A research team at the Robotics and Intelligent Systems laboratory at the University of Oslo’s Department of Informatics is in the process of designing and programming 3D printed robots that can solve complex tasks in situations where humans cannot be present — for instance, in oslohazardous landslide areas, compromised nuclear power plants, or deep mines on faraway planets.
The robotics team has designed three generations of self-learning and self-repairing robots. The first robot, a “chicken robot” the team referred to as “Henriette,” taught itself to walk and leap over obstacles. When Henriette lost a leg, it learned without help from its designers and programmers to move about on the one remaining leg.
The second generation of self-learning robots, developed by masters student Tønnes Nygaard, was designed based on a simulation program that calculated what the robot’s body should look like — for instance, how many legs it should have, how long they would be, and what the robot 4 legdistance between them would be. Basically, the robot designed itself.
The third and most flexible generation thus far was design fully by the simulation program, which suggested the ideal number of legs and joints for the completed, self-learning and self-repairing robot. According to Associate Professor Kyrre Glette, the process works as follows: “We tell the simulation program what we would like the robot to do, how fast it should walk, its size and energy consumption.” The program runs through thousands of possible configurations and arrives at the best models in a process of artificial evolution.
As the team progressed through the three generations of design, the process became more complicated as they wanted the robots to perform increasingly more complex tasks. The robots, which were all produced via 3D printing, are tested for functionality. The team discovered during the tests, however, that the robots’ “real-world functionalities quite often prove[d] to be different from those of the simulated versions,” as Professor Mats Høvin, another team member, noted.
http://www.innomind.org/wp-content/uploads/2015/05/robot-research-team-members-300x200.jpg 300w" sizes="(max-width: 507px) 100vw, 507px" />
Closing the gap between the robots’ capacity to learn and practice at the simulation program stage and the real world is currently the challenge of the robotics team. One challenge they’ve given their robots is to test how they confront obstacles as, ideally, one of the primary functions of the self-learning and self-repairing robot will be to respond on its own to unforeseen problems. For example, one scenario the team provided was this: the robot enters the compromised nuclear power plant and encounters a staircase that had not been expected. It responds by taking a photograph of the staircase, analysing the photograph, and then, equipped with its own printer, printing and installing a part that will allow it to navigate the staircase.
robotIn another scenario, a self-learning, self-repairing robot sent into a deep mine on a distant planet would, for example, need to have the capacity to navigate over uneven terrain, climb boulders, and change direction when necessary. As it encountered problems, it would analyze the situation and respond by possibly adding necessary parts — for instance, augmenting its two- or four-legged design and adding another pair of legs that would allow it to crawl crab-like across a rugged surface (as seen in the video).
3D printing is invaluable both in creating the original models of the robots and in its role as an on-board tool for self-enhancing and -repairing in scenarios like the one cited above. “A 3D printer,” elaborates Høvin, “will construct whatever you want it to, layer by layer. This means you won’t have to bother with molds, and you can produce seemingly impossibly complicated structures as a single piece.”
The University of Oslo uses 3D printers that cost between 400,000 NOK (Norwegian Krone, or around $58,000 USD) and 3,000,000 NOK (or about $440,000 USD). As a general rule, of course, the more expensive the printer, the more sophisticated and the better the detail. It isn’t clear at this stage of the research and prototyping what caliber of 3D printer the self-learning, self-repairing robots will utilize.