REDMOND, Wash., Dec. 12, 2006 — For hundreds of years people have dreamed of the day when autonomous, robot-like machines would perform tasks for us in the physical world. But while robots play a big role in industries such as automobile manufacturing, we still have a way to go before they catch up with their science-fiction counterparts in books like Isaac Asimov’s “I, Robot and movies such as “Star Wars.”
Many observers think that may happen soon. Recent technology advances are opening the door to a new generation of robotic devices that will have capabilities far beyond today’s robots. Microsoft Robotics Studio, a new software development kit, is expected to play an important role by providing robotic developers with a platform that makes it significantly easier to create robotic applications.
PressPass spoke with Tandy Trower, general manager of the Microsoft Robotics Group, about this new product and the changing robotics industry.
PressPass: What are you announcing today?
Tandy Trower, General Manager, Microsoft Robotics Group
Tandy Trower: We’re releasing a software development kit for the robotics community that can be used with a wide variety of hardware, by a wide audience of users, and that supports the development of a wide range of applications. Our goal is to provide an affordable, open platform that makes it easier for robot developers to integrate hardware and software into their designs.
The Microsoft Robotics Studio SDK includes three areas of software. The first is a runtime architecture that can be used for all types of robots, from simple educational robots to sophisticated industrial robots.
Number two is a set of tools that makes programming and debugging robotic applications easier. We take advantage of the programming tools in Microsoft Visual Studio, but it’s not limited to that – robot developers can use almost any language they are comfortable with, and we also provide a visual programming language that makes it incredibly easy for anyone to create a robotic application. In addition, Microsoft Robotics Studio provides a simulation tool that includes a physics engine so that developers can design and test their robotic applications in a three-dimensional virtual environment that can simulate forces like mass, gravity, and friction. It’s also a great classroom tool because it allows students to learn about robotics and program sophisticated robots even when they don’t have access to expensive hardware.
The third part of the package is over 30 tutorials, plus samples with source code that help users get started creating robotics applications.
PressPass: Why now? Why is now the right time for Microsoft to be releasing its first robotics product?
Trower: We started working on this product in late 2004. One thing that got us started was the realization that the robotics industry looks quite a bit like the PC industry did 30 years ago. Today you have very sophisticated industrial robots in factories doing things like building cars. You also have people using robots in education, we’re seeing more and more companies popping up that are trying to create commercial products with broader appeal, you’ve got entertainment robots, and you have a large and enthusiastic hobbyist community.
The parallels with the PC business in the late 1970s are quite striking. In those days, there were IBM mainframes, which are analogous to the big manufacturing robots today. At that time, a variety of companies were creating PCs – even video game companies like Coleco and Atari – and trying to define what the key applications would be that would help justify the hardware. And you had a very active enthusiast audience.
But the problem was that it was a very fragmented industry. In the late 1970s, you had all of these architectures based on different processors and operating software, so there was no consistency and the programming tools were very crude. That made it hard to write applications and typically required rewriting code to move applications from, say, an Apple machine to one from Atari or RadioShack or Commodore. That was the first time that Microsoft provided a common ground for an emerging industry. By developing a version of BASIC and porting it to almost every hardware platform on the market, Bill Gates and Paul Allen provided a common lingua franca across different kinds of hardware, which helped get the software industry up and running.
This same kind of fragmentation exists in the robotics industry. Robots use different operating systems, different hardware, and different processors. So programming these things is very hard, the toolset is limited, and the amount of reusability of code is very limited. When you create a robot application, you often must start from scratch.
PressPass: Given this fragmentation in the robotics field, where does Microsoft Robotics Studio fit in?
Trower: We’re hoping to provide a platform that smoothes out this fragmentation. We’ve created a consistent programming interface so that whether you are creating applications for an educational or industrial robot, it’s the same programming model. This means students who learn to program on a LEGO robot are mastering principles they need for more complex robots.
It’s an open platform in that any hardware or software vendor can easily adapt their products to it, giving them a larger potential audience. And by bridging across the diversity of hardware, it enables a high degree of reusability of code.
So in the same way that the PC community was able to come together around first BASIC, DOS, and then Microsoft Windows, Microsoft Robotics Studio has the potential to be the catalyst that can enable the robotics community to come together. And this has always been Microsoft’s strength – providing that platform that enables new markets to emerge and grow.
PressPass: What was the process that led Microsoft to begin developing a robotics product?
Trower: About three years ago, we started hearing from leaders in the robotics community who were asking Microsoft to consider applying its assets to robotics. Whether they were focused on education, academia, or commercial development, they all came to us with a similar message – that something significant was happening in this space and that there was an important role for Microsoft to play.
Then, in 2004, Bill Gates visited a number of universities to speak about the importance of computer science. After each speech he toured the school’s computer science department, and at almost every school he was shown at least one project that involved robotics. At the time, I was on Bill’s strategic staff, and when he got back he asked me to take a closer look at what was going on. So I spent the next five months talking with professors in universities, teachers using robots in classroom settings, and people on the commercial and industrial side. I also connected with people in Europe and Asia, where significant investments in robotics are being made.
At that time, Microsoft Research (MSR) already had a number of projects underway in the robotics field, both within MSR and in partnership with universities in the U.S. and Europe. One example was a program partly funded by MSR at Cornell University to develop autonomous robots that can program themselves and evolve in response to their physical environment. And at the General Robotics, Automation, Sensing and Perception (GRASP) Lab at the University of Pennsylvania, MSR was supporting research that focused on developing applications for multi-robot teams. These are just two nice examples I could cite.
Talking to all these people was a great experience, partly because the work they are doing is so interesting and partly because when I asked them if they thought Microsoft could help the robotics community, instead of saying, “No, stay out of our business,” virtually everyone said that it would be tremendous if Microsoft could do something that made it easier to build robots so that more people could participate in the industry. It was at that point that we started looking into developing what became Microsoft Robotics Studio.
PressPass: Robots are a big part of popular culture for decades and they’ve been used in industrial settings for many years. Why has it taken so long for robots to become mainstream?
Trower: One reason is that the hardware has been expensive. Sensors like laser range finders for measuring distance can cost about US$10,000. By the time you include sensors, actuators and processors, it can put the cost of building a robot beyond most people’s means.
Probably even more important is the thing I talked about earlier: that fact that software development has just been too hard. So it becomes a world of haves and have-nots. The haves are like the people in the early PC days who had the expertise to build the hardware and the software. Then there is a vast majority of people who would like to use this technology but can’t because of the technical challenges.
PressPass: What has changed that makes robotics look so much more promising now?
Trower: The cost of processing has gone way down and so has the cost of storage. Researchers are using these improvements to solve some of the very hard problems. For example, it is very hard to do vision recognition or navigation with 8-bit processors, but much more feasible with 32-bit processors. With all the computing resources we have now, people are starting to solve these challenges.
At the same time, hardware is getting cheaper, too. For example, that laser range finder that cost $10,000 a few years ago now costs $2,000. In addition, wireless networking is much more common, making it possible to distribute processing.
PressPass: As these hard problems are solved, where do you see the field of robotics headed?
Trower: I believe that robots represent the next era of PC technology. Today, is your PC a robot? You probably would say no. But PCs are starting to include cameras and microphones, primarily for person-to-person communication. Once they are on board, there’s no reason they can’t be used by other pieces of technology. If cameras can do visual recognition – so your PC knows when someone walks in the room whether it’s you or not and changes its behavior accordingly – is it a robot, then? Maybe.
It just seems that it’s a natural progression to the point where our PCs start to interact with you in a much richer way. It doesn’t mean that the PC goes away or that every PC becomes a robot. It’s more subtle than that – it means the technology moves on in wonderful new forms.
In addition, the pervasiveness of wireless networking means that the mobile part of the robot can be a peripheral device that carries a very limited amount of inexpensive processing horsepower but is networked back to the processor on your PC. So the mobile device – the device we think of as the robot – can be very cheap and can be replaced very easily. And if it collects data, the data is stored on the PC so it isn’t lost.
PressPass: What are some of the possible real-world applications that this mobile PC enables?
Trower: If we think of the robot as essentially a mobile PC peripheral that allows the PC to get up off the desktop and interact out in the world, it opens up a very powerful notion of remote presence. We see it already in the form of military robots that can be used to find roadside bombs, for example, and domestic robots that vacuum the floor. In the not-too-distant future, we’ll see probably see robots that play an important role in providing assistance for us as we age and become more dependent. By helping us keep in contact with our families and health care providers, they can enable us to better maintain our independence. Already robots are being used by doctors to check in remotely with their patients. Robots will also be able to handle hazardous materials and dangerous situations in industrial settings or even military scenarios.
PressPass: How big is the market? What is the potential?
Trower: According to the latest statistics from the International Federation of Robotics, about 1 million personal robots will be sold around the world this year. Compared to the hundreds of millions of PCs and mobile devices that will be sold, that’s a very small number. But in South Korea, where the government is really focused on the robotics development, the plan is to put a robot in every home in that country by 2013. The Japanese Robot Association predicts that by 2025, the personal robot industry will be worth more than $50 billion a year worldwide.
PressPass: Is Microsoft Robotics Studio, and the work that your team is doing, the main focus of the robotics work at Microsoft?
Trower: There’s a lot more going on. For example, as part of the ongoing collaboration between Microsoft Research and leading research universities, we’re funding the Institute for Personal Robots in Education hosted at Georgia Tech with Bryn Mawr College. Under a three-year, $1 million partnership, the institute will be developing a new robotics-based curriculum to stimulate student interest in learning computer science.
This program grew out of the work MSR was doing with computer science professors on embedded systems research projects and computer science education enhancement. Many of these professors reported that the compelling nature of robotics makes it a great tool for inspiring students to study computer science. Microsoft Research is supporting this research both as a field of study in itself and as a valuable addition to teaching computer science.
Microsoft is also supporting Carnegie Mellon University’s new Center for Innovative Robotics and has provided seed funding to more than a dozen university robotics research groups around the world in recent years including MIT, the University of Virginia, Tecnológico de Monterrey in Mexico, Pontificia Universidad Católica de Chile, University of Potsdam in Germany, University of Pisa in Italy and many others.
PressPass: In some ways, robots don’t necessarily seem like they fit in with Microsoft’s traditional core businesses. How does robotics fit in with Microsoft’s larger mission?
Trower: There’s very little doubt that robots will be powered by PC technology – even on the industrial side, we’re seeing more and more robots that use PC-based technology. As this industry moves forward it will it will take advantage of technology that already exists on the PC.
It’s the next frontier, and in the same way that we applied our software assets to help create a platform for the PC market, we can help to provide the platform that will unleash the creativity of the robotics community and enable the robotics industry to create the applications that will fuel its growth. So in many ways this work reflects what we have done in the past and is a very natural evolution of the things we do today.