The new $17.2 million Laboratory for Autonomous Systems Research (LASR) at the Naval Research Laboratory (NRL) in Washington, D.C., is developing autonomous firefighting robots that will give our warfighters more capability.
LASR director Alan Schultz—who also has the title of director of the Navy Center for Applied Research in Artificial Intelligence at NRL— says the new facility will support a broad range of multi-disciplinary basic and applied research related to autonomous systems.
An interdisciplinary team of researchers at LASR is collectively developing technology for advanced shipboard damage control that includes the development of an autonomous humanoid robot that could fight fires on the next generation of combatants. At LASR, robots are being developed to do difficult and dangerous jobs, but they can also think for themselves and advise their human teammates.
For example, NRL’s Lucas and Octavia are robots that are programmed to work as a part of a ship’s damage control team. These robots can track commands and gestures, , and will respond with “voice,” gestures, and a few facial expressions of their own if they don’t understand or are confused by a command or instruction. Tell them to do the wrong thing, and they’ll ask for clarification.
The robots have human facial features, and can even raise an eyebrow when perplexed. Dr. Greg Trafton, a roboticist at the NCARAI Information Technology Division says that Lucas and Octavia appear “human-like” so they work well with their human counterparts, but more importantly, use computational cognitive models that allow the robot to reason similarly to humans.
The next generation of “mobile, dexterous and social” firefighting robots is SAFFIR (Shipboard Autonomous Firefighting Robot), which will be able to walk around the ship, and use equipment such as hoses and extinguishers just like humans do.
In addition to the different scientists at NRL, universities such as Virginia Tech and the University of Pennsylvania are involved with NRL’s firefighting robot project. SAFFIR is being developed by Virginia Tech, based on their CHARLI-L1 robot.
Because the shipboard environment has been designed for humans, with narrow and sometimes crowded passageways, ladders, and doors with handles, SAFFIR is being designed with a human form so it can hit the deck running.
Communicating in a damage control situation is difficult for humans in loud, hot, smoke-filled spaces. So how will a robot perform? Think of how noisy it is on an aircraft carrier flight deck, yet communications is very important. The Sailors and air crews communicate through gestures that are anticipated and understood. By allowing the robot to understand both gestures and language, there is a better chance that the robot can resolve the intent without making a misunderstanding.
“A crisis situation is stressful for humans,” Trafton says. “Humans respond to facial expressions. Having a robot with real, calm expressions can help to keep a tense situation under control. ”
Schultz says the goal is for the robot to understand spatial perspective, and even possess what is known as theory of mind so it has the ability to think about what other people think. “If the robot can ‘model’ another person, it can think about what that person can and will do, wants to do, and is capable of doing.”
Octavia and Lucas, as well as SAFFIR in the future, are equipped with a firefighting backpack. The robots have a camera and infrared sensor so they can locate heat sources and direct a water or foam stream at the hot spot. They can hold a hose, and manipulate the nozzle handle or the trigger on an extinguisher.
Schultz says the robots have to be able to move around the ship on their own, but they are not completely autonomous. “They have to work with human teammates. We’re developing models and algorithms and computational cognitive modeling to create a peer, a team member.”
“The robot needs to understand and instill trust in its human team members,” Schultz says. “We start with models of human reasoning and processing to learn how humans solve tasks. Then we apply that to robots.”
The robots are programmed with algorithms to follow the damage control team leader—and understand what the team leader is focusing on. When the team leader points, gestures or speaks to the robot, it will know what is being communicated, and what to do.
Schultz says that NRL is uniquely positioned to lead work in this field because of the “breadth and depth of our science, and because we have the underlying science needed for all aspects of autonomous systems. As I went around to NRL’s various divisions to brief our scientists about the new facility, I realized that virtually all of our research divisions have a role, whether it’s bio-molecular researchers developing CBRNe (chemical, biological, radiological, nuclear and enhanced conventional explosives) sensors, material scientists developing novel ways to embed antennas or electronics into vehicle structures, or psychologists studying how our warfighters will work with these systems.”
“The potential for cross-disciplinary work is huge, and I am already seeing folks working together to solve the bigger problems in autonomous systems that cannot be solved when these groups work in isolation,” Schultz says.