Automated Becomes Autonomous: Smart Helpers for Industry 4.0

It is not only in the automotive industry that autonomous driving represents the Holy Grail of progress. The idea of having transportation vehicles drive without human operators has been explored and successfully implemented in warehouses and production environments since as early as the 1950s . Navigation at that time still depended on permanently installed fixed guide rails, but the advanced sensor technology now used in grid, laser, and contour navigation has significantly increased the degree of freedom enjoyed by the trucks. Nevertheless, there is a lack of critical features that would allow the trucks to truly move around freely and detect obstacles without the intervention of humans and, above all, without running the risk of collisions. The KION Group is currently significantly involved in several research initiatives to determine precisely how this can be achieved in both smart factories and warehouses of the future.

Semi-autonomous vehicles are already being deployed in many production environments. While it is standard for a forklift to recognize a static obstacle and brake, fully autonomous driving is still a vision of the future. It is in this context that the project “IMOCO under Industry4.E” (short for “Intelligent Motion Control”) was recently launched, and in which the KION subsidiary STILL is significantly involved: Using AI and advanced sensors and communications, it aims to enable intelligent trucks to move around production halls or warehouses completely autonomously, avoiding obstacles and intelligently navigating their way around them. The IMOCO project is supported by the German Federal Ministry of Education and Research as well as the European Union with the research incubator ECSEL (Electronic Components and Systems for European Leadership). Many project partners are involved including Fraunhofer IML.

The technical challenges are immense: To begin with, the truck must be able to perceive its surroundings using a wide range of sensors (cameras, laser scanners, and radar). This not only includes spatial objects such as shelves, but also signs, markings, and displays. The second stage involves the truck understanding what it perceives and learning to classify objects: Are they static (such as shelves), movable (such as pallets), or even dynamic (such as other trucks and people)? Self-localization skills (where am I?) are enhanced and an understanding of assigned tasks (what should I do?) is added. In the final stage, the truck is expected to perform its tasks autonomously: autonomous navigation to the destination, load detection and handling, drive through the warehouse including automated decision-making, such as avoiding obstacles and finding a logical place to put down a pallet. These are typical processes that can be handled by autonomous transport fleets in the future.

The project is scheduled for completion in the fourth quarter of 2024. The ultimate objective is to significantly expand the autonomous capabilities of the trucks and open up a broader range of applications in human-machine production environments.

A project addressing similar issues was launched back in September 2018. “Deep PTL” is the name of the research collaboration (a combination of “deep learning” and the acronym for production, transportation, and logistics) that aims to give self-driving trucks the ability to “see” more, resulting in increased autonomy in the warehouse.

Together with scientists from the University of Freiburg and sensor specialists from Sick AG, KION Group engineers are working on the project to find artificial intelligence models capable of recognizing objects with as few parameters as possible. The reason is that it is important for a forklift not only to brake in front of obstacles, but also to be able to distinguish whether the obstacle is a human being or just plastic sheeting flapping in the wind, for example. This is where Deep Learning comes into play, i.e., specific optimization methods with which neural networks can mathematically solve the kinds of problem that humans solve intuitively. “We use Deep Learning methods for this purpose—they are extremely powerful in recognizing objects and the reason why there is currently so much hype about the potential of artificial intelligence in many different industries,” explains Patrick Erbts, Research Manager Technology & Innovation at KION. There are, however, a number of challenges when it comes to implementation. Firstly, the processing capacity of the computer chips on board the trucks is limited, restricting learning. Secondly, the visibility conditions in warehouse and production environments are often not optimal, resulting in erroneous object recognition. The basic idea of the project is to bring together the data from different sensors. To do this, a mobile platform was specially developed and equipped with the latest sensor technology. This is needed for training and enhancing the neural network. The project is due to be completed in 2022.

How can autonomous technologies be combined even more effectively with human capabilities? And what safety features are needed to allow trucks to operate together with people in the same area? These questions were the focus of the QBIIK, which was successfully completed in 2020. In addition to the KION subsidiary STILL, project partners included Audi Sport GmbH and the Karlsruhe Institute of Technology (KIT).

The research task was to investigate and develop the autonomous production supply of a “supermarket” using a mobile robot. In the automotive industry, a supermarket refers to the place where assemblers are provided with the components they need for production—in other words, a warehouse close to the production line. As part of QBIIK , an autonomous order picking system with learning capabilities, the STILL iGo neo, was equipped with a gripping robot to autonomously drive to the supermarket and collect the ordered goods. A human-machine interface made it possible, if necessary, to remotely request human assistance via a virtual reality user interface if, for example, gripping problems occurred. When this happened, the human took control of the robot and carried out the recognition and gripping processes. This enabled the robot system to learn from humans how to deal with new work processes and perform new work steps independently. The prototype successfully passed the test run at Audi’s production warehouse environment.

All these projects represent important milestones on the road to Intralogistics 4.0. The KION Group and its brands are contributing all their expertise and experience to these, and other key research projects, with the aim of gaining an early understanding of future industry requirements. Only then can we provide our customers with the best solutions for their future success.