黑料福利网

Lazy robotics. That鈥檚 how Ren茅 van de Molengraft cheekily describes the focus of his research group at 黑料福利网鈥檚 Department of Mechanical Engineering. 鈥淲e design robots that only do what is necessary to complete their task, thereby consuming as little energy and resources as possible. No one needs an autonomous robot that can both efficiently stack heavy boxes in a warehouse and pour a cup of coffee without spilling.鈥�

The biggest challenge in robotics is making robots autonomous in a human living and working environment that inherently is full of variations. 鈥淎utonomous robots must be able to think for themselves and automatically carry out a dedicated task 鈥� while interacting with the real world. To enable them to handle new situations on their own without being instructed by some human operator, they need to understand their environment,鈥� explains Ren茅 van de Molengraft, Chair of the Robotics research group. 

Robots need to learn things that we, as humans, don鈥檛 have to think twice about. How does something feel? What do sounds tell you? And even: how does something smell?  鈥淲e learn things by experiencing them, but how do you translate that knowledge and those skills to a language robots understand? They need to be able to perceive through multiple senses, detecting and interpreting qualities like shape, texture, and temperature.鈥�

Stripping down to the core

The research group focusses on task-oriented robots. 鈥淚t makes no sense to develop a robot that can do everything a human can do. You don鈥檛 need that in any setting.鈥� The researchers therefore start by analyzing the task at hand, and the possible variations therein.

鈥淎nd then, the most important thing for us as researchers is to determine what knowledge and skills are and aren鈥檛 important,鈥� Van de Molengraft states. 鈥淢ost robots need to carry their own energy in the form of battery packs. Anything that drains that battery without contributing to the required result, should thus be left out.鈥�

That is easier said than done. 鈥淥ne approach is to have the robot learn from videos. So, if you are developing a laundry-folding robot, you might feed it with hundreds of videos of different people folding laundry and have it determine the best strategies under different circumstances. Personally, I do not believe in that approach, simply because we do not have the required data to train any robot for any task.鈥�

Understand the world

His group therefore takes another approach: 鈥淲e build model and knowledge-based systems that can reason on their own about changes and variations, and decide what, under the given circumstances, would be the best strategy to complete the task at hand. To achieve this, we build so-called world models that combine data and physical models. The robot uses this world model to simulate possible scenarios and predict what will happen if it takes a certain action.鈥�

Football as a demonstrator

To demonstrate what these autonomous systems are capable of and to advance the field as a whole, Van de Molengraft has been active in the RoboCup competition for over two decades with the 黑料福利网 team Tech United he set up. 鈥楢ctually, it was the RoboCup competition that got me interested in robotics in the first place,鈥� he recollects. 鈥淥riginally, I was working on motion control systems, for example to position wafer stages in lithography machines down to the micrometer or even lower. Around 2005, I visited a RoboCup tournament, where Philips was competing with their team. I immediately thought that at 黑料福利网, we could do better than the other teams.鈥� Having won nine world titles over the past twenty years, he has certainly proven himself right. 

Where the general public might regard a robot football competition to be a gimmick for engineers, it has a serious undertone, Van de Molengraft emphasizes. 鈥淲e use the annual competition as a platform for validating our research, and for exchanging the latest technological developments amongst the robotics community.  Learning autonomous systems to play football poses a multitude of challenges upon the technology. On the pitch, there鈥檚 constant, unpredictable interaction 鈥� not only with the ball, but also with an opposing player. At any given time, as a robot, you鈥檙e actively being challenged. If we can make systems that can handle those adversarial conditions, many other tasks in less dynamic and more quiet conditions become a lot simpler.鈥�

Solving labor shortages

Current and expected labor shortages in various sectors are a major driver for the development of autonomous robots, he states. 鈥淭here are many sectors where we are facing labor shortages. Think of horticulture, logistics, and construction. But also in healthcare. We have a severe and growing staff shortage there. It would be great if we could free up time for healthcare workers by having robots perform certain tasks.鈥�  

Even though there are still many challenges to be solved before autonomous robots will beat the human world champion footballers, as is the ultimate goal of the RoboCup competition, Van de Molengraft sees ample opportunities for the current generations of systems to be of use. 鈥淥ne thing current autonomous systems are well-suited for, is for performing inspection tasks. This could involve visual inspection, picking up objects, scanning, sensing, and so on. That labor-intensive work is currently still done by humans.鈥�

Collaborations

Since robotics is an inherent multidisciplinary field, the group has a lot of collaborations with others, both from inside and outside of university. 鈥淓ven though our group itself hosts multiple areas of expertise, ranging from manipulation and artificial skins to mobile robots and world modeling, we collaborate with several other 黑料福利网 departments for example with Electrical Engineering on sensing and hardware development, with Industrial Design on soft robotics, and with Built Environment on autonomous navigation in buildings.鈥� 

Real-world use cases provided by industry serve as an important source of inspiration for the research, the robotics professor explains. 鈥淲e build systems that should have a measurable impact in a specific domain. Knowledge of that domain comes from these companies. What鈥檚 more, through genuine collaboration, our vision also comes to life within the companies, that are slowly but surely starting to embrace our 鈥榣azy robotics鈥� approach.鈥� 

Brainport as robotics hub

With the ever increasing energy demands from our digitizing society comes the responsibility to develop efficient robotics, Van de Molengraft states. 鈥淚n addition to that, especially when robotics are entering the real-world involving people, ethics should be a crucial ingredient of every development. We can鈥檛 work with black-box AI solutions. As the Netherlands and Europe, we need to ensure that we develop a strong AI and robotics sector.  

Eindhoven can play a key role in this. This region has all of the required expertise and know-how. What we need now is coordination and entrepreneurship to make Eindhoven a global leader in robotics. I hope the Casimir Institute can play a role in this. For example, by building an ecosystem of scientists, robotics engineers, and domain experts, and demonstrating what this technology is already capable of. In the end, we need to do more with less. And the smarter we get, the lazier we can be.鈥�