The Defense Advanced Research Projects Agency announced a major new initiative to create robotic autonomous manipulators that mimic the human hands, an agency program manager said.
For the past several decades, the research agency and the robotics community have concentrated their efforts on programming ground robots to get from point A to point B, said Robert Mandelbaum, a DARPA program manager who focuses on robotics and autonomous systems. That challenge has for the most part been tackled, he said.
The autonomous robotics manipulation program will take on a new goal, creating an inexpensive hand-like device that is as adaptable as a human appendage.
“We are looking for software that can span a variety of spaces. We’re not looking for any particular application,” he said at an Association for Unmanned Vehicle Systems International conference.
In other words, for any single task, engineers can design a robotic tool that can do better than a hand. For example, for putting in a screw, they would simply design an arm with a screwdriver on the end.
“We want general adaptability. We want a hand that can do multiple tasks,” Mandelbaum said.
There are myriad possible applications including counter-mine, the ability to remove rubble in search-and-rescue missions, weapons support, explosive ordnance disposal, casualty care and prostheses. They are also needed in extreme environments such as space, he said.
As far as prosthetics, there has been progress in pushing World War II-era technology for artificial arms into the modern age. There are now robotic appendages, but for each individual action there must be a command. DARPA wants to create arms and hands that will mimic the human mind’s ability to pick up a grape without having to think about what each individual finger must do, he said.
“We would like to give that kind of low-level control to prosthetic arms to really give them the ability to have the same dexterous manipulation as regular people,” he said.
Unlike industrial robots with arms that do the same action over and over again, the challenge will be to create arms that can do one task well, and then be able to a completely different job immediately afterwards, he said.
The program will ask teams working on the program to accomplish three major operations.
The first will be for the robot to open up a duffel bag, search through the contents and find a revolver that is hiding inside. That requires the ability to handle flexible material like clothing, which has been a challenge for robots in the past, he said. That will require “force feedback,” or the ability to feel for objects, rather than just relying on visual cues. It will also require bimanual coordination, the ability to use two hands at once. It must use one hand to hold onto the bag and the other to open the zipper.
The next task will be rubble removal in a search-and-rescue scenario. The robot will have to pick up oddly shaped objects with both arms. The third challenge will require that the arms insert one object into another — in this case a shell into a mortar.
There will be two tracks: software and hardware.
For the software, DARPA will provide the hardware in the form of robotic arms that the teams can upload its software into. The agency doesn’t want the competitors to spend their time and money integrating software into different models of robotic arms. The robot will start out with one arm picking up and using rigid objects “just to get the teams’ feet wet” and then graduate to two arms and more difficult to manipulate objects such as cloth, he said.
The main challenge in the hardware track will be creating inexpensive arms.
If these hands are to be ubiquitous, they cannot have astronomically high price tags, he said.