The DVL project, “Developmental Learning Algorithms for Embedded Agents”, (2002-06) EPSRC, explored several foundational ideas in developmental robot learning. Based on a hand/eye system made from an Adept robot arm and motorized pan/tilt camera.
The REVERB project (2005-2010) EPSRC, “Reverse Engineering the VERtebrate Brain”, was alarge consortium of UK neuroscientists, computer scientists and engineers modelled some of the action selection mechanisms found in the Basal Ganglia. Involved a Schunk robot arm and tactile hand, and a RoboSoft fast saccade binocular camera system.REVERB Website.
The ROSSI project (2008-2011, EC FP7) examined the nature of the internal mental models necessary for robots and humans to interact and cooperate. We explored both robot grasping with the above three fingered tactile robot, and human grasping and manipulation with data capture equipment. Partners: Università di Bologna, Italy; Università degli Studi di Parma, Italy; Universität zu Lübeck, Germany; Högskolan I Skövde, Sweden; and the Middle East Technical University in Istanbul. ROSSI Video.
The IM-CLeVeR project (2010-2013, EC FP7) implemented novel software on the iCub humanoid robot in order to investigate developmental learning processes. Development in the human infant is restricted by a series of constraints, which restrain the infant's action repertoire and sensing capabilities. We performed longitudinal experiments in which the constraints are lifted with maturity and experience. Partners: Consiglio Nazionale delle Ricerche, Rome, Italy, Universita Campus Bio Medico di Roma, Italy; Frankfurt Institute for Advanced Studies, Germany; Scuola Universitaria Professionale della Svizzera Italiana, Switzerland; University of Sheffield, UK; University of Ulster, UK. IM-CLeVeR Website
The MoDeL Project (2015-18) EPSRC, “Developmental algorithms for robotics: Understanding the world of objects, interactions and tools” has the goal of designing, implementing, and understanding a mechanism that will drive robots to autonomously develop new behaviors and learn about novel events for which they have no prior experience. There are two components being investigated: (1) intrinsic motivation for autonomous operation is provided by a play generator that produces behavior analogous to infant play with objects; and (2) the encounters with these new environmental experiences is formative in the perception of objects and other sensorimotor experience.
Our hypothesis is that generating infant-like play in robots is an effective and powerful way of building real world understanding, and also driving the exploration and practice of skills, and the discovery of novel behavior and new action possibilities.
This research project is a psychologically-inspired investigation, with our computational models being implemented as novel software on an iCub humanoid robot. The experiments will include solitary play with objects, interactive play with a human participant, and simple tool use.
Scientific Advisory Panel: Dr Jacqueline Fagard, Paris Descartes University; Prof Merideth Gattis, Cardiff University; Dr Giorgio Metta, IIT, Genova, Italy; Dr Keith Mowbray, QinetiQ, UK ; Prof. Kevin O’Regan, Paris Descartes University; Dr B. Stjerne Thomsen, The Lego Foundation, Denmark; Dr David Whitebread, Cambridge University, UK.