Using bio-electrical signals to influence the social behaviours of domesticated robots

Abstract

Several emerging computer devices read bio-electrical signals (e.g., electro-corticographic signals, skin biopotential or facial muscle tension) and translate them into computer-understandable input. We investigated how one low-cost commercially-available device could be used to control a domestic robot. First, we used the device to issue direct motion commands; while we could control the device somewhat, it proved difficult to do reliably. Second, we interpreted one class of signals as suggestive of emotional stress, and used that as an emotional parameter to influence (but not directly control) robot behaviour. In this case, the robot would react to human stress by staying out of the person's way. Our work suggests that affecting behaviour may be a reasonable way to leverage such devices.

A Survey of Autonomous Human Affect Detection Methods for Social Robots Engaged in Natural HRI

“…Several of the affect-detection systems designed for social robots have used dimensional models, mainly consisting of valence and arousal scales for recognition from facial expressions [76,77,125], body language [5], voice [81,82], physiological signals [4,70,86,193], and multi-modal inputs [193,203]. A small number of systems have utilized alternative affect classification scales, such as accessibility [79], engagement [61], predictability [81], stance [90], speed regularity and extent [203], stress [19,85], anxiety [84,190], and aversion and affinity [213].…”

“…HRI studies have also been conducted with robots using dimensional models for affect classification using facial expressions [76][77][78], body language [5,79,80], voice [81,82], physiological signals [4,22,70,[83][84][85][86][87], and multi-modal systems [88][89][90]. The most common model used in HRI is the two-dimensional circumplex (valence-arousal) model.…”

“…These systems can also be classified based on the proposed HRI scenarios: collaborative [70,190,193], assistive [4,6,[84][85][86], and multi-purpose [72].…”

“…In [85], the OCZ NIA TM headband was utilized to measure muscle tension in a person's forehead in order to estimate his/her stress level. This stress level was then used to control the social behaviors of the Roomba®robot vacuum cleaner.…”

Design principles for robot inclusive spaces: A case study with Roomba

“…Even though Roomba has been studied extensively over the last decade, but previous efforts focuses on the improvement of mechanical design [19], control algorithms [20], multi-robot co-operation [21], human robot interaction [22], and autonomy [23] with no attention on designing a friendly space for Roomba to operate and therefore to improve its performance.…”