Developed by researchers at the University of Massachusetts at Amherst, uBot-5 is a promising robotic platform for mobile manipulation in social telepresence applications (Fig. 3). Designed to be economical yet highly capable, durable, and safe to operate, it’s equipped with an LCD touchscreen monitor and a webcam.
Funded by the European Union (EU), researchers with the COSPAL (Cognitive Systems Using Perception-Action Learning) program have combined techniques from artificial intelligence for symbolic reasoning with artificial neural networks that associate percepts and states in a bidirectional manner.
Feedback loops are established through the continuous and symbolic parts of the system, which allow perception-action feedback at several levels. After an initial bootstrapping phase, incremental learning techniques train the system simultaneously at different levels, allowing adaptation and exploration (Fig. 4). The COSPAL architecture is expected to enable the design of systems that show autonomous behavior.
“Developing systems in classical artificial intelligence is essentially a top-down approach, whereas in artificial neural networks it is a bottom-up approach,” says Michael Felsberg, a researcher at the Computer Vision Laboratory of Linköping University in Sweden and member of the COSPAL program.
“The problem is that, used individually, these systems have major shortcomings when it comes to developing advanced artificial cognitive- system architectures. Using an artificial neural network is too trivial to solve complex tasks, while classical artificial intelligence cannot solve them if a system has not been programmed to do so,” says Felsberg.
Through its Unit E5 “Cognition” program, the EU has funded a five-year project to study the cognition and implementation of a robot the size of a two-year-old child, called the iCub (Fig. 5). The RobotCub Consortium will use it to study cognition through biologically motivated algorithms. The ultimate goal for the opensource- software project is a robot version with 54 degrees of freedom, with seven for each arm, nine for each hand, six for the head, three for the torso and spine, and six for each leg.
And, a major breakthrough in robotic-awareness capability comes from Evolution Robotics. Its NorthStar 2.0 autonomous navigation system enables office and home robots to truly be aware of their environment to perform everyday tasks with total autonomy. The first product to use NorthStar, which combines GPS, radar, and auto-pilot technologies, is the Rovio robot from WowWee Group Ltd.
DEVELOPMENT TOOLS AND STANDARDS There’s now a greater effort to provide robotic systems designers with the right hardware and software development platforms. An example of this trend was founded by inventor Dean Kamen in 1989, called FIRST (For Inspiration and Recognition of Science and Technology), which focuses on the youngest designers. Its accessible and innovative student competitions build self-confidence, knowledge, and life skills while motivating young people to pursue opportunities in science, technology, and engineering.
The Microsoft Robotics Studio, a Windowsbased environment for robot control and simulation, targets academic, hobbyist, and commercial developers. It handles a wide variety of robot hardware, with features like a visual programming tool, the Microsoft Visual Programming Language for creating and debugging robot applications, Web-based and Windows-based interfaces, 3D simulation (including hardware acceleration), a lightweight services-oriented runtime, easy access to a robot’s sensors and actuators via a .NET-based concurrent library implementation, and support for a number of languages, including C# and Visual Basic .NET, JScript, and IronPython.
CoroWare designs and develops unmanned software for robotics applications. As one of the first third-party companies to support the development of the Robotics Studio, it recently launched PlusPack, a collection of applications services, tools, assets, and utilities that complement the platform.
Many designers use National Instruments’ LabVIEW and graphical system design to create autonomous vehicles, humanoid robots, fixed-base industrial robots, and personal/home service robots around the world. The company’s suite of development tools is a common staple for developers of robotic platforms.
Several companies are trying to help robot manufacturers and end users with a universal robotic platform associated with state-of-theart software modules (for speech recognition, face detection, and other functions) to simplify the development of robotic products. France’s Gostai is working on various applications for home robotics, such as home surveillance, elderly care, and entertainment, with the focus of making these applications robot-independent.
“A key issue on the market is that robots are incompatible, which makes application development for various hardware parts a difficult task,” says Gostai CEO Jean-Christophe Baillie. “We solve this problem with our Urbi technology, a kind of Java for Robots that is compatible with 15 different robots on market. Urbi is in many ways similar to Microsoft’s Robotics Studio, but with key differences. These include greater simplicity and more flexibility. It also works with Linux as well as Windows.”
Yet achieving some sort of standardization for robotics will be challenging. Some experts believe the most useful technologies will become the de facto standard. They point out that it’s difficult to develop standards without knowing the target applications, the hardware requirements, the computational requirements, and other factors. Down the road, they see top-down design approaches, bottom-up implementations, and iterative refinements to more closely suit end applications.
We may be a long way off in developing personal robots that can match human emotional complexity and live side-by-side with human beings, but this isn’t so distant. In Japan, where robots are taken for granted, robots serve as receptionists, make sushi, plant rice, tend paddies, serve tea, greet company guests, and chatter away at public displays. The level of R&D for robotics in Japan far surpasses the rest of the world.
Nonetheless, there’s a lot of optimism worldwide. Robosoft’s home-centric Robuter is based on the company’s robuBOX generic robotic middleware, which in turn is based on Microsoft’s Robotics Studio, allowing the robot to be customized to any task. Robuter combines the advantages of Internet and robotics technologies to aid people that are handicapped or have limited mobility. Robosoft foresees the commercialization of this system by 2011.
Reprints
