Towards situational awareness from robotic group motion

The control of multiple robots in the context of tele-exploration tasks is often attentionally taxing, resulting in a loss of situational awareness for operators. Unmanned aerial vehicle swarms require significantly more multitasking than controlling a plane, thus making it necessary to devise intuitive feedback sources and control methods for these robots. The purpose of this article is to examine a swarm's nonverbal behaviour as a possible way to increase situational awareness and reduce the operators cognitive load by soliciting intuitions about the swarm's behaviour. To progress on the definition of a database of nonverbal expressions for robot swarms, we first define categories of communicative intents based on spontaneous descriptions of common swarm behaviours. The obtained typology confirms that the first two levels (as defined by Endsley: elements of environment and comprehension of the situation) can be shared through swarms motion-based communication. We then investigate group motion parameters potentially connected to these communicative intents. Results are that synchronized movement and tendency to form figures help convey meaningful information to the operator. We then discuss how this can be applied to realistic scenarios for the intuitive command of remote robotic teams.

I. Introduction

Mobile robots are becoming more prevalent in a variety of tele-exploration tasks such as surveillance of large areas, emergency response and military deployment. However, fully autonomous robotic systems are sensitive to unconstrained and dynamic environments, and human operator inputs are required in many critical scenarios. With scenarios that bene-fit from an increased number of deployed units, the operator's load also increases quickly and diverts his or her attention from gathering contextual information from the robots. In the end, most tele-operation tasks are about gathering knowledge on the deployment site, i.e. situational awareness (SA) for the mission. With larger robotic teams, the information to be processed by the operator may become overwhelming. For instance, an Unmanned Aerial Vehicles (UAV) swarm creates a large amount of data and requires significantly more multitasking than controlling a single UAV, because multiple vehicles must be supervised at once and the data collected by these UAVs - which may be of different types (visual, infrared, audio, etc.) and from different perspectives - must be aggregated by the operator. This can overwhelm operators and lead to loss of SA [1]. Since it was shown that adding multiple types of feedback modalities reduce cognitive workload [2] and improve operator SA, we aim at exploring one specific to swarms: group motion.