The robot revolution has been expected for decades, but still has not arrived. Every year, automakers and their suppliers purchase thousands of robots, a number that still amounts to vastly fewer than what has been predicted. Robots simply are not used as widely as they could be, due to several barriers that have persisted far longer than expected. These barriers to adoption can ultimately become missed opportunities to increase the country’s manufacturing capacity and improve its supply chains.
With the pandemic exacerbating shortages in the availability of goods and shining a light on how easily critical supply chains can be disrupted, there has never been a better time to focus on technologies and solutions that can help strengthen the country’s domestic capabilities. Recent executive orders show that this will be a focus of the Biden administration, so now is the time to act.
Barriers to Adoption
There are, generally speaking, three barriers to wider adoption of robotic automation in the automotive industry. First and foremost, the cost remains too high. Cost includes not just the purchase of the robots themselves, but also several other significant related costs. Typical robotic workcells require extensive engineering coordination. Simply deploying one or more robots in a work cell is an expensive task. Then, once the work cell has been engineered, the robots must be programmed. Robot programming is the bane of many engineers—and of the manufacturers that have to pay them to do so. The high cost of deploying robotics cannot be amortized by anyone producing small product volumes or low-value products and has even proven difficult for suppliers to afford to roll out at scale. Recent findings by the International Federation of Robotics show that, on average 75 percent of the lifetime operating costs for industrial robots come from programming. With every task change, the application must be re-programmed.
A second barrier to adoption is inflexibility. Once you have engineered your workcell and programmed your robots, you are set—just as long as you never change what you are doing in any way. Any change—be it a new robot or a variation in the manufacturing process—requires reprogramming (at a minimum), and quite likely begins a process of re-engineering and revalidating the work cell to change robot placement. This inflexibility makes robotics unfeasible for anyone producing a variety of products with small volumes.
The third barrier is the poor marginal benefits of adding robots to work cells. Programming a single robot is a challenge; programming multiple robots to work in a shared space, without collisions, is an extraordinarily difficult task that consumes months of engineering time. In fact, multi-robot programming is so difficult that engineers make simplifications to shorten programming time at the expense of greatly reduced efficiency.
The most common simplification is the use of “interference zones,” which prohibit more than one robot from being in any space that could be reached by more than one robot, even though, in practice, multiple robots could frequently share such spaces without collision. Due to the use of interference zones, it is not uncommon to find that a work cell with four robots achieves performance that is less than two times that of one robot. This low efficiency of multi-robot work cells drives down the use of robots, even in companies that can afford them.
If we are going to unlock the potential of robotics, we need to lower the barriers to adoption. We want everyone in the automotive sector to be able to make greater use of robotics. Because the cost of robots and engineering time is unlikely to decrease, the key levers in our control are the time and expense of deployment and programming.
To achieve our goals, we need robots that are adaptable to their current situation, enabling them to operate in relatively unstructured workcells. Adaptability, in turn, depends on two capabilities: reliable vision and fast motion planning:
- Reliable vision—enables robots to observe their surroundings and react to them. Democratization requires that vision be not just reliable, but also relatively inexpensive. Fortunately, many good, low-cost options exist for vision, and they continue to get better and cheaper.
- Motion planning—the process of computing and coordinating how to get a robot from its current pose to its desired pose without collision—must be fast enough to adapt to dynamic environments, particularly for environments that include people. Motion planning performance has historically not sufficed for general purpose applications, leading to robots that either react slowly, or operate without vision at all (e.g., in a work cell that is highly engineered, such that the robots never need to react). Recent advances in academia and industry, however, suggest that the motion planning bottleneck will soon fall.
Democratization also necessitates robots that can be programmed quickly while still achieving high performance. Currently, we can achieve either high performance or tolerable (but still long) programming times, and industry has consistently chosen the latter. The only solution for reducing programming time and improving performance is greater automation of the programming itself.
Asking engineers to reason about the trajectories of multiple arms while choreographing all their movements is not a path to success. The automotive industry needs new software tools that remove most or all that effort from the engineers, so they can simply specify the tasks they want the robots to perform, and the software generates the robot programs. These advances would improve any robot programmer’s productivity in much the same way that general (non-robot) programming in a high-level language like Java or Python enables far greater programmer productivity than programming in assembly language. The key in both cases is providing a higher level of abstraction to the programmer and using software tools to automatically convert this easier-to-write code into what is needed at the lower level.
The robotics industry has not nearly reached its potential in the automotive manufacturing sector and supply chain, but with innovations in just a couple of key areas far greater value could be realized. Democratizing robotics would enable many automotive companies to introduce additional robots and profit greatly from them. Furthermore, we could add to the value proposition by reducing costs and improving the marginal benefit of adding robots to work cells. The end result would be a large increase in domestic manufacturing capacity and an opportunity to create more reliable domestic supply chains, helping to better prepare the country for any future pandemic-level disruptions in materials or labor.