Soon, people and autonomous transport robots (ATR) will work together. The robots will be given the same opportunities as humans to feel and see the full spectra and thus be able to take greater responsibility for what they do.
This can be made possible as researchers at Chalmers and Volvo develop systems of artificial intelligence with cameras, in Volvo's factories, where information from the cameras is the robots' main source of information. Researchers from Chalmers have been working together with Volvo Group Truck Operations (GTO) for a long time and now the vision is starting to become reality.
Flexible and advanced technology
Admittedly, self-driving trucks are no new news in the automotive industry, but so far, they generally use simple technology – such as following a magnetic loop or a strip of tape on the floor. The control and monitoring of the factory's material flows that Volvo Trucks now outlines will be significantly more flexible and advanced.
“Fundamentally, it is the need for more environmentally friendly transport that drives the development”, says Knut Åkesson, Professor of Automation at the Department of Electrical Engineering at Chalmers University of Technology and the one who leads the work from Chalmers' side.
Today, it is not a given which technology will replace the diesel engine, and within the Volvo Group, work is being done in parallel to develop several alternatives. Manufacturing trucks with several different powertrains in the same factory means that the amount of components required to manufacture them increases greatly.This, in turn, means that the need to improve the efficiency of internal transport increases.
“Here we see that a type of self-driving transport robot can give us the flexibility to use the floor space in the factory in the best way and at the same time also give us the efficiency we must have in our internal transports”, says Per-Lage Götvall, Senior Research Leader at Volvo GTO and the one who leads the work within the Volvo Group.
For several years, Knut Åkesson and his team have worked closely with the Volvo GTO to develop the GPSS (Generic Photogrammetry based Sensor System), image analysis using machine learning. A collaboration that has been very valuable and fun.
If the work to develop the GPSS system and its transport robots is as successful as Volvo believes, it will mean a huge change in the factories' internal transport flows, from the traditional "fishbone layout" to more "Just-in-time". The system is then expected to be implemented in most of Volvo's more than 30 factories worldwide.
Efficiency and safety
But it is not a trivial task to create a traffic system where, in the end, hundreds of self-driving transport robots will be able to function efficiently, safely, and friction-free in an environment where people and materials are constantly in motion.
"Compared to normal road traffic, the factory floor still constitutes a limited and controlled environment and just like in traffic, where common rules make everything work, similar principles can also work in factories. This means that all transport robots can be controlled and coordinated by an "omniscient" system”, Knut Åkesson explains.
Planning all material transport in time and space, based on what production requires, is an important part of the system. In addition, it will prevent transport robots from driving in each other's way.
“But then there are all the dynamic obstacles, in the form of people and traditional forklifts, which the system cannot control", Knut Åkesson adds.
Traffic monitoring systems and machine learning
This is where the GPSS technology comes in. The idea is that the cameras in the ceiling should detect and classify all static and moving obstacles on the factory floor and thus guide the transport robots to their destination without collisions and accidents. The cameras become the joint "eyes" of the robot fleet.
"If a human does something unexpected, such as suddenly going to the side, the traffic monitoring system must immediately detect this and ensure that the transport robot that is next to it acts so that an accident is avoided”, says Knut Åkesson.
Machine learning
What makes a traffic monitoring system like this possible today is the rapid development of computers' ability to interpret images in recent years. This is done with the help of machine learning and so-called artificial neural networks – that is, self-learning algorithms that are gradually trained to cope with their tasks. A bit like the way the human brain works.
"We are also trying to teach the system to predict how people will move in the next few seconds. This means that we can optimize in real time how the robot trucks should move in each unique situation that arises, all based on the information from the cameras", says Knut Åkesson.
Improved working environment and sustainability
But how will it feel to be in a work environment where humans and robots work so closely together, and is the fact that Volvo uses cameras something that could become a problem?
"Since the images and videos generated by the cameras will only be analysed by computers and not at all seen by a human eye, we see the cameras more as "optical sensors" rather than cameras, but of course, basically they are digital cameras we use”, says Per-Lage Götvall.
"In the discussions we have had with union representatives at Volvo, everyone has been positive that Volvo is driving the development. In the project, we are also happy that the relevant personnel categories can be involved in the design of GPSS, and those who are involved are all very positive about what we are doing”, Per-Lage Götvall continues.
Knut adds that the ultimate purpose of "giving eyes to the factory" is to be able to control it better and thus also to make it cooperate better between the people and robots who work there.
"One could see it as a matter of improved working environment and thus also as a question of sustainability”, says Knut Åkesson.
"Optimisation of logistics is also important from a sustainability perspective. The aim is to consume as few resources as possible. For example, the transport robots must drive the shortest possible route and thus minimise energy consumption” Knut Åkesson continues.
Test environment
Much of the algorithm development that the researchers work with takes place in a part of Volvo Trucks' factory in Tuve on Hisingen called the Pilot plant - a kind of experimental factory in the factory where new products and manufacturing methods are tested and developed.
There, the researchers can test that the transport robots really behave in the way that the increasingly refined algorithms aim for in a realistic environment.
Knut Åkesson believes that it will take a few years before Volvo Trucks is ready to introduce the new logistics system into production.
"The idea is that it should be able to be introduced in small steps. First in the pilot plant, then in a smaller part of the factory and finally in the entire truck factory”, says Knut Åkesson.
Per-Lage Götvall adds:
"Since it is a completely new technology and a new way of "socialising" with autonomous robots, the "precautionary principle" requires that we take small steps and continuously evaluate what is happening. As part of this, we have industrial doctoral student Atieh Hanna. Her area of research is to see how we can develop safe but efficient systems and methods for workgroups where autonomous robots are included as a natural part".
About the collaboration
The collaboration between Chalmers and Volvo Trucks has been going on during various project phases since autumn 2019 and has involved a handful of senior researchers, three doctoral students and a large number of students at master’s and bachelor's level. For Chalmers, the current project period runs until summer 2023, but a continuation cannot be ruled out.
Facts
- The collaboration between Chalmers and Volvo AB regarding the truck factory of the future is part of the long-term partnership that was renewed for another three years at the latest in February 2022.
- The project now continues at Chalmers under the name AIMCor (AI-enhanced Mobile Manipulation Robot for Core Industrial Applications).
Initiator and main manager at Volvo Trucks is Per-Lage Götvall, Senior Research Leader at the company. - Researchers and engineers at Volvo who are involved in the development of GPSS are Associate Professor Kristoffer Bengtsson, Adamya Shukla, industrial doctoral students Erik Brorsson and Atieh Hanna.
- Chalmers researchers who are or have been involved include, in addition to Knut Åkesson, Yiannis Karayiannidis and Emmanuel Dean, as well as the three doctoral students Sabino Francesco Roselli, Ze Zhang, Endre Eres and Martin Dahl, who recently publicly defended his doctoral thesis.
Text: Sandra Tavakoli
For more information, contact:
Knut Åkesson, Professor of Automation at the Department of Electrical Engineering, Chalmers University of Technology
knut.akesson@chalmers.se
Per-Lage Götvall, Senior Research Leader, Volvo Trucks
per-lage.gotvall@volvo.com