The RoboEarth project aims to build a cloud computing platform for robotics. After four years of research, scientists at
Eindhoven University of Technology (TU/e), Philips and four other European universities will present this online platform through which robots can learn new skills from each other worldwide.
It is part to a €5.6m ($7.6m) European Union research project known as RoboEarth. The work is a collaboration between six European universities and Philips, the Dutch electronics company.
The project seeks to create robots’ very own cloud: a vast network, database and computation engine “where robots can share information and learn from each other about their behavior and their environment.”
RoboEarth, then, is cloud storage and computing for robots: its database is intended to store knowledge created by both humans and robots in a robot-readable open format. To date, that knowledge is basic: maps to help mobile robots navigate; task information like how to pick up a cup; and object-recognitiondata such as digital models of real-world objects.
The system also lets robots offload some of their computational tasks to “a powerful and secure” cloud-computing engine.
Most robots today perform discrete, pre-programmed tasks related to a small set of objects in a controlled environment. Some already use data-sharing systems, although almost all are proprietary. For example, the “autonomous robots” made by Kiva Systems, which is owned by online retailer Amazon, pool data about the constantly changing warehouses in which they operate, enabling them to navigate and work more efficiently.
The challenge, notes Markus Waibel of the Institute for Dynamic Systems and Control in Zurich, and one of the scientists behind RoboEarth, is that the “nuanced and complicated” nature of life outside these controlled environments cannot be defined by a limited set of specifications. In other words, to perform complex and useful tasks in the unstructured world in which humans actually live, robots will need to share knowledge and learn from each other’s experiences. They will also have to learn patterns that humans take for granted. For example, milk is usually kept in a refrigerator; the strange objects next to plates are usually knives and forks; glass objects break easily; and so on. In short, they will have to evolve and adapt to the real world, and do so autonomously.
A second robot, accessing the data supplied by robot number one, unerringly picked up the juice and carried it to the bed. Unfortunately this test did not end in total success as Amigo then dropped the milk to the floor after delivering it to the bed-ridden patient.. But as Dr Waibel—who was the “patient” in question—points out, the demonstration still proved its point.
RoboEarth and a handful of similar initiatives, including Google’s cloud-robotics ROS platform, do, however, raise some questions.
As with many internet of things developments, security is variable, and physical machines that work alongside people in the real world will be tempting targets. Although RoboEarth’s cloud may itself be secure, it will take only one vulnerable node (robot) for hackers to gain control—a scary thought in a world where robots might be care-givers with access to private medical information and other personal data.
Lee Tien, a senior attorney at the Electronic Frontier Foundation, which seeks to protect individual rights in a digital world, worries that the fundamental requirements of projects such as RoboEarth—the collection, storage and sharing of as much data as possible—are at odds with the goal of “privacy by design” that many policymakers now support. If a robot is caring for a sick or elderly person, what happens if there is a family dispute over that care? Or a contentious divorce? Is all the robo-cloud data simply up for grabs by whomever manages to obtain or subpoena it? “The number of safeguards that will need to be put in place is staggering”, Tien told the Economist.