‘Robots don’t need to be powerful, fast or precise any more’: Suryansh Chandra

“They cannot compete with a human in terms of intelligence. They cannot compete with a human in terms of perception. But a robot is very good at doing a narrow set of things,” says Suryansh Chandra. “If you want to pick something up and put it somewhere, and you need to do that task 2,000 times a day, let a robot do it.”

Chandra is co-founder and co-CEO of Automata, a London-based SME that makes desktop robot arms for industrial manufacture, test, machine tending and inspection. His company’s literature claims that its EVA robot is the ‘first robot designed with small-batch production in mind’ and that shop-floor workers can learn to program the machine in a quarter of an hour. With a price tag of just under £5,000, EVA is also bringing robots within reach of tasks that had once been thought of as too small to be automated. No wonder that the 37-year-old architect thinks EVA is poised to fulfil his dream of ‘democratising’ the world of industrial desktop robots. Recent changes in market dynamics and manufacturing labour trends also mean the capabilities required of today’s robot arms are changing. “They don’t need to be extremely powerful, fast or precise anymore. They just need to compete with a human.”

There’s nothing particularly new in robot arms. They can be traced back to the automotive assembly lines of the 1960s. “These robots were basically meant to do the heavy lifting,” says Chandra, who describes the early incarnations as “a mechanically operated crane-like device that will pick up heavy things at Point A and move them to Point B simply because it is too intensive for people to do”. Back in the heyday of post-Industrial Revolution factory labour, serialisation of production in high volume “had just begun. Factory jobs were a new thing and people were excited about it because there were lots of jobs. Healthcare and factory floor welfare was starting to become a thing. And that’s when people started to look at these devices as something that could augment labour.”

Since the early days of car assembly, other industries have also increased their production volumes, while robots have “got better and living standards have improved, and opportunities in white-collar industries have boomed massively”. Part of what this industrially led societal change means is that “more and more people have been driven away from labour-intensive jobs and into comfortable plush office-chair alternatives where you sit in front of a computer and process data”. Meanwhile, in the past four decades, the rise of eastern economies – particularly China – where labour costs have been low historically, has created the opportunity to capitalise on rising living standards of the west. What this means is that “there are very few blue-collar jobs in the UK now, and those that do exist are in constant competition with China and, increasingly, Vietnam, Malaysia and Indonesia.”

Chandra says “the pressure to automate in order to offer the same products and services at a competitive price has reached a critical point for the UK and European manufacturer. What is likely to happen over the next few years, as we get to higher levels of automation, and better efficiency and throughput, is that manufacturing will make a comeback. Yet the jobs won’t be the same. There will be more high-skilled, value-added jobs in which people will be no longer required to do what they have been historically bad at: repetitive, menial labour, where we have seen error-rates going up and throughput rates going down.”

This is particularly the case on unsupervised night-shifts because “people are simply not meant to deposit glue repetitively hundreds of times in exactly the same spot for eight hours. That’s just mind-numbing.” These are the jobs, says Chandra, where it makes most sense to automate. Chandra cites further examples of repetitive product testing, such as pushing buttons on a car dashboard or opening and closing the hinge of a laptop, swiping entry security fobs on a reader millions of times over three months.

“Those are jobs that fundamentally people are not designed to do. They’re not good at doing them and they don’t enjoy doing them. Yet they need to be done and the default is to send them to China or Indonesia, where that will happen,” which is where, says Chandra, as competitive entities, British manufacturers lose out. “But customers want these jobs done in the UK because they want them done closer to their customer. The solution is to bring back automation to do some of these jobs. These are the underlying drivers for people to buy robots.”

Robots are only one part of the automation landscape, says Chandra, but it is the part that Automata is plugging into, which requires a decent understanding of the economic backdrop leading up to today’s market conditions. This is because conventionally, “the entry price-points of complex robots has been high. This means that there are only a certain number of tasks where it makes economic sense to automate, and most of them are high-volume, high-throughput tasks, which is why automotive and electronic production lines do so well in automation.”

Chandra explains the UK, with a 17 per cent share, has one of the largest aerospace industries in the world. “But because this industry is lower volume and lower throughput, suddenly automation starts to make less sense. With a smaller-scale automation solution taking, say, three to six months to set up, if your batch will only take two months to produce and then you’re done, it might not make sense to build that line.”

This sort of task – where current automation options “don’t make much sense” – is precisely the ground that Automata is pitching for with its EVA desktop robot arm. In order to create sense, something in the equation has to change, which is where capital cost of hardware and time-to-install can make a difference, which Chandra sees as equal in terms of cost. While there was once a stratum of potential jobs considered too small and time-​consuming to be worth setting up as a robotics application, this can now be addressed with cheaper robots. With EVA coming in at a shade under £5,000, low-volume, low-throughput jobs such as machine tending, product testing, inspection and sorting become an economically viable proposition.

‘Robots don’t need to be extremely powerful, fast or precise any more’

The availability of lower-cost hardware introduces a subtle but critical shift in the way smaller manufacturers can shape their attitude towards installing robots. While they were once brought in to do the sort of repetitive work that was difficult for humans to sustain over long periods of time, or work that was physically too demanding, “what we are seeing is that machines have become cheaper and more capable, while labour has become more expensive and there are fewer and fewer people that want to do blue-collar jobs. Robots are starting to infiltrate that area and the question becomes ‘what can we automate?’ That’s why iPhones and laptops are assembled by robots. It’s not because there are heavy parts or that people can’t do it. It’s because it’s cheaper for a robot to do it.” When you get to that point you bring in robots. In that respect, says Chandra, not much has changed apart from that the cost of human labour is getting more expensive.

Chandra admits there’s a ‘big leap’ from being an architect who understands the evolution of a post-Industrial Revolution robotic landscape to becoming an entrepreneur confronting the reality of actually designing and building a commercially available robot arm. Having grown up in India where he studied architecture, he moved to London 12 years ago at the age of 25 to become an architect, “because I was very good at being creative, but pretty bad at mathematics”, and where he also became a British citizen. His initial post-graduate work was at a “pretty avant-garde” school of architecture, the Architectural Association School of Architecture in London, probably the most prestigious and competitive such organisation in the world and renowned for radical thinking. After graduating, he spent six years at Zaha Hadid Architects where he rose through the ranks, from working in the parametric modelling workshops to lead architect at around the time of the 2009 economic crisis in the UK that the “architecture business somehow survived”.

Chandra describes Zaha Hadid’s organisation as fundamentally challenging conventional thinking: “Why should anything be straight? Or vertical? Or horizontal?” It was in the research group that Chandra teamed up with colleague Mostafa ElSayed, who would go on to be the other co-founder of Automata. “Our job was to understand how to build some of these exceptionally radical shapes.” Yet it wasn’t simply the case of an architect coming to the research group with a complex design and asking the tech-guys to ‘make it so’. 

“We had an input. We could say: ‘this is not how you design stuff, because this is completely disconnected with how you build things in reality. We were the bridge between architecture and engineering in that we’d work closely with machines and materials and putting that knowledge back into the design process.” Chandra describes his research group as an ‘autocorrect function’.  “If you think of something physically or gravitationally impossible to create, we’d go back with more feasible alternatives in a short collaborative feedback loop.”

Chandra goes on to describe the ‘curvy roof’ of the Magazine restaurant in Hyde Park, inside which is a series of curved metal pipes. “To understand how to get the three-dimensional complex curves required by the architect, we had to understand the machines that bent the pipes and work with those machines to produce the desired curve.” The nearby Math Gallery in the Science Museum is another example of how the end result evolves out of an understanding of the engineering material.

Yet it wasn’t until Chandra worked on an installation called ‘Arum’ for the Venice Biennale architectural show in 2012 that the idea of creating his own robot really took root. The sculpture is made up of 500 metal panels, all of which had to be bent into a precise shape. Inspired by technology in the automotive industry, “we had designed this sculpture on the assumption that we’d be able to use industrial robots to fold these panels.” Yet what Chandra quickly realised was that while it took 30 minutes to fold each panel using a robot, the same could be achieved manually in a sixth of the time, a scenario that hardly fits into any efficiency algorithm. He was using robots “because we could. This was the pivotal moment. Why was using robots so hard when they were supposed to be the promised land of automation? We initially thought that maybe this was because we didn’t understand robots well enough, so we decided to get one to play about with. It turns out that when you buy a robot without any specific project in mind, and by default, a very small budget, that doesn’t work. With the naivety of an architect, we thought, if we can make buildings, we can also build robots. How hard can it be, right? It’s just a bunch of motors.”

That was what Chandra calls the “ignorance with which we started a curiosity project” that was originally simply to find a way to use robots in the design process to better effect. “We bought some off-the-shelf electronics and mechanical parts. Then we tried to put it together. That’s how our first robot was born, and it turns out that it’s very difficult to build a robot. It took many, many iterations and generations.”

After a year and “six or seven prototypes”, Chandra became convinced his idea “had legs”. Importantly so did his colleague Mostafa ElSayed, and between them, having decided that “building robots is more fun than architecture”, they simultaneously turned their backs on opportunities to spend four years doing work on their doctorates in order to build robots.

It was a case of “we could go and do our PhDs for four years and then come out of that and decide what to do. Or we could do it now. That’s how we took the leap of faith, quit our jobs and set out to commercialise what we were doing. We thought it would take six months. It had to because we only had enough money to do this for that amount of time. However, within four months ABB had expressed an interest in what we were doing and, along with a few other investors, we were able to take on a team of engineers that could actually solve the problems we were starting to face. And that is the genesis of Automata.”