Towards Teammate-Aware Active Search for Human-Multi-Robot Teams in Adverse Environments

Sherif Bakr, Brian Reily, Christopher M. Reardon

Confirmation: I have read and agree with the workshop's policy on behalf of myself and my co-authors.

Keywords: Human-Multi-Robot Teams

Abstract:

In the context of challenging and dangerous environments,
robot teams are increasingly employed in various
applications such as search-and-rescue missions in unstable,
post-disaster areas, mine rescue operations, and military patrols
in contested zones. This paper addresses the challenge of
teammate-aware active search, focusing on the robots’ ability
to locate targets of interest and maintain communication with
teammates to ensure safe operation under adversity. Our approach
leverages multi-agent reinforcement learning techniques
to enable robots to robustly search for targets of interest using
multi-sensory information while maintaining communication
with at least one other teammate. The robots utilize a prior
map indicating probability distributions of potential targets
in the environment, enhancing their search efficiency. Human
operators are integrated as part of the agent team. Humans
can provide real-time input and feedback to adjust the robots’
strategies based on their observations and capabilities that
robots do not possess. This collaboration allows for an exchange
of information between the robots and the human member, utilizing
both the speed of robots and the understanding of human
members. This synergy between the high robotic precision and
speed and the human intuition creates a robust framework
for teammate-aware active search operations. By incorporating
the human component into the loop, this approach insures
that the human perspective remains central and critical to the
mission. The interactive AI system prioritizes human situational
awareness, allowing operators to make adjustments in real-time.
Through this integration, we aim to create a balance between
the strengths of both, humans and robots, ensuring successful
outcomes in adverse and complex conditions.

Submission Number: 9