CPM Report No.: 98-38
By: Aude Billard, Kerstin Dautenhahn and Gillian Hayes
Date: August 21st 1998
A paper presented at the workshop on Socially Situated Intelligence, held at SAB'98, the Fifth International Conference of the Society for Adaptive Behavior, University of Zürich, 17 - 21 August 1998.
Published as: Aude Billard, Kerstin Dautenhahn and Gillian Hayes (1998). Experiments on human-robot communication with Robota, an imitative learning and communication doll robot. In Edmonds, B. and Dautenhahn, K. (eds.), Socially Situated Intelligence: a workshop held at SAB'98, August 1998, Zürich. University of Zürich Technical Report, 4-16.
Imitation and communication behaviours are important means of interaction between humans and robots. In experiments on robot teaching by demonstration, imitation and communication behaviours can be used by the demonstrator to drive the robot's attention to the demonstrated task. In a children game, they play an important role to engage the interaction between the child and the robot and to stimulate the child's interest. In this work, we study how imitation skills can be used for teaching a robot a symbolic communication system to describe its actions and perceptions.
We report on experiments in which we study human-robot interactions using a doll robot. Robota is a robot, whose shape is similar to that of a doll, and which has the capacity to learn, imitate and communicate. Through simple phototaxis behaviour, the robot can imitate (mirror) the arms and head's movements of a demonstrator. The robot is controlled by a Dynamical Recurrent Associative memory Architecture (DRAMA), which allows learning of time series. We carry out experiments, where the robot is taught to perform different sequences of actions and to label these action sequences with different 'names'. In further experiments, the robot is taught combination of words, which form grammatically correct sentences, to describe its actions and perceptions of touch on different parts of its body. Finally, we carry out tests with 5-years old children, who teach the robot words to label different parts of its body and simple action sequences.
Results demonstrated the validity of the imitative strategy for teaching a robot complex sequences and combinations of sensor and actuator inputs. In particular, tests with children suggest that the imitative and communicative behaviours of Robota makes it an interesting toy for children. Moreover, if Robota was provided with more complex actuator capabilities, it would also make an interesting robotic platform for research on human-robot interaction and especially on robot learning from demonstration and by imitation.