We are still waiting for robots that make our everyday life easier by doing the dishes and cleaning our homes. Of course, we have machines such as Roombas that can vacuum, sort of, but why do we still have to bother doing all those boring things ourselves?
Why aren’t robots doing my laundry yet? Well... maybe they are. Using a very strict definition of robot, and an even stricter definition of doing laundry. Perhaps your dry cleaner uses some kind of automated device for handling large amounts of clothing. But that’s not what we mean with robots doing laundry, of course. So why is it that there is still no robot coming into my bedroom every morning, picking up my dirty clothes from the floor, and putting them in my washing machine? The simple answer is: doing laundry is just not that easy.
Everyday activities such as doing your laundry or doing the dishes are - while easy for most humans - actually quite complex. Think of all the steps you have to go through. First, you have to identify pieces of clothing on the floor (or your laundry basket, depending on personal preference), a non-trivial task for automated systems, drawing from expertise in the field of computer vision: is that your faux fur vest or your cat?
Next, you have to determine if it’s clean or dirty, an even less trivial task, even for myself. Now we need to pick up this piece of clothing to transport it to the washing machine, a mechanical task drawing from knowledge in the field of mechanical engineering. Even if we can combine all this knowledge (in the broadest sense of the word) in a robot, we would still only have a robot that can do laundry, leaving us with dozens of robots in our house performing various tasks. Unfortunately, my house is not large enough for all these robots, so I’d prefer a kind of robo-maid that can do all kinds of chores. And I think most people would.
I am not arguing that the field of robotics is not advanced enough: robots in some form are used in a wide array of fields, from car manufacturing to pharmaceutical research. In these settings they perform repetitive tasks often faster and more accurately than a human could. And it is exactly this nature of tasks that is the problem. While everyday tasks seem repetitive to us, they are in fact far too heterogeneous to automate easily. A lot of arguably mundane tasks require knowledge that we take for granted. How do different textile fabrics behave, and do some colors bleed more than others? Humans learn this by experience and observation of others, but in robot systems a lot of information is often hard-coded, making them less flexible by being unable to learn or update knowledge. So how can we make robots that are able to perform a wide variety of tasks, without giving them too much a priori knowledge?
Investigating issues like this is part of our work as a member of the RoboHow consortium, an EU-funded project aimed at creating an autonomous robot able to perform a wide variety of tasks. The role of cognitive scientists and psychologists is perhaps best illustrated by roboticist Rod Grupen: “At bottom, robotics is about us. It is the discipline of emulating our lives, of wondering how we work.” To make robots that perform like humans, it is necessary to look at how humans perform such tasks. And who are better equipped to answer that question than psychologists?