Wanted: Robot to load dishwasher. Must be careful

Chris Roberts, head of industrial robotics at Cambridge Consultants, provides an overview of the jobs robots could potentially do in future

Cambridge Consultants recently announced a new technology for automating one of the most unloved jobs in a commercial kitchen – loading the dishwasher.

This technology combines innovative robotics with cutting edge AI to achieve something hitherto impossible – detecting and handling a wide variety of dirty crockery haphazardly arranged on a tray.

However, the fundamental techniques used in this product can be applied to a wider range of situations and problems. There are many different tasks that it is only now possible to automate.

There is no single technology revolution that has enabled this, it is the combination of several disparate building blocks, that have each evolved to the point where new levels of automation are economically viable.

Things like machine vision, neural network algorithms, data storage and processing power have been around for a while, but have all got incrementally better, cheaper or more accessible; and that has enabled a new set of applications.

Not all tasks that could be automated with this technology should be. The key question is how to spot something that is both technically feasible and economically viable – yet hasn’t already been solved.

What other applications can use this technology or some variation of it?

The answer is those applications that combine three characteristics:

• a semi-structured task;
• a large potential install base; and
• a difficulty in recruiting labour.

We see applications in industries as diverse as agritech, construction, logistics and domestic automation that an all make use of these technologies to enable new products.

‘Semi-structured’ tasks

This technology – the combination of neural networks, advanced sensing and clever electromechanical design – enables automation in places where the task was previously too varied to be automated.

But there are still limits to what can be done. Although the system is designed to handle some variation in the input, it is still doing essentially the same action with the items it handles.

So, while I can imagine this technology being applied in a domestic setting to, say, load and unload a kitchen dishwasher or washing machine, I can’t see it being used to put the clothes or plates away again.

There are simply too many intermediate tasks – crossing the kitchen, avoiding colliding with people, opening and closing drawers – to make this easy to automate.

However, there are industries where semi-structured tasks are ripe for automation, for example logistics and warehousing.

It is not unusual to find automation of the movement of bulk goods around a warehouse, but in general, the fine movement of individual items to and from a tote or carton is still done by hand, because of the variety of items the system would need to handle.

This technology could be customised to help fit that need, and it’s a great candidate because although the items are varied, the task itself is always the same.

Similarly, the construction industry could benefit from more automation.

Again, using a single robot to do all the tasks needed to construct a building is probably still science fiction, but this technology can enable automation of individual tasks, such as fetching and carrying on building sites, or transferring markings from plans to the real world so that builders know where to dig or cut.

Large install base

It is important to realise that the hardware and software used in this design is customised to the specific task of loading a dishwasher.

While the techniques and know-how can transfer from one product to the next, unlike with traditional robot arms, you can’t manufacture a “general purpose” system like this that can accomplish many things in different contexts.

This means that a new product needs to be developed for each new task to be accomplished, and that costs money. In turn, this means that industries that can best benefit from this technology are ones where the cost of development can be amortised over a large installed base of products.

Typically, this means that the company that develops the technology needs to be able to sell it to a wide variety of users, rather than having a single user own the technology for competitive advantage.

For example, there are a lot of applications in the Agritech business – picking and sorting individual apples, strawberries or other high-value crops – and while individual farmers may be the ones struggling with labour shortages and wanting to automate, they generally don’t have the scale where automation makes economic sense.

Labour shortage

This is the trickiest aspect of finding the right application. There’s a lot of media coverage around robots and AI especially “stealing jobs” – but this is not what we see happening among our clients.

Whatever you do to spread the cost, it remains true that automation, and especially innovative automation, is expensive to implement, at least in a short timescale.

The most successful applications are ones where political pressure, socio-economic changes or changes in expectations mean that it is hard to find people willing to do that job.

In the end, people automate not to save money, but because the job is too boring, dangerous or difficult to recruit for.

As part of the project that developed the dishwashing robot, we explored other aspects of catering that could usefully have robots added to them.

One idea that came up was to automate the process of cooking – but we rejected this. Not only because teaching a robot to taste food is really very hard, but also because that is an easier job to recruit for.

When you’re unemployed, almost any job is better than no job, but people dream of being a chef – no one wants to be a dishwasher loader or wants their children to do that job.

Future

So, will we see a whole range of tasks being automated?

I think the answer will be yes, where it is technically feasible, economically viable and socially acceptable. 

There are lots of exciting developments in the field of automation that could reap benefits in many different industries, and for the innovative companies that are prepared to invest in development, the rewards will be huge.