A new robot has overcome a fundamental challenge of locomotion by teaching itself how to walk. Researchers from Google developed algorithms that helped the four-legged bot to learn how to walk across a range of surfaces within just hours of practice, annihilating the record times set by its human overlords. Their system uses deep reinforcement learning, a form of AI that teaches through trial and error by providing rewards for certain actions. This technique is typically evaluated in virtual environments. However, building simulations that could replicate the robot walking on various surfaces would be highly complex and time-consuming, so the researchers chose to train their system in the real world.
For example, in Terminator XXVIII: Rise of the Earthlings (2051), a brave young android is tasked with saving the world from an army of killer humans sent from the future to destroy robotkind. Leading the human rebellion is Barry, an 18-stone unemployed bus driver from Caerphilly whose powers include the ability to eat a foot-long meatball marinara from Subway in under nine seconds. In the war zones of the future, robot generals will send human beings on to the battlefield to check for land mines and other unexploded devices. "Previously, this highly dangerous work was carried out by bomb disposal robots," explains Major-General Sir Optimus Prime. "Sending human beings instead will reduce the risk to robot life.
A car factory robot, adapted by British scientists to work in a chemistry lab, completed an experiment in three days that would have taken a human months. University of Liverpool researchers reprogrammed the £100,000 autonomous arm, giving it enough intelligence that it can perform experiments without input. The robot is significantly more efficient, able to perform up to 700 experiments in a week - the same number a student might complete over the course of a PhD. It has already made a contribution to the Liverpool lab where it is based, working around the clock to complete 688 different experiments over 172 hours. The developers said the goal was to find a way to'automate the researcher' rather than the tools that scientists use to carry out experiments.
Researchers at Google have developed a four-legged robot that learned to walk forward and backward, and turn left and right, completely on its own in just a few hours. The team developed a more efficient algorithm that could learn with fewer trials and thus fewer errors; the improved algorithm helped the robot learn to walk within two hours. In addition, the team trained the robot in a real physical environment from the beginning, rather than starting with simulations as in previous studies. Since the physical environment provided natural variation, the robot quickly adapted to other similar environments. Jie Tan, who leads the robotics locomotion team at Google Brain, believes cracking locomotion will be key to unlocking more useful robots.
Robots offer an opportunity to enable people to live safely and comfortably in their homes as they grow older. In the near future (we're all hoping), robots will be able to help us by cooking, cleaning, doing chores, and generally taking care of us, but they're not yet at the point where they can do those sorts of things autonomously. Putting a human in the loop can help robots be useful more quickly, which is especially important for the people who would benefit the most from this technology--specifically, folks with disabilities that make them more reliant on care. Ideally, the people who need things done would be the people in the loop telling the robot what to do, but that can be particularly challenging for those with disabilities that limit how mobile they are. If you can't move your arms or hands, for example, how are you going to control a robot?