Scouting disaster-affected areas is a tricky and risky task for humans to do manually. In manual attempts, humans must carry out search and rescue operations fast enough to assist as many victims as possible despite the extremely perilous conditions. Due to the associated inefficiencies and risks in manual search and rescue attempts, many human lives of survivors and rescue workers have been lost in the past in several earthquakes and nuclear crises. Autonomous robotic systems are now being developed to work with specialist remote operators, who attempt to locate casualty signs from the data sent by robots in the field and direct medical and other support teams to disaster sites as required. The CSIRO’s Robotics and Autonomous Systems group recently implemented such a demonstrator system, winning 2nd place at the 2021 DARPA Subterranean Challenge Finals. It consists of a heterogeneous robot fleet, which can be deployed to an unknown disaster site to explore the site and locate signs of victims autonomously.
These types of autonomous robotics systems reduce the risks to search and rescue crews, but they can be further improved by ensuring experts and robots work together effectively. As part of the CSIRO’s Collaborative Intelligence Future Science Platform, we are now investigating how a human robot team can best work together, leveraging the intelligence and experience of both. We are specifically researching how to provide humans and robots members with optimal situational awareness about each other and the operating environment to collaborate better. One approach being studied is using users’ biosignals (e.g., eye gaze, speech and heart rate) to generate estimates of user’s situational awareness, stress and workload and use that knowledge to make the system more adaptive to the remote operator. Another approach is to provide the human-robot team with the option to communicate using natural language to reduce the situational awareness gaps, while visual and haptic interactions also support the user to receive updates efficiently. The outcomes of our research are to ensure an effective human-robot collaboration, with superior performance comparing to humans or robots on their own, and support human trust in the robots. We intend to explore these issues in diverse domains, e.g., search and rescue, smart farming, mining and even space.
