Experiments in learning by imitation - Grounding and use of communication in robotic agents

Citation
A. Billard et K. Dautenhahn, Experiments in learning by imitation - Grounding and use of communication in robotic agents, ADAPT BEHAV, 7(3-4), 1999, pp. 415-438
Citations number
57
Categorie Soggetti
Psycology
Journal title
ADAPTIVE BEHAVIOR
ISSN journal
10597123 → ACNP
Volume
7
Issue
3-4
Year of publication
1999
Pages
415 - 438
Database
ISI
SICI code
1059-7123(199924)7:3-4<415:EILBI->2.0.ZU;2-N
Abstract
Social behaviour and in particular social learning are key mechanisms for t he cohesion and evolution of primate societies. Similarly, social skills mi ght be desirable for artificial agents who are expected to interact with ot her natural or artificial agents. We view learning, communication and imita tion as important capabilities to possess by social artificial agents and s tudy how these skills can be designed and used by physically embodied auton omous robots. We study grounding and use of communication among heterogeneo us agents. In particular, we investigate the role of social interactions fo r sharing of context and building of joint attention among communicative ag ents. Grounding and use of communication is investigated through simulation s within a group of autonomous agents. Results show that social behaviour b enefit the agents in two circumstances: (I) agents capable of following one another, and in this way imitating each other's movements, develop faster and better a common understanding of the language; (2) furthermore, the age nts' capability of communicating with one another via a common vocabulary b enefits to the group and to each agent individually as it speeds up the tra nsmission of information. We use a connectionist model, based on Hebbian as sociative learning, for the learning of the word-signal pairs. This work fo llows robotic experiments [6, 5, 7] in which a physical autonomous robot wa s taught a vocabulary to describe its perceptions of objects, movement, inc lination and orientation. The robot was taught: either by a human instructo r or by another robot. The teacher-learner robot experiments were based on an imitative strategy whereby the learner robot followed the teacher robot. The work of this payer demonstrates scaling up of this movement imitative strategy for transmitting a vocabulary across a group of robotic agents, i. e. from a teacher agent to several learner agents. In particular, it shows that imitative behaviour is necessary for the grounding of the agents' prop rioceptions and speeds up the grounding of exteroceptions. These studies st ress the importance of behavioural social mechanisms in addition to general cognitive abilities of associativity for grounding communication in embodi ed agents. In particular, it shows that a simple movement imitation strateg y is an interesting scenario for the transmission of a language, as it is a n easy means of getting the agents to share a common context of perceptions , which is a prerequisite for a common understanding of the language to dev elop. It is thus suggested that a behaviour-oriented approach might be more appropriate than a pure cognitivist one which is dominating in related stu dies of the mechanisms involved in grounding communication.