Democratization of Automation
This post originally appeared in Xconomy.
The Next Generation of Industrial Robotics
In the coming years, artificial intelligence will have many transformative impacts, but the one perhaps most resembling science fiction will be the rise of autonomous physical systems: real-world systems that can make decisions for themselves. Of these autonomous systems, self-driving cars have captured the most attention. But the field of manufacturing, which represents one-sixth of the global economy, stands to be impacted as much by AI-powered automation as the cars it produces on its assembly lines.
The Evolved Robot
Robots – machines capable of carrying out a series of actions automatically – have been in use in manufacturing for decades. However, these robots have not had much in the way of intelligence or autonomy. Traditional industrial robots have tended to be extremely expensive, dangerous to humans (requiring the robots to be isolated in cages), difficult to program, and narrow in their application. For these reasons, robots have been utilized by only a small percentage of manufacturers, and for a small portion of their production processes. Today, thanks to a number of technology enablers, a new generation of industrial robots is emerging that will reshape the way factory floors operate.
In contrast to the current generation of industrial robots, the next generation has the following characteristics:
- Inexpensive – typically tens of thousands of dollars, versus hundreds of thousands or millions
- Collaborative – able to work safely alongside humans
- Easy to program – can be programmed to perform a task in 2 – 4 hours, versus a traditional set-up time of over a month
- Flexible – can be trained to perform a variety of tasks
- Adaptable – can handle unstructured environments and variation in how tasks are presented, and even self-optimize over time
Two points here deserve expanding upon. The first is that, although the next generation of robots has a lower sticker price than prior generations, it is these robots’ other attributes that produce truly game-changing cost savings. “In a typical industrial arm application, up to three-quarters of the cost goes to a systems integrator for the installation, application development, programming, and maintenance and only one quarter goes to the cost of the hardware,” according to Tom Ryden, Executive Director of the MassRobotics innovation hub in Boston. Modern collaborative robots’ flexibility and ease of set-up have the potential to reduce or eliminate the need for these costly systems integrators.
The second point is that the adaptability of these next-generation robots is fundamental to achieving greater levels of automation. Today’s robots do not handle variation well, but for some tasks it is either expensive and time-consuming or simply not possible to engineer out all the variation in a production process. New robots will increasingly be able to handle complex unstructured tasks like dealing with unsorted parts or deformable objects. This will dramatically expand the set of tasks a robot is capable of doing.
The Big Bang
What is enabling this new generation of robots? What makes them so much cheaper and smarter? There are a few factors that make possible today what was not possible a few years ago:
- Cost of components – a depth sensor, which robots rely on to perceive their surroundings, cost tens of thousands or hundreds of thousands of dollars just a few years ago, but now costs on the order of $100
- Standardization – the open source Robot Operating System (ROS) provides a common hardware abstraction with thousands of user-contributed libraries, while any new robotic arm produced today has an API that lets external software control it, allowing roboticists to focus on pushing the boundaries of what a robot can do rather than reinventing the wheel
- Artificial intelligence – advances in AI, particularly deep learning, allow robots to understand their surroundings and make decisions based on their sensor-driven perception of those surroundings
- Advances in silicon – GPUs and more specialized chipsets (for example dedicated processors for robotic motion planning) enable massive computations to be done on the robot itself, without incurring the latency or bandwidth overhead of sending data to the cloud
As this next generation of robotics emerges – cheaper, more user-friendly, and much more intelligent – we believe it will have some significant consequences for the manufacturing landscape because it will not be incremental but truly transformational. Here’s why:
- Robotics will be accessible to a much larger set of manufacturers. The price point and ease of programming will allow small and mid-size manufacturers to leverage robotics for the first time. This democratization of automation could portend an increase in the global competitiveness of small and mid-size producers. Typically when an expensive technology once restricted to a small audience becomes inexpensive and accessible to a large audience, it creates the potential for transformative impact.
- Robots will shorten the innovation cycle. Being able to quickly train a robot on a new production process, rather than set up a whole new production plant overseas, can have a tremendous impact on the speed with which new products can be brought to market. The next generation of robots will enable production processes to be reconfigured much more rapidly than in the past. As part of this trend, a number of companies are mounting robotic arms to autonomous mobile platforms that enable automatic reconfiguration of work cells as demand changes. “In a manufacturing environment these AMRs – autonomous mobile robots – dramatically increase flexibility,” according to MassRobotics’ Ryden.
- Robots will address the manufacturing labor shortage. Skilled manufacturing labor is retiring and the replacement pipeline is sparse due to a shortage of STEM graduates and negative perceptions of manufacturing work among millennials. The global workforce is also shrinking. China will lose 90 million workers by 2040 due to demographic shifts. Japan and Korea will lose 10% – 15% of their workforces during that time. The working-age population is also declining in the U.S. and Europe. For this reason, increasing levels of automation are in some sense imperative if the manufacturing industry is to maintain or grow its output.
There are a number of exciting industrial robotics startups advancing the state of the art in this space, tackling key elements of the value chain from enabling technology (at both the hardware and software layers) to complete robotics systems. The efforts of these startups, along with innovation occurring in research labs and at larger companies, will lead to dramatic improvements in the capabilities of robots for manufacturing. The next few years will be exciting ones for the industry as these next-generation technologies become more widely adopted.
More broadly, we believe autonomous physical systems will play an increasingly important role in our lives, and will be an area ripe for innovation over the next decade or more. The interaction of AI with the physical world will impact many industries, such as agriculture, energy, and transportation & warehousing. Robots and cyber-physical systems will seem increasingly less exotic and ever more commonplace as they become a critical enabler for businesses of all sizes.