Interactive learning in human-robot collaboration

Abstract

In this papel: we investigated interactive learning between humon subjects and robot experimentally, and its essential characteristics ore examined using the dynamical systems approach. Our research concentrated on the navigation system of a specially developed humanoid robot called Robovie and seven human subjects wliose eyes were covered, making them dependent on the robot f o r directions. We compared the usual feed-forword neural network (FFNN) without recursive connections and the recurrent neural network (R"). Although theperformances obtained with both the RNN and the FFNN improved in the early stages of learning, as the subject changed the operation by learning on its own. all performances gradually become unstable and failed. Results of a questionnaire given to the subjects confirmed that the FFNN gives better mental impressions, especially from the aspect of operabili@. When the robot used a consolidation-learning algorithm using the rehearsal outputs of the RNN, the perfarmonce improved even when interactive learning continued for a long time.The questionnaire results then also confirmed that the subject's mental impressions of the R" improved significantly. The dynamical systems analysis of RNNs support these differences.

Active Learning for Autonomous Intelligent Agents: Exploration, Curiosity, and Interaction

“…Another reason to use interactive systems is that when the users train the system they might become more comfortable with using it and accept it. See the work from (Ogata et al, 2003) for a study on this subject. The queries of the robot will have the dual goal of allowing the robot to deal with its own limitations and give the user information about the robot's uncertainty on the task being learned (Fong et al, 2003;Chao et al, 2010).…”

Multi-method learning and assimilation

“…For example, Fitzpatrick and Metta [21] studied learning object affordances as causal relations, while Ogata et al [22] focussed on learning a walking hand-in-hand behavior. However, most researches involved with machine learning of humanoids thus far have focused on a single task category, adopting only a single learning method.…”

Progress and prospects of the human–robot collaboration

“…Similar to the humans' anticipatory (feed-forward [Shadmehr and Mussa-Ivaldi, 1994b]) or feed-back [Todorov and Jordan, 2002] mechanisms to develop a suitable motor behaviour, the collaborative robots' response to sensory input can be achieved through model/knowledge based techniques [Tamei and Shibata, 2011, Ogata et al, 2003, Kimura et al, 1999, Magnanimo et al, 2014, the implementation of feedback controllers with pre-set interaction modalities [Peternel et al, 2016c, Donner andBuss, 2016a] or a combined approach [Rozo et al, 2013, Peternel et al, 2016b, Lawitzky et al, 2012b, Palunko et al, 2014. A key strategy in this direction is the establishment of a shared authority framework in which the significant capabilities of both humans and robots can be exploited.…”