The next generation of robots

We live in an era of digitalisation which has created a virtual world of data where we can store and retrieve information that can be shared on the internet every day all over the world. So humans now have not only the physical world but also an information world on the cloud where we can interact with data. And with the creation of the internet of things, the data-cloud is getting direct access to our real physical world and can perform functions in it.
Professor Siegwart

Robots are the most evident example of this new step of connecting the data world with the physical world. Thanks to the world of data, some software and IT companies have grown up to be the most valuable companies in the world. Now we see first examples of new robotics companies conquering the world market, such as DJI, a Chinese company which has grown in the last 12 years to a unicorn in the drone market.

Of course, robots are already performing a lot of functions such as building cars. Without them, humans would still have to build the large number of cars, which I think would be nearly impossible. Imagine that people had to spray-paint the cars – it would be extremely unhealthy. But while we can be happy to have these robots, they are only able to perform their job in very structured environment, protected behind fences. 

The next generation of robots which is slowly developing are not behind fences, but are learning to do things like walk, climb stairs and do some jobs that humans do. However today they are still very slow, inflexible and very limited in the tasks they can fulfil, even if some overselling YouTube videos makes us believe something different. 

Humans can not only see and understand their surroundings, but also feel and interact with it and each other – and this is what the next generation robots must learn. For example, they need to have not just position control, but force control for tactile and soft interaction with objects and humans. Other key challenges for expanding the use of robots are:

  • Dealing with uncertain and partially available information
  • Seeing, feeling and understanding their environment
  • Intuitive human-machine interfaces
  • Ability to learn and adapt every day

Next I would like to set out a few examples, mainly of what we are doing in our research lab, which give you a feeling about what is happening in research for the next generation of robots. The first group of examples are about the design of robots for different tasks. The second group will present perception and navigation. 

The first example is the result of a project by ETH students in the last year of their bachelor degrees. The students said that people who are handicapped should be able to do whatever other people can, and designed a wheelchair which can climb stairs and thus move to most places where healthy people can go. The project continued during their master’s, when they got some funding and they are now starting to sell their wheelchair Scewo.

The ANYbotics mobile robot

The second example is the walking quadruped AnyMal, which also started with student and PhD research projects. Industrial robots are typically fixed to the ground, and can only do things where their arms can reach. For the future, it would be nice to have robots which can go everywhere and anywhere. To do so they need legs. Humans and animals have muscles to generate forces and tendons that absorb shocks and can store energy. Similarly AnyMal has a motor to generate forces, and a spring that absorbs shocks and stores energy. Thanks to these bio-inspired actuators, AnyMal can move in very challenging terrains and has thus the potential to conduct complex inspection and intervention tasks. 

Helicopter-type drones have very limited flight duration, whereas fixed-wing airplanes need specific infrastructure for take-off and landing. Another group of ETH students decided to combine the advantages of helicopter-type drones and fixed wing flight. The resulting novel vertical take-off and landing Wingtra was developed into a product that is now sold all over the world. 

Flight duration with fixed-wing drones was pushed even further by the autonomous systems lab. Equipped with solar panels on the wings, flying robots can now stay in the air for days and weeks, and can be used for tasks such as search and rescue at sea – replacing much more expensive human-piloted airplanes. 

It will take much longer to deliver people’s ambitions than many think

If such robots are to carry out autonomous tasks, however, they need sensors to see and understand the environment. Laser distance measurement is one option that is expensive, though it will find a use in cars when the price falls. But cameras are evolving fast and are a much cheaper option, and techniques are now evolving that can create 3D vision which is essential for flying robots and are also proving useful for augmented reality applications. 

Meanwhile flying robots which can see are being used to inspect industrial sites, offshore platforms and wind turbines which are difficult for humans to access. And combinations of flying robots and ground robots – with different sensors – will soon be able to survey fields from the air to get information about whether they need more water or fertiliser that can be remedied by the ground robots – which may be able to detect and destroy weeds.

Humans can not only see and understand their surroundings, but also feel and interact with it and each other – and this is what the next generation robots must learn

Robots will also be used in construction for various purposes. These include building very odd-shaped walls which they can do better than humans, cleaning buildings and painting them. But there are some harder jobs that will take longer to hand over to robots, such as housework which includes complicated tasks involving interactions with children and the complexities of cleaning a kitchen. 

In conclusion, robotics is booming and everybody is talking about robots. But it will take much longer to deliver people’s ambitions than many think. It’s not the same as implementing new business ideas like Uber or Airbnb which are less complex and can scale faster. Building the next generation of robots will be difficult: it needs position mechanics combined with AI, as well as top talent and research. But I hope that my examples have demonstrated the potential for robotics and the progress that has already been achieved.

This article is part of the 'Lakestar Briefing', a periodical publication about Lakestar's portfolio companies and our network of inspiring minds we like to work with. If you wish to subscribe please click here to fill out your name and email and we will be in touch.