Proactive Intention Recognition for Joint Human-Robot Search and Rescue Missions Through Monte-Carlo Planning in POMDP Environments

Abstract

Proactively perceiving others' intentions is a crucial skill to effectively interact in unstructured, dynamic and novel environments. This work proposes a first step towards embedding this skill in support robots for search and rescue missions. Predicting the responders' intentions, indeed, will enable exploration approaches which will identify and prioritise areas that are more relevant for the responder and, thus, for the task, leading to the development of safer, more robust and efficient joint exploration strategies. More specifically, this paper presents an active intention recognition paradigm to perceive, even under sensory constraints, not only the target's position but also the first responder's movements, which can provide information on his/her intentions (e.g. reaching the position where he/she expects the target to be). This mechanism is implemented by employing an extension of Monte-Carlo-based planning techniques for partially observable environments, where the reward function is augmented with an entropy reduction bonus. We test in simulation several configurations of reward augmentation, both information theoretic and not, as well as belief state approximations and obtain substantial improvements over the basic approach.

Editorial: Active Vision and Perception in Human-Robot Collaboration

“…Scale normative frameworks to deal with realistic tasks and environments (see Van de Maele et al and Ognibene and Demiris, 2013 ; Lee et al, 2015 ; Donnarumma et al, 2017a ; Ognibene et al, 2019b ).…”

“…This is because they are contingent on previous observations, hierarchically organized (Proietti et al, 2021 ), and must extend over time, space and scene elements which may not be always visible (Ognibene et al, 2013 ). While some active recognition systems and normative models for action and social interactions have already been proposed (Ognibene and Demiris, 2013 ; Lee et al, 2015 ; Donnarumma et al, 2017a ; Ognibene et al, 2019b ), it is not completely clear what strategy humans adopt in such tasks, not least because of the heterogeneity of the stimuli. Salatiello et al introduce a validated generative model of social interactions that can generate highly-controlled stimuli useful for conducting behavioral and neuroimaging studies, but also for the development and validation of computational models.…”

Learning State-Variable Relationships in POMCP: A Framework for Mobile Robots

“…Using the MRF, we push the belief probabilities toward states that agree with this knowledge. Methodologies for optimally updating POMDP beliefs to reduce uncertainty on the true state have been proposed by Stachniss et al (2005) , Araya et al (2010) , Veiga (2015) , Ognibene et al (2019) , Fischer and Tas (2020) , and Thomas et al (2020) . However, these methods mainly focus on introducing the belief into the reward function to allow the definition of information gain goals, otherwise not definable, in the context of POMDP.…”

Monitoring and Mapping of Crop Fields with UAV Swarms Based on Information Gain

“…To this end, we exploit the Information Gain (IG), an informationtheoretic measure of the expected reduction in uncertainty from additional observations of a specific area. The usage of information theory for exploration and mapping has been demonstrated across application domains [6,16,22,21,20], and precision agriculture in particular [14,24]. Here, we exploit IG to support exploration and coordination among robots.…”