Robotic revolution: why now? A hardware perspective
This trend will have profound long-term economical and societal consequences. Our report “New Robotics and Drones 2018-2038: Technologies, Forecasts, Players” forecasts that the market for robots and drones will grow from $66 billion in 2018 to more than $210 billion in 2028. Interestingly, much of these new robotics will come from new (vs old) robots and drones.
We foresee that these new robots and drones will come to represent 46% of the market in 2028, up from almost zero.
Now the question is why now? This is a fair question because people have always sought to automate more tasks but they have been limited by the economic and technical limitations of automation process. Therefore, the answer must lie in the fact that advances in hardware and software components have rendered them significantly better and cheaper over time, turning once fantastical robotic ideas into a real commercial opportunity.
In this article, we draw from New Robotics and Drones 2018-2038: Technologies, Forecasts, Players to consider developments in price-performance ratio of hardware components to make up a new robotics: transistors, memory, sensors, energy storage, and electric motors. We will leave the discussion of advances in software aspects including deep learning to another article.
The anatomy of a new robot
The schematic below shows the hardware anatomy of a new robot. Here, we have assumed a mobile robot. Its intelligence unit consists of computing elements (e.g., transistors and memory), as well as sensing ones (e.g., GPS, camera, IMS, ultrasound, IR, LIDAR, wheel encoder, etc). Its mobility functions include electric motors and energy storage. At a higher level, there is a supporting infrastructure such as the internet, cloud computing and storage and GPS network.
In the remainder of this article we will consider how all of these hardware components have evolved over time. This drastic improvement in price-performance has been the wealth creation engine of our time, boosting global productivity. It is a trend that we all intuitively appreciate, as we have all lived through this rapid transformation over the past decades.
Exponential trends have brought us here
Consider the chart below. First focus on the ‘computing’ line. It shows the cost of computing as a function of time from around 1940 to about now, divided by the cost at the initial point in the series. The unit of measurement is the cost of a device capable of calculating 1600 million instruction per second. We can see that there has been a phenomenal 12 orders of magnitude fall in the cost of computing. This is just incredible.
This has of course been accompanied with increase in performance. Indeed, Moore’s Law is well known. Interesting, at 1971, Intel 4004 chip packed only 2250 transistors in a 12mm2 chip. Fast forward to 2017. The Apple A11 packs around 4.3 billion transistors in a 89mm2 chip and the Centriq2400 packs around 18 billion in a 398mm2 chip. This too is phenomenal and has of course entailed an incremental year-on-year shrinkage of device size. This exponential growth has helped make new robotics a commercial possibility.
Now focus on all the memory related lines in the chart. The unit of cost here is $/Mbit. Here too there has been a phenomenal cost reduction in hard drives, RAMs and solid-state memories. In fact, the rate of change in this industry has been so rapid that it has turned it into an academic case study for all those wishing to study the phenomena of disruption.
Now let us consider cameras (CMOS image sensors). Here too the cost has exponentially fallen. Other technologies showing rapid cost falls include MEMS and other sensors. We speculate the LIDARs will also start to rapidly fall in price as the transition towards improved mechanical control and solid-state versions take place. It is therefore evident that the cost of sensing (data acquisition) and computing (data processing) has dramatically fallen over the year. The size and power consumption of these devices have also dramatically fallen. All these exponential trends have operated for multiple decades, bringing us to a point today that suddenly complex new robots (collaborative, mobile and intelligent) are becoming commercially viable.
Now let us consider two aspects of two hardware components that make up the mobility unit for mobile robots: energy storage and electric vehicles. First, consider energy storage. Here, the growth rates are limited by chemistry and are not exponential. But that is not to say they are unsubstantial. The chart shows the Li ion batteries’ energy packing capability has grown by 3 and 2.4 times in two decades from 1991 to 2011 in Wh/Kg and Wh/L terms, respectively. This improvement has continued since and is only likely to accelerate thanks to further work on advanced and post Li ion batteries. Furthermore, the cost has fallen significantly too. This is set to accelerate thanks to investments to scale up production to meet the rising demand from the electric vehicle industry.
Next consider the electric motor, the unsung hero of mobile robotics. The chart shows the reduction in mass of a 5HP (3.7kW) electric motor with time relative to its mass in 1910. We can see that the electric motors now outputs the same power using a much smaller size and weight. They are also more efficient and less expensive. These trends explain why commercialising electrically powered mobile robots and drones in now possible and popular.
About the author:
Dr Khasha Ghaffarzadeh is Research Director at IDTechEx – www.IDTechEx.com/robotics