June 18, 2019
In the future, your co-worker might be a robot. Luis Sentis, a professor in the department of Aerospace Engineering and Engineering Mechanics is leading research that focuses on making that collaboration as natural as possible – with robots that move like humans and learn like them, too.
The Office of Naval Research recently awarded Sentis a $2.7 million grant that funds research into these robots. Sentis is the principal investigator of the project and collaborating on the work with Apptronik, a UT research spin-off company, and the artificial intelligence company Thinking Robots. The goal of the research is to investigate the productivity of mixed human-robot teams during logistics and search tasks. The companies will help with software integration so that the robots are able to walk, talk and efficiently carry out tasks alongside actual humans.
Most importantly, that work should come natural to both human and robot alike, Sentis said. That means no specialized robotics training for the human operator nor clunky or unsafe movements for the robot.
“Humans working with humanoids should be more productive than humans alone,” Sentis said.
The software for the “brains” of the robot is called DIARC (distributed integrated affect, reflection, cognition) and was developed by Tufts University and Thinking Robots. DIARC allows the robot to perform deductive reasoning, a feature that allows it to make educated assumptions based on prior knowledge. In addition, it can learn new knowledge through verbal instruction, watching videos or watching other humans or robots at work – the same ways people are taught.
The software for the “brawn” of the robot is called DYNACORE. It was developed by UT’s Human Centered Robotics Lab, which is led by Sentis, and organizes the movements of the robot so that the limbs move fluidly and in sync with each other as well as adjust to accommodate new terrain and obstacles for effective manipulation of the objects.
“When you have lots of body appendages and teaming agents, the problem is how to program robots so you can tell them what to do,” Sentis said. “There’s mobility, safety, manipulation, human-robot interaction and more all which require whole body control and multi-robot coordination.”
Sentis and his research team will be working to integrate both DIARC and DYNACORE in a humanoid robot built and serviced by Apptronik. The humanoid shape isn’t just for looks. Arms and legs allow a robot to navigate human environments as humans do, allowing precise navigation and manipulation in tight environments, such as ships, homes or even space outposts. This idea is why NASA selected the humanoid form for its Valkyrie robot, an astronaut assistant that Sentis helped design and test but is not directly related to the grant research.
The robot body for this project is in development. The legs and lower body are complete while the upper body is still being built. Sentis said that the research will integrate everything together – DIARC, DYNACORE and the body parts they’ll ultimately be controlling – and put the robot through its paces by performing team-oriented projects that involve working alongside people and other more task-specific robots.
