TU/e wants to make packing robots in distribution centres more efficient by not necessarily letting them avoid collisions. By making robots take into account the impact of a collision with an object, they should be able to pick up objects faster, among other things.
Robots are a key hub for packing goods in distribution centres, fulfilment centres and other logistics locations. However, the intelligence of such robots is limited. For instance, they cannot predict the impact of any collision they cause. In practice, this means that robot arms come to a halt just before they collide with an object or or the surface on which an object is placed. This measure prevents the robot from colliding too hard with the object, causing damage.
However, this measure limits their actions. Robots, for example, are not able to place goods on a conveyor belt or in a crate in a fast and fluid motion, for example.
Putting collisions to good use
Current robots thus deliberately try to avoid collisions rather than putting them to good use. For example, a robot arm can pick up an object faster if the object can be picked up without having to come to a standstill first. Thus, according to the researchers, robots, while precise and reliable, are poor at performing dynamic tasks where they come into contact with objects.
"Humans are much better at predicting the effect of a collision, and therefore a lot faster. They are able to place, or even throw, an item onto a conveyor belt in one quick, fluid motion, without damaging the item, the conveyor belt or themselves. The same problem also occurs when placing items in a crate or box. In this, humans are not only faster than robots, they are also much better at filling the crate or box as efficiently and quickly as possible," TU/e writes on its website.
Poorly paid, repetitive and unhealthy
However, human staff are relatively difficult to recruit. For example, packing products is relatively poorly paid, repetitive and unhealthy work, for which there is little enthusiasm. TU/e is therefore joining forces with two other universities, a knowledge institute and four companies. The parties will conduct research into 'impact-aware robot manipulation'. The project should lead to smarter logistics robots, capable of relieving human workers.
Within the project, TU/e researchers are collaborating with colleagues from the Technical University of Munich (TUM), the École polytechnique fédérale de Lausanne (EPFL), and the French research institute CRNS. The parties are collaborating with Dutch companies Vanderlande and Smart Robotics, Germany's Franka Emika and Sweden's Algoryx.
Making robots impact-aware
Together, they are studying how to make logistics robots impact-aware. In time, this should lead to new robots that can reliably determine the impact of a collision with their surroundings or other objects. This will enable them to 'exploit' this collision to pick up or put down objects faster.
To make this possible, the researchers want:
- deploying artificial intelligence to teach robots to make movements without causing damage in the process;
- develop advanced sensors that allow robots to map the impact of a collision.
Building into existing robots
To simplify the embrace of their technology, the researchers want to make it integrable into existing robots. Alessandro Saccon, working at the Faculty of Mechanical Engineering and research leader on behalf of TU/e, expects that this will allow them to work 10 per cent faster.
"Impact-aware robots could play a big role in logistics," says Saccon. "E-commerce has boomed, and in a tight labour market it is becoming increasingly difficult to find workers. Moreover, work in distribution centres is often mind-numbing, and bad for health. Impact-aware robots can then offer a solution, freeing workers' hands for other work."
Impact Aware Manipulation
The parties are joining forces within the project I.AM, which stands for Impact Aware Manipulation. The programme will receive a 4.4 million euro grant from the European Union. TU/e is coordinating the four-year project.
Further (technical) information about the research can be found in the paper‘Predicting the post-impact velocity of a robotic arm’, published by Saccon in collaboration with Vincent Padois of the French National Institute for Research in Computer Science and Control and Ilias Aouaj of TU/e.
Author: Wouter Hoeffnagel
