Learn AI and Human-Robot Interaction through Asimov's I, Robot Stories (2019)
Robotics Through Science Fiction (MIT Press Fall 2018)
2012 Defense Science Board Study: The Role of Autonomy in DoD Systems
ACM Webinar on Computing for Disasters (2018)
Disaster Robotics (MIT Press 2014)
Computing for Disasters: A Report from the Community Workshop
Research Topics and Resources
My primary research is in artificial intelligence for mobile robots as applied to disaster robotics. As my analyses have shown that 50% of the terminal failures in disaster robotics are due to human error, a significant portion of my work is in human-robot interaction. My research uses a field methodology, working with responders and agency stakeholders to determine gaps that lead to the formulation of applied and fundamental research thrusts. We have participated in 27 disasters or incidents and over 35 exercises gathering data spanning urban search and rescue, structural inspection, hurricanes, flooding, mudslides, mine disasters, radiological events, and wilderness search and rescue.
Disaster deployments: see the materials at the Center for Robot-Assisted Search and Rescue crasar.org. Our most recent deployments have been for the Hawaiian volcanic eruption and Hurricanes Harvey and Irma. We are continuing to support the use of robots- especially unmanned marine vehicle such as EMILY, wireless communications, and AI for preventing drowning deaths from boat refugees. This is with funding from donations and the National Science Foundation.
Our fundamental research covers:
- Human-robot interaction. One theme is how distributed team members can use natural user interfaces to communicate with the robot and with other decision makers in real-time. We can predict unsafe human factors pre-conditions using our shared roles model without the extensive task and user model required by methods such as SOAR. Our Skywriter interface for small UAVs that allows a user with no experience to exploit a visual common ground with the robot and robot Pilot and use sketch, pointing, and multi-touch. It has been licensed. Another theme is our Survivor Buddy project which explores how trapped victims will react to a rescue robot, how victims can use gestures to control their "buddy robot" and how medical practitioners will use robots. Our work in UAVs for evacuation discovered that people don't appear to respond to UAVs the same way they do to ground robots, which introduces safety concerns.
- Autonomy. Our work follows the behavior-based robotics paradigm, using insights from cognitive and biological studies of intelligence. One theme is how to share responsibility of autonomous roles between the human and robot, especially what autonomous capabilities must the robot reliably perform as the human cannot react fast enough or would require more manpower. One theme is fault detection and recovery, using AI generate-and-test methods to rapidly recover, deferring fault identification to after the robot is in a safe state. Another theme is determining appropriate path planning strategies for working in three dimensions, either ground robots searching a collapsed building, a marine vehicle inspecting the substructure of a bridge, or a UAV inspecting a building or mudslide. We conducted the first work in marsupial robots and our AirJoey project under funding from DOE and the National Science Foundation NRI program is
exploring how to use a small UAV or UGV to provide external viewpoints to help a primary robot perform complex tasks such as opening a door.
- Dimensionless numbers and metrics for capturing field performance. Our work has established metrics for traversability that are independent of the scale of the robot and environment. We have identified constraints that impede the actual deployment of a robot in the field. We have created new methods for field work, including viewpoint-oriented cognitive work analysis and a formal computational model of the data-to-decision process.
Another way to think about our research is by platform. We work with ground, aerial, and marine vehicles. Because we are a field robotics group and deploy to disasters, we prefer to using commercially available, hardened robots rather than build our own. We often use platforms donated through the CRASAR Roboticists Without Borders program.
- Small unmanned aerial vehicles. We work with both small fixed-wing and rotor-craft having the first deployment of either for a disaster (Hurricane Katrina). Our ongoing outdoor rotor-craft work since 2005 has concentrated on structural inspection (flying with 10 feet of structures with GPS denied) and indoor flight, hazardous materials spills including chemical and radiological since 2011, and geospatial reconstruction of flooding, debris, and other elements of disasters since 2013. Our indoor rotor-craft work is split between general indoor navigation and using SUAS for evacuation.
- Man-packable and man-portable ground vehicles. Our UGV work focuses on autonomy and UGV-UAV cooperation. Our newest work is in collaboration with CMU and Georgia Tech with snake robots working in granular materials (aka burrowing robots).
- Unmanned surface and underwater marine vehicles. Our UMV work focuses on underwater structural inspection, marsupial USV-ROV teams, and victim recovery. We also explore UMV-UAV teaming as a result of our experiences assisting with the Tohoku Tsunami response.
Calendar of Events
ImPACT Tough Robot Challenge Nov 12-15, 2018
Winter Institute: FirstNet for Disasters Oct. 22-23, 2018
IROS Workshop on Human-Aiding Robotics Oct. 5, 2018
IROS Workshop on Tough Robot Challenge Oct. 1, 2018
IEEE Safety Security and Rescue Robotics Aug. 5-8, 2018