Digital technology is advancing, which opens up new possibilities for remote control and automation. In 2024, researchers at Mistra Digital Forest used a full-scale forwarder to test two new methods for automated log grabbing and loading. Both solutions can improve productivity as well as improving the working environment in the forest.
Tomas Semberg. Photo: Skogforsk.
For a machine operator, spotting a log on the ground with the naked eye, and understanding how to grab and move it is no big deal. For a computer, on the other hand, it is a complex task to solve using visual information alone. Now, researchers at Skogforsk have succeeded in developing a programme that allows a driverless forwarder to ‘see’, grab and move individual logs by processing information from a camera sensor mounted on the crane.
– We have looked at solutions based on the conditions that currently exist in forestry. This is a simple method and a reliable system that is easy to install on almost any type of machine, says Tobias Semberg, a civil engineer at Skogforsk.
Within Mistra Digital Forest, automatic log picking is being further developed and integrated with solutions for remote control of forwarders, also developed by Skogforsk. The result is a tool that allows the operator to switch between manual and automatic mode, when loading logs from a remote station.
– Being able to adapt the steering to different situations provides much sought-after flexibility and allows for improvements in the working environment, as well as increasing productivity. In the long term, this may mean that one operator controls several machines at the same time, says Tobias Semberg.
AI model trained to grab from log piles
Arvid Fälldin. Photo: private.
The Troëdsson Forestry Teleoperation Lab in Uppsala is home to the forwarder that has made large-scale tests possible. This is where Skogforsk is working to develop remote control and automation for forestry.
The remote control lab’s forwarder has also played an important role in testing the AI models for forestry automation that are being developed by researchers at Umeå University. Last year, an AI model that paves the way for self-driving forwarders was tested, and in 2024 an AI model that can grab multiple logs from log piles has been tested.
– The AI model is trained in a realistic simulation to grab from randomly placed log piles rather than grabbing logs that are spread out individually, one by one on the ground. The AI faces much the same challenge as we do in the game Pick Up Sticks, it is not able to pick up logs randomly but needs some kind of logic for how best to proceed, says Arvid Fälldin, PhD student at Umeå University, he continues:
– Before working on the model, we asked operators about their strategies, but many found it difficult to specify their reasoning, saying that they did what ‘felt right’. This illustrates the challenge of formulating simple logical rules that a robot is able to follow.
‘The model has the potential to work very well’
In these early tests on a physical forestry machine, there is a kind of a glitch when the AI moves out of its virtual training environment into the real world. This is where, researchers often need to make adjustments along the way before the AI works well in the physical environment. Reducing this gap by allowing the AI to train on the most realistic data possible is currently a major area of research, as it would streamline the implementation of AI models.
– We are still at too early a stage to say anything conclusive, but judging by what we can deduce from this year’s tests, the model for grasping logs seems to have the potential to work very well, says Arvid Fälldin.
Both research areas show how forwarding can be automated in the future. But in order to continue the journey towards fully self-propelled machines, further development of automation solutions for unloading is needed.
– Both projects pave the way for tools that can make life easier for people who work in the forest, Tobias Semberg concludes.