Solving the Correspondence Problem Between Dissimilarly Embodied Robotic Arms Using the ALICE Imitation Mechanism

Abstract

Imitation is a powerful mechanism whereby knowledge may be transferred between agents (both biological and artificial). A crucial problem in imitation is the correspondence problem, mapping action sequences of the model and the imitator agent. This problem becomes particularly obvious when the two agents do not share the same embodiment and affordances. This paper describes work with our general imitation mechanism called ALICE (Action Learning for Imitation via Correspondence between Embodiments) that specifically addresses the correspondence problem. The mechanism has been implemented in two different software test-beds. The previous implementation, chessworld, is briefly summarised and the current robotic arm manipulator implementation is presented in this paper. Using the robotic arm test-bed we present proof of concept for the social transmission of behavioural patterns through groups of heterogeneous agents. We also present experiments that illustrate the impact of synchronization, loose perceptual matching and proprioception on the imitative performance. The robustness and adaptive nature of the ALICE mechanism is further illustrated with examples where the imitator agent embodiment is changing during the initial and later stages of the learning process.