Automation

Moon project can contribute to the development of new forestry machinery

In Japan, excavators are being developed for lunar missions, while in France, remotely operated maintenance robots are being used in a fusion reactor. At the same time, in Umeå, a self-driving forest machine is being trained as part of one of Mistra Digital Forest's projects. All these projects utilize the same simulation software, and the forestry research can benefit from the experiences of other industries.

The simulation software used to train both humans and AI in complex tasks is developed by the Swedish company Algoryx. The company is a spin-off from Umeå University, and its co-founder, Martin Servin, leads the research group Digital Physics at Umeå University, which is involved in several ongoing projects within Mistra Digital Forest.

Martin Servin, Umeå universitet
Martin Servin, Umeå University. Photo: Johan Olsson

The technology has led to numerous diverse collaborations, and the experiences gained from them can often be applied to other projects. Two ambitious initiatives that the forestry industry can particularly benefit from are currently underway and are relevant to the research conducted within Mistra Digital Forest. One of them is funded by the Wallenberg Foundations' research program on autonomous systems. It focuses, among other things, on the control of robot arms mounted on mobile platforms with wheel drive. Another initiative is the European research project I-AM, which brings together broad expertise in robotics research:

Kenneth Bodin, Algoryx
Kenneth Bodin, Photo: Algoryx

– In those initiatives, the focus is on addressing the energy losses that occur when current industrial robots have to pause their motion to grip an object. The deceleration is due to the lack of accurate models for grasping moving objects. With millions of robots, these brief pauses become extremely costly, says Kenneth Bodin, CEO of Algoryx (also with a background from Umeå University).

Martin Servin continues:

– The projects are developing models that, like humans, have an inherent understanding of the environment. The same ideas can be applied to forestry machinery as we explore models that can be used for decision support, remote control, and eventually autonomous machines in the forestry industry.

The forest and the moon have many similarities

In a project where the Japanese government and Komatsu are developing machines for unmanned operations on the moon, the initial development work takes place in Algoryx's simulated environment. One challenge is to replicate the moon's low gravity and largely unknown terrain. The developers also need to understand how the machines respond to these unfamiliar conditions.

– Forestry faces similar challenges. The terrain in the forest is diverse and, in many aspects, unknown, which calls for smart automation solutions and remote control interfaces. The development of driverless machines also presents an opportunity to rethink their design. How should a machine be designed when we no longer have to consider accommodating a driver? We can benefit from the moon project in forestry research, says Martin.

So how can it look like?

Example scene for generating synthetic training data through realistic physics simulation and virtual camera. With the simulator, it is easy to create diverse training data by modifying the scene, such as vegetation, lighting conditions, the position of logs, and the camera. Simulation: Algroryx Simulations.

Digital physics in Mistra Digital Forest

The research group in Digital Physics at Umeå University focuses on the science that enables the creation of virtual worlds with realistic physics. When combined with deep machine learning (AI), this enables the development of robotic and smart machines capable of digging, loading, and transporting in challenging terrains. Within Mistra Digital Forest, laser-scanned terrains and physics-based simulation models are utilized to develop fully and partially autonomous forest machines. Image: Algoryx Simulations

Algoryx

Algoryx provides simulation solutions to industrial and research environments worldwide. Their software creates a virtual counterpart of the physics and geometry of machines and their surroundings, bringing them to life in the form of a simulator. Algoryx's customer base includes some of the world's leading manufacturers of autonomous construction equipment. Prominent companies like MDA and the Canadian Space Agency use the software for the development of robotic arms for the Artemis mission and Lunar Gateway. ABB integrates the technology into the world's most widely used robot software, ABB Robot Studio. Algoryx's simulations also enable the creation of underwater robots in the Netherlands, autonomous harbor cranes in Germany, and self-driving mining machines in Sweden. Image: Algoryx Simulations.