Improving Emergency Response and Human- Robotic Performance 8 th David Gertman, David J. Bruemmer, and R. Scott Hartley Idaho National Laboratory th Annual IEEE Conference on Human Factors and Power Plants Joint with Human Performance and Root Cause Trending (HPRCT) August 26-31st 2007 Monterey California
Overview Emergency response is a critical part of preparedness for chemical, biological, radiological, and nuclear (CBRN) incidents at nuclear power plants (NPPs) Response can be improved through collaborative use of new robotic capabilities Gains include reduced operator exposure, workload, and stress and improved mapping. New robotic response capabilities include casualty identification and mapping of radiation levels, interior walls, major landmarks, and obstacles Robotic advances include simplified tasking and control, incorporation of response-oriented behaviors, mapping and localization, data fusion, sensor abstraction, reduced communications, and reduced manpower requirements
Background Emergency response is covered by the US NRC Federal Radiological Emergency Response Plan (FRERP) (1996) Any peacetime radiological emergency Authority for response is detailed in Executive Order 12241 US NPPs participate in reviewed emergency exercises every two years (source NEI 2007) NRC requires plants to conduct training drill during alternate years The extent to which robots are deployed is decided on a local level. We want to support the decision making process We want to design- build- test single and multi-robotic collaborative behaviors to be used to aid rapid exploration and event characterization
Our Philosophy Regarding Robot-Human Collaboration How much autonomy is best? We believe in mixed initiative control as the logical answer to this question, especially where the operator s view of the robot world is constrained We pursue the answer combining 3-D visualization, improved sensor capability, enhanced robot initiative and intelligence including spatial reasoning, and improved metaphors for control At the end of the day, the answer is context and user specific
Determine User Issues Seek input from responders such as explosive ordinance disposal (EOD) technicians and weapons of mass destruction (WMD) civil response teams (CST) for dirty bomb detection Many with robot experience in-theater Responder issues: Interface design, system weight, communications drop-out, high operator work load, personnel protective gear and heat stress, need for back-up team, difficulty in mapping interiors and hazards, establishing safety boundaries
Develop Robot Behaviors and Controls Focus development on user issues Develop exploration, navigation, and radiation field mapping behaviors for indoors/outdoors Incorporate simultaneous localization and mapping (SLAM) Incorporate perceptual capabilities for 3D ranging and utilize for detecting and avoiding negative obstacles Incorporate radiological sensor and controls into robots Develop radiological field abstractions and indications Incorporate human presence/casualty detection behaviors
Design Approach to HRI INL overarching design concepts include Intelligence should reside on the robot itself Keep complexity from the user Develop next generation controller Enhance data fusion Employ end-users as designers and participants Develop communications protocols whereby robot can communicate from within shielded structures
Tasks and Objectives Develop and Assess Behaviors, Tools, Controls & Interface Establish performance metrics, assessment methods, and acceptance criteria for experiments. Conduct human factors experiments and review of activities Conduct Experiments verify usefulness of potential behaviors: Obstacle avoidance, guarded motion, mapping and localization, path planning, waypoint navigation.
Design Experiment, Metrics, and Assessment Methods Experiment Training delivered on-site Independent variables experience, level of robot autonomy Three modes of robot control (all have video feed) Joystick, Joystick with Map, Target Mode with Map Counterbalancing Facility configuration, source placement, order of display conditions Metrics Robot initiative - Localization accuracy Time to localize sources - Distance traveled Situational awareness - Workload Debriefs - Preference
Dirty Bomb Experiment (July 07) Building 613 irobot Packbot Radiation Detector INL Critical Infrastructure Test Range Complex
Critical Infrastructure Test Range Complex
PER-613 Experimental Facility
Dirty Bomb Experiment Objectives Responding to the needs of EOD and CSTs Click and go behavior to go down range and return, but allow direct human control within critical / target area Response: Provide dynamic autonomy tasking tools for navigation and manipulation. Want better situation awareness at the target area Response: Localize depth mosaics of selected entities / terrain within 3D visualization of robot and map data Currently using paper and pencil to draw map and hazards Response: Multiple threats displayed in real-time within fusion of robot-generated map and terrain data Need to know / display safe vs. unsafe zones in real-time Response: Scalable thresholding with automatic abstraction / normalization of hazard sensor data Communications is major limitation for robot use Response: Multi-modal, long range communications adapt to changing connectivity & bandwidth
Dirty Bomb Experiment Objectives Responding to the needs of EOD and Civil Response Teams Click and go behavior to go down range and return, but want more direct human control within critical / target area Response: Provide dynamic autonomy tasking tools for navigation and manipulation. Want better situation awareness at the target area Response: Localize depth mosaics of selected entities / terrain within 3D visualization of robot and map data Currently using paper and pencil to draw map and hazards Response: Multiple threats displayed in real-time within fusion of robot-generated map and terrain data Need to know / display safe vs. unsafe zones in real-time Response: Scalable thresholding with automatic abstraction / normalization of hazard sensor data Communications is major limitation for robot use Response: Multi-modal, long range comms. adapt to changing connectivity & bandwidth
General visualization
FY-07 Source localization findings Joystick Joystick + Map Target + Map EOD % Excellent 57% 67% 100% % Fail 29% 0% 0% CST % Excellent 43% 50% 50% % Fail 43% 17% 0% NE % Excellent 50% 50% 75% % fail 25% %0% 25%
FY-07 Dirty bomb study findings Map Analyzation Score for End Users 25.00 20.00 Score 15.00 10.00 Video Only Video and Map Target 5.00 0.00 2 3 4 5 6 7 13 8 9 10 11 12 14 15 Subject #
Conclusions Increased autonomy was associated with user preference and superior task performance Work is a work in progress, additional testing scheduled for October Employ end-users, the performance differences can be significant Efforts require a team Deploying intelligence on the robot can lead to substantial performance gains Use of robots in emergency response has the potential to reduce exposure and personnel requirements but the behaviors must be acceptable to end users
INL Government and Industry Collaboration Partners NASA Ames Research Center DoD Technical Support Working Group SPAWAR 101st Weapons of Mass Destruction Civil Support Team (WMD- CST) from the Idaho National Guard Joint Ground Robotics Enterprise (JGRE)Program Nuclear Regulatory Commission (NRC) Office of Nuclear Security and Incident Response Vendors: irobot, Foster-Miller, Segway, SRI University of Texas Las Vegas Law Enforcement RAJANT Corp Others: FEMA, DHS