With the rapid growth in the robotics industry, humans and robots have been working side by side for years. So it’s not surprising that in the current situation caused by the COVID-19 pandemic, people have begun to ask if robots might be a viable solution to controlling the spread of the virus.
Social robots like Pepper or the Singapore-based robot dog could be part of the new force of frontliners in the struggle with COVID-19 and their existence has fueled a growing demand for robots, especially in this season. As an indicator, Hyundai Robotics, which was only officially launched in May of this year, has already received an order for 370 robots.
While it’s true that there are many robot use cases, in a range of industries — from manufacturing to agriculture — consumer robotics isn’t as simple as building and deploying. Sure, the robot vacuum may be a lifesaver in many homes, but when it comes to robots that are meant to be released into the open world (like delivery robots or the robot dog), there are a lot more variables to consider.
These variables make a world of a difference between robots used in smart homes versus those deployed on the streets. Home-use robotics have the benefits of a controlled environment, where they may have the time to “learn” about the space they inhabit.
“Mistakes” made by robots at home are also less costly compared to those made by robots out in the open. A vacuum robot missing a spot is trivial compared to a delivery robot falling into a manhole.
Although robotics has been explored for deliveries, a service that’s growing in importance in the wake of this pandemic, none of the existing robots are ready to be deployed at scale. Even the more adventurous players in the consumer robotics industry like Starship Technologies have only managed to deploy 12 robots in five places within the United States.
As with any other autonomous vehicle, these sorts of consumer robots don’t have adequate software to handle the complexity of all the uncertain situations that may arise when out in the “real world”. They have not reached the level of reliability required for them to be produced and deployed at scale.
Companies like Postmates have found a solution to this for now and it’s a solution that creates more job opportunities for humans as well — teleoperations. In this scenario, human employees are “fleet supervisors”, helping to guide the robots towards making the right choices or even using a hand-held remote controller to steer the robot when needed.
Of course, this involves having to develop complex systems for the teleoperator to utilise — either in the office or in their homes — that include setting up workstations, performing internet upgrades and rolling out standard operating procedures.
Why it’s not as simple as “plug and play”
For robots to even begin to learn how to move autonomously through the streets, they have to be fed a huge amount of quality training data.
You probably don’t remember the first time you were brought out of the house as a child, but imagine being a toddler having to navigate through the world for the first time. Even walking on uneven ground was a struggle. It was something we all had to learn.
Over the years, we learnt how to look left and right before crossing the street. We learnt that on wide streets, we have to use a pedestrian crossing. That it was probably a better idea not to run across a highway where vehicles are likely to be traveling at much higher speeds. We learnt how to zig zag in a crowd to avoid colliding with another pedestrian. And even after years of learning how to physically navigate the world, we still have accidents.
A robot meant to conduct itself autonomously on the streets would have to go through even more rigorous training, for every single city that it’s deployed in. Imagine the amount of data that would be required for training! And even then, would that data be able to account for every single possibility that may occur?
Not everyone’s ready for avant-garde innovation
When new tools are launched online, there’s always a chance that there are still bugs in the system. It depends on its early adopters to understand that these bugs exist, and to respond to them with constructive criticism. Not everyone is cut out to be an early adopter.
With the consumer robotics industry however, many expect these robots to work perfectly right off the bat. Besides that, there are so many other non-tech considerations to take into account when releasing these sorts of robots into the world.
After a couple of self-driving cars were involved in accidents, people began to ask questions related to law and policies that at the time had not been updated to take cases like these into account.
Deploying consumer robots at scale in the “open world” will take time. Even if a robot is deployed successfully in one city, it doesn’t mean it would immediately be successful in another.
Each place has its nuances and robots would have to go through custom training for each. One thing that might make this part easier is for robotics companies to find training data that’s accurately labeled by people who are familiar with a particular region.
That said, consumer robotics is still a difficult industry to be in as these robotics companies have to manage the cost of training robots and maintaining the hardware, while balancing it out with the cost of sales and marketing.
When we think about consumer robotics, we think about what an exciting industry this could be and all the benefits that having robots in our day-to-day could bring.
Ask anyone if they’d want robots to deliver their supermarket purchases or drive them from one place to another, there’s a high chance the answer is “yes”. But there are so many other factors to think about.
It may be a difficult industry, but as with anything that used to “be impossible”, it’s just a matter of time.