Monday, December 5, 2011

Robotics is a Highly Fragmented Industry With Few Common Standards or Platforms

by Frank Tobe, Publisher, The Robot Report
with Manoj Sahi, roboticist and educator, India

 
Imagine being present at the birth of a new industry. It is an industry based on groundbreaking new technologies, wherein a handful of well-established corporations sell highly specialized devices for business use, and a fast-growing number of start-up companies produce innovative toys, gadgets for hobbyists and other interesting niche products. But it is also a highly fragmented industry with few common standards or platforms. Projects are complex, progress is slow, and practical applications are relatively rare. In fact, for all the excitement and promise, no one can say with any certainty when -- or even if -- this industry will achieve critical mass. If it does, though, it may well change the world.
Bill Gates, Scientific American Reports
2008 Special Edition on Robotics
Bill Gates clearly defined what I see to be prevalent in robotics today -- that there really are two different worlds operating, often together, more often not: 
  • Industrial robotics -- wherein a handful of well-established corporations sell highly specialized devices for business use. KUKA, FANUC, ABB  and Yaskawa Motoman are amongst the mostly European and Japanese vendors providing handling, welding, cleanroom and assembling robots to the automobile, electronics and a few other industries. [According to the International Federation of Robotics in their annual World Industrial Robotics 2011 report, 118,337 robot units were sold in 2010 bringing the worldwide count of industrial robots in place and working to 1,035,000.]

  • Service robotics -- faster-growing, with start-up companies and research labs producing innovative toys, gadgets for hobbyists and niche products for healthcare, dairy and agriculture, home and yard, space, defense, security and surveillance, and scientific research. [According to the International Federation of Robotics in their annual World Service Robotics 2011 report, 2,203,241 total service-sector robots were sold in 2010 composed of 13,741 field or high-end service robot units (of which 45% were for defense applications), about 1.4 million vacuum and floor cleaners, 36,500 lawn-mowing robots and 753,000 entertainment robots.]
As Mr. Gates said, robotics is a highly fragmented industry with few common standards or platforms even though there are well-financed pushes toward standard connectors, open-source software, modularity and shared libraries. Recently there's been a clammer for (and against) a universal open-source robotic operating system. The proponents main argument is that such a system is beneficial particularly for rapid prototyping in the development of new systems while almost all of the industrial robot manufacturers argue that their robots are performing mission critical tasks which require secure and fail-safe operating systems - which they have developed and optimized over the past 50 years.

The diversity of the arguments are, to some extent, merited based on the unique needs of the various end users:
  • Industrial, commercial, healthcare and defense users do need security, reliability and efficiency and robot vendors have a vested interest in protecting their value-added software and hardware solutions.
  • Academics and emerging businesses need openness and sharing so that re-invention isn't so prevalent as it has been in the past, and a full range of features and simulation capabilities are available so that far-ranging teams can collaborate on research projects using the same software. 

Industrial robot control systems are complex in part because of the need to integrate the robot into the end user's manufacturing system, partly because of the many algorithms which need to be processed to make the robot and all of its moving parts work, and because, up until recently, it was necessary to distribute the computing workload of the various activities of the robot, but also because control systems haven't kept up with the availability of faster and more capable CPUs, connectors and communication methods. All of these are reasons leading to robot manufacturers becoming more OEM-like as most of them aren't too capable in developing engineering solutions for new applications (thus their dependence on system integrators).

Service robotics, on the other hand, are more mobile and self-contained, carrying with them (or communicating via wifi) their control system functions. They tend to be either mobile or portable and either loosely connected to the end-user's internal system or operate either autonomously or by modified remote control. Consequently, software like ROS brings lots of new capabilities which the original robot manufacturers are not able to do. Service robotic providers and research labs and facilities see the value and enjoy the benefits of shared open source development - they see it as good and useful for the overall robotics community -- yet most of the start-up companies lock down their finished products so that end users are assured that their products are safe, secure and reliable and the resulting software is reduced to enable the product to fully function but not be overloaded with unused functionality.

Adding to this fragmentation and complexity are a few macro-level near-term trends:
  • Commoditization of hardware, hardware modules, and software.
    • Markus Waibel, senior researcher at Zurich's ETH IDSC and participant in the EU's RoboEarth library-sharing project, feels that there is a good chance that we will see a split into companies specializing in robotics software vs. robotics hardware in the near future. This is evidenced by the current trend toward modularization and commoditization of robotic hardware. This happened in the early PC industry and there are obvious benefits. "Actually, I would go a step further and argue that this split has already happened in robotics research, with most research platforms at least supporting [Willow Garage's] ROS," said Waibel. 
    • Tom Wagner, CTO of iRobot, says: "Open interfaces; not open source. That way 3rd party equipment and payloads can work harmoniously with iRobot and other robot manufacturer's products while still providing a safe, secure and reliable product to the end user." 
    • Many needed solutions have already been solved yet are re-invented regularly within academia, partly because of cost, partly to aid in the learning process and sometimes out of just plain whimsy. 
    • Forums for bringing together scientists, engineers, users and executives, such as the recent RoboBusiness 2011 event in Boston, attempt to thwart reinvention by being small and personal so that people have the time to explain their problems and explore various solutions. At that Boston forum, by having the time to meet and explore, one Boston agricultural startup was able to solve a vexing navigation problem by purchasing a solution module from another startup that had already solved that exact problem.
    • Industrial robot manufacturers are beginning to shift into the service sector. At a recent conference in San Francisco, KUKA, a large industrial robot manufacturer, announced that they were proceeding along a path outlined in an internal roadmap to enter the service robotics business. Although no other industrial vendor has indicated similar intentions, the progression to enlarge their businesses by adapting their robots and robotic solutions into the service sector appears to be inevitable.
      • Key points in KUKA's roadmap are enabling mobility with SLAM navigation, making their robots lighter in weight, sensitive to their surroundings so that they are safe for nearby workers, and providing motion simulator trainability for future applications.
    • Bosch also has a roadmap for developing robotic consumer devices that work within Bosch's existing product family. Cost, safety and convenience are their primary concerns. A laundry app is a perfect matchup of technology and Bosch's product line thus, at IROS, they were all over the Berkeley team that trained a PR2 robot to fold laundry.
  • SMEs (small and medium sized enterprises) are seen as a major near-term market for robot expansion. This market needs easily trainable low-cost co-robots that are flexible and safe enough to handle multiple small-volume tasks alongside humans. This is the bread and butter application for SMEs. Incidentally, this is another example of robot manufacturers becoming commodity robot and module providers - in this case providing a large quantity of low-cost co-robots and letting the SME's provide their own customization and integration.
    • Big consumer product companies such as Toyota, Samsung and others, have announced their intentions to enter the medical robotics and consumer healthcare marketplace in 2013 and 2014. This may have as much positive impact on the service robotics sector as Foxconn is having in the industrial market (see below). 
    • Foxconn entering the robotics manufacturing business with their own manufacturing and assembly factories as clients.
      • They've announced a planned deployment of 1 million robots within 3-4 years.
      • Little is known of the details of the types, quantities or activities of the robots nor of the 500,000 human laborers that will be displaced.
    • The Cloud is being introduced into robotics as it is in other areas of business - a way to offload processing and data to a place better equipped to handle the high number of CPU cycles needed to process vision, recognition, communication and navigation, most often simultaneously.
    It is into this fray -- this two-world ascendant industry of promise, fragmentation and complexity - that the question of whether a universal robotic control system is needed or not has been asked and not clearly answered.

    There appears to be an obvious and logical answer:
    • Robot manufacturers will soon enable open-source systems such as Willow Garage's ROS to connect and front-end their proprietary internal operating system. Yaskawa Motoman has already done this.
    • They will also enable an as-yet undeveloped universal interface so that using specialized devices and modules from various 3rd party vendors can occur.
    • There won't be a commercial version of ROS nor a universal robotic operating system.
    • There will, however, within academia and the emerging service robotics sector, be a greater dependence and use of ROS because, as Markus Waibel says, "There are obvious economies of scale benefits - just have a look at the ROS contributors list and it is very clear that even now no single robotics manufacturer can match this brainpower."