability for much richer forms of imitation (see Iacoboni et al. [62.99]), pantomime, and even language (see Arbib and Rizzo￾latti [62.100] and Arbib [62.101]). Indeed, it has been suggested that mirror neurons underlie the motor the￾ory of speech perception of Liberman et al. [62.102], which holds that speech perception rests on the ability to recognize the motor acts that produce speech sounds. Section 62.4.1 reviews basic neurophysiological data on mirror neurons in the macaque, and presents both the Fagg–Arbib–Rizzolatti–Sakata (FARS) model of canonical neurons (unlike mirror neurons, these are active when the monkey executes an action but not when he observes it) and the mirror neuron system (MNS) model of mirror neurons and their supporting brain re￾gions. Section 62.4.2 then uses Bayes’s rule to offer a new, probabilistic view of the mirror system’s role in action recognition, and demonstrates the operation of the new model in the context of studies with two robots. Fi￾nally, Sect. 62.4.3 briefly shifts the emphasis of our study of mirror neurons to imitation, which in fact is the area that has most captured the imagination of roboticists. 62.4.1 Mirror Neurons and the Recognition of Hand Movements Area F5 in the premotor cortex of the macaque contains, among others, neurons which fire when the monkey ex￾ecutes a specific manual action, e.g., one neuron might fire when the monkey performs a precision pinch, an￾other when it executes a power grasp. (In discussing neurorobotics, it seems unnecessary to explain in any detail the areas like F5, AIP, and STS described here Part G 62.4