Can AI manage automatic loading of timber trucks?

For several years, a research group at Umeå University has been training AI models to autonomously carry out different work elements in forestry. In a new Mistra Digital Forest project, is testing how close this technique is to being applied commercially in systems for automatic loading of timber trucks.

In 2024, the first AI-controlled self-driving forwarder was tested at Skogforsk's remote control lab in Jälla, Uppland. The tests showed that AI models trained in advance in a simulation, can be transferred to operate a real forestry machine. This is something that a new Mistra Digital Forest project is building on. In the project, AI models will be trained to autonomously load logs from piles at the roadside collection points, known as the landing. The hope is that if the solution is successful, it can be implemented in the near future by forest companies and equipment suppliers. Umeå University, Skogforsk and Södra are participating in the project.

Establishing partnerships with suppliers 

The project builds on basic research of recent years, where researchers in digital physics at Umeå University have developed AI models for controlling forwarders in grasping and loading multiple logs in the forest. The results have been promising but have only been tested in simulation so far. Now the researchers are going to test the developed method on a real machine.

– By demonstrating proof of concept and gaining insights into the potential and the limitations of the different solutions, we are able to understand what it takes to arrive at a commercial product in a relatively short time. The AI model and simulator will be available to stakeholders interested in further development, and an important part of the project is to establish collaborations with potential suppliers, says Martin Servin, leader of the project and of the research group in digital physics at Umeå University.

Safety is a powerful incentive

The strong demand for semi- and fully-autonomous solutions in the industry further reinforces the need for this project and development in the area.

“One powerful incentive for automation in the forestry sector is improved workplace safety. In a future scenario, the timber truck driver could remain in the cab during loading, which would significantly reduce the risk of workplace accidents,” says Joel Persson, Manager of Value Chain and In-House Units at project partner Södra.

At the same time, automated loading frees up time for other tasks and for breaks — something that could increase truck utilisation.

–The availability of skilled labour is another critical issue. A timber truck driver needs to master many different tasks, and if we can automate the crane and loading process, we’ll lower the threshold for training new employees in the profession, says Joel Persson.

AI with a sensitive touch 

The job requires finesse. With a large number of simulated loading scenarios, the AI is trained to gain deep insight into how the logs, the crane and the grapple behave. This allows it to ‘learn’ which logs to choose, and how to fill the grapple safely and efficiently with the right bundle of logs.

– We have had promising results from the basic research, and now we are assessing whether the technology is mature enough to be put into machines in the near future. The project is also a step towards automatic loading with forwarders in the forest, which is another interesting area of development in forestry, says Martin Servin.