Beyond Gazing, Pointing, and Reaching A Survey of Developmental Robotics Max Lungarella Neuroscience Research Institute Tsukuba AIST Central 2 Tsukuba 305-8568, Japan max.lungarella@aist.go.jp Giorgio Metta LIRA-Lab, DIST Univ. of Genova 16145 Genova, Italy pasa@dist.unige.it Abstract Developmental robotics is an emerging field lo- cated at the intersection of developmental psy- chology and robotics, that has lately attracted quite some attention. This paper gives a survey of a variety of research projects dealing with or in- spired by developmental issues, and outlines pos- sible future directions. 1. Introduction Judging from the number of recent and forthcoming con- ferences and symposia, there is an undeniable and in- creasing interest in a rapidly growing research area lo- cated at the intersection of developmental psychology and robotics that has come to be known as developmen- tal robotics. 1 Developmental robotics constitutes an in- terdisciplinary two-pronged approach to robotics, which on one side employs robots to instantiate and investi- gate models originating from developmental psychology or developmental neuroscience, and on the other hand, seeks to design better robotic systems by applying in- sights gained from studies on ontogenetic development. We believe that the growth of the affinity between devel- opmental psychology and robotics has been promoted by at least two primary driving forces: Engineers are seeking for novel methodologies ori- ented toward the advancement of robotics, and the construction of better, that is, more autonomous, and more adaptable robotic systems. In that sense, stud- ies on infant development provide a valuable source of inspiration (Asada et al., 2001; Brooks et al., 1998; Metta, 2000). Robots can be employed as research tools for the in- vestigation of embodied models of action and cogni- tion (see Sporns, 2002, for instance). Neuroscientists 1 Developmental robotics and Epigenetic robotics are very similar research endeavours. They share problems and challenges, and have a common vision. Epigenetic robotics focuses primarily on cognitive and social development (Zlatev and Balkenius, 2001). Developmen- tal robotics encompasses a broader spectrum of issues, and investigates also morphological development, and the acquisition of motor skills. and developmental psychologists, but also engineers, may gain considerable insights from trying to embed their models into robots. This approach is also known as synthetic neural modelling (Reeke et al., 1990), or synthetic methodology (Pfeifer, 2002; Pfeifer and Scheier, 1999). In many aspects developmental robotics is similar to biorobotics, which can be defined as the “intersection of biology and robotics” (Webb, 2001, p. 1033). Biorobotics addresses biological questions by building physical and biomimetic models of animals, and strives to advance en- gineering by integrating aspects of animal biomechanics and neural control into the construction of robotic sys- tems. The main goals of this article are: To survey the state of the art of developmental robotics, and to motivate the use of “robots as cognitive tools.” We maintain that on- togenetic development can be a source of inspiration, as well as a valid design alternative for the roboticist, and that robots represent a new, and powerful research tool for the cognitive scientist. In the following section, we give an overview of the various concurrent research threads. After a discussion of the implications of the developmental approach for robotics research, we point to future research directions and conclude. 2. Research Landscape This section is a survey of a variety of research projects dealing with or inspired by developmental issues. Table 2 gives a representative sampling of studies, and is not in- tended to be fully comprehensive. For the inclusion of studies we adopted the two following criteria: The study had to provide clear evidence for robotic experiments. Computer-based models of real sys- tems, such as avatars, or other sophisticated simula- tors, were discarded a priori. In other words, the sys- tem had to be situated in the real world, since only the world crystallizes the “really hard” issues (Brooks, 1991).