by Andreas Holzinger, Karl Stampfer, Arne Nothdurft, Christoph Gollob (University of Natural Resources and Life Sciences Vienna) and Peter Kieseberg (University of Applied Sciences, St.Poelten)
New challenges, especially due to climate change, warrant rethinking classical forestry in many ways, ranging from neophytes and overspread of pests to drought and fires, previously rare in many places like middle Europe. Digital transformation in future smart agriculture and forestry requires a human-centred artificial intelligence (HCAI) approach, incorporating sociological, ethical and legal issues. Human intelligence should be augmented – not replaced – by artificial intelligence, like "power steering for the brain".
Especially in the case of small-scale forestry, many forest owners work in their own forests and much of the work is done manually. But digital and intelligent tools are spreading in this area, for example, forest owners can be shown individual tree information via head-up displays in their helmets or VR glasses. In addition, assistance can be provided for value-optimised logging to increase efficiency and yield. Conversely, the AI can learn from the forestry workers’ expertise and concrete actions at any time during the work process. A first step towards autonomous practice is to temporally decouple the collection of environmental data from a forest machine and its AI-controlled functions. For example, 3D scanners can be used to create digital twins of the forest (this often takes place as part of the forest inventory anyway) and autonomous, automated or even augmented processes can be integrated into the forest machine on the basis of the digital twin. This saves the time-consuming mapping and navigation of the environment during operation. Robots (e.g., quadrupeds, see Figure 1) are already affordable and offer a good opportunity to test such a process. Throughout the process, the forester needs to be specifically involved in making decisions.
![Figure 1: While the hardware is working, the software needs additional capabilities – experts in the loop can provide the solution [1].](/images/stories/EN130/holzinger1.jpg "Figure 1: While the hardware is working, the software needs additional capabilities – experts in the loop can provide the solution [1].")
Figure 1: While the hardware is working, the software needs additional capabilities – experts in the loop can provide the solution [1].
The introduction of AI can be a game changer in many current environmental issues surrounding forestry; one example is the analysis of forest fires. Major research efforts have been spent on the detection and classification of forest fires, especially in cases of large, isolated places in the countryside. Classification of forest fires is required in order to discern wanted fires (fires native to the ecology) from unwanted fires. Still, there are several questions that require additional attention by the academic world. Intelligent methods could be used to target the following key research issues: (i) the automated detection of fire nests, (ii) the modelling of risks for fires based on forestry data, (iii) the setup and development of low-cost sensors for data gathering and actual fire detection. One key takeaway from the fires in Lower Austria in 2020 was the fact that the actual detection was not a problem in such a densely populated area. Still, the topology of the area had a severe impact on the actual control of the fire: extinction was often done from airplanes, thus putting out the fires was not the key problem. The key issue was that the fires were re-igniting constantly from fire nests. Combating these nests was a key problem, as the area was very karstic and thus firefighters needed to climb up to the potential locations for fire nests – a tedious and dangerous task that required a lot of manual effort and time. Thus, the detection of fire nests from airborne drones would be very important in order to increase the effectiveness of combating fires.
Still, while detection is certainly important in combatting fires, the reasons why fires break out need to be understood as does whether data can show which areas (i) are especially prone to fires and (ii) increase the growth of a fire at a disproportionate rate. Information from previous fires and also sensors targeting new parameters need to be developed, not only for pure environmental factors like temperature and humidity, but also for man-made issues like highways (glass and smoking hazards). A low-cost infrastructure based on low-energy sensor and communication equipment needs to be developed and applied.
The previous example regarding forest fires is just one of many aspects where data and the respectively generated digital twins could add great value, with respect to increased sustainability of forestry, as well as security and safety concerns [2]. Still, generating good digital twins requires a lot of research efforts: what data is required to be collected at what granularity and what error variables exist. In addition, the collection of such parameters is very different from tradition digital twins in industrial operational technology (OT) systems. One solution lies in the application of flying drones combined with (inexpensive) ground sensors, as outlined in the forest fire example above. Still, the dense coverage with foil and trees provides a lot of problems for most current low-cost, low-energy communication techniques. Furthermore, energy consumption is a problem for many potential solutions in smart forestry, as well as the unsure ground to navigate. Compared to questions in smart farming, the wide variety of different plants requires advanced decision making on the software side in order to make ground-based robots fully navigable and operable without them stalling – again, decisions that cannot be made at design time, but require a human in the loop. A human expert (e.g., forester, farmer) has a wealth of practical experiential knowledge that needs to be fed to the robot to enhance their operations, combining "natural intelligence" with "artificial intelligence" [3].
To this end, we propose three pioneering research areas (see Figure 2) that we have identified as the most important and promising research areas for the coming years, based on our experience, namely (1) Intelligent sensor information fusion, (2) Robotics and embodied intelligence, and (3) Augmentation, explanation, and verification. The results of this research will not only enhance efficiency in forestry, but will also allow the development of new technologies. Furthermore, we also see ample opportunities for enhancing forest workers’ safety by outsourcing dangerous tasks to robots.
![Figure 2: Three frontier research areas with agile human-centred design [1].](/images/stories/EN130/holzinger2.png "Figure 2: Three frontier research areas with agile human-centred design [1].")
Figure 2: Three frontier research areas with agile human-centred design [1].
Link:
[L1] https://human-centered.ai
References:
[1] A. Holzinger, et al.: “Digital Transformation in Smart Farm and Forest Operations needs Human-Centered AI: Challenges and Future Directions”, Sensors, 22, (8), 3043, 2022. doi:10.3390/s22083043
[2] A. Holzinger, et al.: „ Digital Transformation for Sustainable Development Goals (SDGs) - a Security, Safety and Privacy Perspective on AI, Springer LNCS 12844, pp. 1-20, 2021. doi:10.1007/978-3-030-84060-0_1
[3] A. Holzinger: “The Next Frontier: AI We Can Really Trust”, in M. Kamp (ed.) Proc. of the ECML PKDD 2021, CCIS 1524, Springer Nature, pp. 1-14, 2021. doi:10.1007/978-3-030-93736-2_33
Please contact:
Andreas Holzinger, University of Natural Resources and Life Sciences Vienna, Austria
