Linde MH and the TH Aschaffenburg present the results of the research project "KAnIS - Cooperative Autonomous Intralogistics Systems" © Linde Material Handling

From next year, the vehicles are to be further developed and tested in order to take on four specific material flow tasks at the plant in the future: the transport of pallet cages and the transport of pallets with batteries, as well as the transport of vehicle frames and driver protection roofs, which are brought from the pre-assembly lines to the main assembly lines on special load carriers. The first two applications are purely outdoor operations, while in the other two the forklifts travel both in and between the halls. Gradients of 8 percent have to be overcome, and other AGVs and manually operated vehicles also operate in the halls.

To ensure that the four KAnIS forklift trucks can pick up pallets, pallet cages and metal frames with their forks even if they are not exactly aligned with the floor, the vehicles have a mobile camera mounted between the forks. It measures the pockets of the load carrier so that the forks can be positioned correctly over the sideshift. The vehicle frame, the battery door and the counterweight were also adapted in terms of design. "Our aim was to integrate the safety scanner, cameras and sensors into the vehicle contour as far as possible so that the dimensions remain as close as possible to the standard forklift truck," says Hanke.

In the halls, the vehicles localize themselves using laser scanners, and outside using differential GPS (Global Positioning System), a method for increasing the accuracy of GPS, as well as additional local sensors at the transition from the indoor to the outdoor area. Unlike the manual forklifts, the automated vehicles always travel backwards on the defined routes so that the load cannot slip off the forks in the event of an emergency stop.

Real-time communication with forklifts and infrastructure

A particular focus of the research project was on the automated forklift trucks' perception of their surroundings in order to ensure their reliable interaction with other road users. To this end, the vehicles are equipped with additional 3D scanners and HD cameras in addition to the sensors of the personal protection system. The camera data forms the basis for detecting and classifying objects using AI algorithms and then localizing them in order to adjust the truck's driving speed and slow it down to a standstill.

Thanks to real-time communication between the vehicles used and stationary 3D cameras, the trucks do not have to rely solely on their own sensors outside. © Linde Material Handling

But that was not all. A further question dealt with critical situations that arise when road users are in concealed areas that cannot be seen by the truck's sensors and are approaching the truck's path. This is where the cooperation of the forklift trucks comes into play, because if another forklift truck is in the vicinity, it could provide the relevant information. However, the prerequisite is real-time transmission of the perception data. In order to achieve these low latencies, a private 5G network was set up at the Aschaffenburg plant. The perception data is transmitted from the forklifts to an edge server, which uses the locally detected objects to create a global list of all detected objects and sends this back to the forklifts.

The test was conducted with a crash test dummy that suddenly emerges from behind a wall and runs into the path of travel. Without cooperative behavior, the automated forklift truck cannot stop in time and crashes into the dummy. If it receives real-time information from a nearby forklift truck, the vehicle is able to recognize the dangerous situation in advance and can brake in time. However, as it cannot always be assumed that a second forklift truck is nearby, eight stationary 3D laser scanners have been installed at junctions and gateways along the routes that the KAnIS forklift trucks will travel in future. The local object lists of the stationary laser scanners are also merged on the edge server and the information is made available to all vehicles.

"Fast wireless networks are the prerequisite for autonomous forklifts to be able to act cooperatively in outdoor areas and react to unforeseen traffic situations in real time," emphasized Prof. Dr. Klaus Zindler, Vice President Research and Transfer at the TH , at the event. "Our goal is to develop general standards and algorithms using AI methods that can then be flexibly applied to different vehicles or applications and continue to learn."

Cleaning system for sensors, battery charging by robot

A further work package examined how the optical sensors close to the ground can be cleaned if they have been soiled by splashing water in the rain or wet road surfaces. This is because if reliable object detection is no longer possible, the personnel protection system automatically brings the truck to a safe state and it stops. To prevent this, the project team developed a cleaning system that uses compressed air to blow away the droplets of dirty water that collect on the laser scanners.

Another project team investigated possible solutions for the autonomous charging of forklift truck batteries. The result was in favor of an AI-based robot that connects the charging plug to the charging socket of the forklift truck. The rear of the truck was modified accordingly and an automatically driven charging flap was added to protect the charging socket from dirt and splash water.