If you have always been a Terminator movie fan and firm believer of the “robots are going to take over the world” theory, then here is something to prove you are right. Scientists at the Italian Institute of Technology have developed this little iCub, which has learned to shoot a bow and arrow, all by itself. It hit the bull’s eye after eight tries. Eight. I don’t think I could do that. Imagine a robot that can make a kill once in every 8 times, it hits its target. Pretty scary, isn’t it?
The learning algorithm, called ARCHER (Augmented Reward Chained Regression) algorithm, was developed and optimized specifically for problems like the archery training, which have a smooth solution space and prior knowledge about the goal to be achieved. In the case of archery, we know that hitting the center corresponds to the maximum reward we can get. Using this prior information about the task, we can view the position of the arrow’s tip as an augmented reward.
ARCHER uses a chained local regression process that iteratively estimates new policy parameters which have a greater probability of leading to the achievement of the goal of the task, based on the experience so far. An advantage of ARCHER over other learning algorithms is that it makes use of richer feedback information about the result of a rollout.
RobotCub is a 5 years long project funded by the European Commission through Unit E5 “Cognitive Systems, Interaction & Robotics“. iCub Robot also looks pretty cute in a head dress.
The iCub archery project will be presented at the Humanoids 2010 conference in December. If you can’t make the conference, but want a better idea of the algorithm behind this intriguing robotics project, check out the paper (pdf).
TechChunks is a Technology Geek, Web Entrepreneur, SEO Consultant and Social Media Evangelist. Prior to starting this blog, TechChunks has spent many productive years as a Software Engineer, Wordpress Blogger, Corporate Trainer, Frequent Conference Speaker and Workshop Leader. Circle TechChunks on Google+!