The effect of display type on operator prediction of future swarm states

Abstract

Abstract-Large teams of robots that operate collectively, whose behavior emerges from local interactions with neighbors, are known as swarms. While significant progress has been made improving the hardware, communication capabilities, and autonomous operation of these swarms, we still have much to learn about how human operators control and interact with them. This research is necessary if real world swarms are to be deployed in the future. The study presented here investigates different methods of displaying information about the swarm state to operators, and asks them to make predictions about the swarm's future state. In the study, participants are shown swarms performing one of three different behaviors, and are asked to use the information available from the display to make their predictions. Results show that summarizing the swarm's current state to just an average position and bounding ellipse allowed predictions as accurate as those made when full state information was shown. Furthermore, two leader-based methods were used, whereby the operators were shown only a small subset of the swarm. However, such display methods were inferior for prediction than either the summary center and ellipse or full information methods. With these results, and with participant feedback about the helpfulness of the four display types, we hope future studies can make more informed decision about interface design when it comes to the control of swarms.

Transparency: Transitioning From Human–Machine Systems to Human-Swarm Systems

“…Abstract swarm visualizations were found to perform better, with respect to correct decisions made, versus individual swarm entity visualizations (Seiffert et al, 2015b); however, some evidence contradicts this finding (Roundtree et al, 2018). A human operator's ability to predict a swarm's future state was evaluated based on three swarm visualizations (Walker, Lewis, & Sycara, 2016), as well as using predictive classifiers to indicate swarm motion, from flocking to a torus (Brown, Goodrich, Jung, & Kerman, 2016). Accurate predictions were only feasible when the human operator allowed a swarm to stabilize before issuing new commands, which is referred to as neglect benevolence.…”

Understanding and facilitating predictability for engagement in learning

“…Thus we opted for responsive robot actions, which are actions that are at risk of being perceived as unpredictable when the event the robot is responding to is not clear -a scenario that may well occur in real HRI's. In previous studies on predictability in HRI, behavioural predictability was also measured by asking participants to make a prediction (Takayama et al, 2011;Lichtenthäler et al, 2012a,b;Driggs-Campbell and Bajcsy, 2016;Walker et al, 2016) and indicate their certainty of their prediction (Takayama et al, 2011). However, in these studies, the robot's behavioural predictability was manipulated through limiting the information available to participants by having the robot perform fewer communicative actions (Takayama et al, 2011) or by presenting different and less information in a graphical user interface (Walker et al, 2016), or through how the robot moved (Lichtenthäler et al, 2012a,b;Driggs-Campbell and Bajcsy, 2016;Walker et al, 2016).…”

“…Key is to take the attributed predictability of a robot as an important outcome measure that can be used to evaluate different designs (e.g. as in Walker et al (2016)), and can lead to humanrobot interactions that are easier to understand. While we cannot draw conclusions regarding the behavioural predictability of a robot -due to a ceiling effect -we do believe that both behavioural and attributed predictability are important for effective HRI, given the central role of predictability in human perception.…”

“…While these motions can result in the same motion in certain situations, in highly ambiguous situations, legible motions allow a person to predict its goal, but its trajectory is unpredictable, and vice versa for predictable motions . Following on from predicting the end-state of a robot action, other studies are about predicting what action the robot will take after the on-going action (Takayama et al, 2011;Lichtenthäler et al, 2012b;Mubin and Bartneck, 2015;Driggs-Campbell and Bajcsy, 2016;Walker et al, 2016). And finally, people can also attribute predictability as a quality of a robot Dragan and Srinivasa, 2014a;Dragan et al, 2015;Mubin and Bartneck, 2015;Driggs-Campbell and Bajcsy, 2016).…”

“…In such studies, behavioural predictability has been measured through behavioural observations, for example by asking participants to make a prediction (e.g. Driggs-Campbell and Bajcsy, 2016;Takayama et al, 2011;Lichtenthäler et al, 2012a,b;Walker et al, 2016), or subjectively through self report, by asking participants how predictable they judge the robot to be (e.g. Dragan and Srinivasa, 2014a;Dragan et al, 2015;Mubin and Bartneck, 2015).…”