Smart forestry – a forestry 4.0 approach to intelligent and fully integrated timber harvesting

ABSTRACT

Forestry in general and the wood supply chain in particular are facing a wide range of challenges. Digitalization seems to be able to provide a remedy, but the forest-based sector’s high degree of heterogeneity and the large number of actors hampers the search for solutions – in Germany and beyond. In the presented Smart Forestry approach, we apply concepts from Forestry 4.0 to achieve intelligent and fully integrated timber harvesting. It is based on digital twins that are networked and orchestrated in an Internet of Things. First, we reengineer existing forestry processes, identify the relevant objects and messages, and specify templates for digital twins. The necessary Internet of Things technologies are selected and enhanced for use in forestry. Finally, we build first prototypes of the target system that are the basis for upcoming reference experiments in practice – from the forest owner via production in the forest and logistics to the timber buyer.

KEYWORDS:

• Digital twins
• IoT
• wood supply chain
• edge computing
• forest machine data

Introduction

In Germany and many other countries, the forest-based sector contains numerous actors and has highly heterogeneous structures. Moreover, modern-day forest industry faces challenges of economic and regulatory nature, as well as by changing environmental conditions. Due to climate change, forest fires, droughts, and insects, disturbances increase in severity and extent, especially in the conifer plantations common in Europe (Seidl et al. 2017). This increases tree mortality and threatens continuous timber supplies (Allen et al. 2010; Lou and Chen 2013, 2015). Technology and digitalization aim to optimize and manage processes better. While the forest-based sector innovates and is now digitalizing many aspects (e.g. usage of remote sensing and simulation, high degree of mechanization in timber harvest with computerized decision support) (Müller et al. 2019), it still lacks networking and end-to-end digitalization of value-added processes (Scholz et al. 2018; Baycheva-Merger 2019).

The manufacturing industry is facing comparable challenges and is meeting them with approaches that are summarized under the term “Industry 4.0” (Kagermann et al. 2013). “Forestry 4.0” (or “Forest 4.0”) applies these approaches to the forest-based sector (Reitz et al. 2019; Müller et al. 2019; Feng and Audy 2020; Matthews 2021; Sahal et al. 2021; He and Turner 2021; Karjadi et al. 2022; Rubí et al. 2022; Singh et al. 2022; Chen et al. 2022; Chen and Roßmann 2022; Spencer and Torres 2023; Torres et al. 2023).

Smart Forestry uses Forestry 4.0 methods to yield an intelligent and fully integrated timber harvesting process. The approach builds on the digital twin (DT) concept, networking all actors along the wood supply chain. It enables the communication of all actors on an even footing, to monitor, control and optimize the timber harvesting process and allows its integration into upstream as well as downstream process stages. The anticipated advantages of a continuous data flow (Labelle and Kemmerer 2022) through heterogeneous forest information and communications technology therefore makes the DT-based Smart Forestry concept highly attractive.

In summary, Smart Forestry aims at transferring the current wood supply chain to a redesigned, DT-based value network with a focus on commissioning, harvesting operations, and mill acceptance. This article presents the initial situation as well as the concepts and technologies used for the redesign. The results focus on the redesigned target subprocesses, the identified further development of concepts and technologies required for this purpose, and the realization of the subprocesses in real-world systems and machinery. An integration of the overall process in real harvesting operations in BaySF forests will follow in the next project phase.

State of the art

Current forestry processes

Analysis of the current processes revealed multiple media discontinuities throughout the wood supply chain between and within stakeholders (Hug 2004; Von Bodelschwing 2006). For example, a work order may be generated as an Excel file, is then converted to a PDF file and sent via e-mail or even paper-printed and handed over to the forest contractor. Subsequently, they manually insert the bucking instructions into the price list of their harvester’s on-board computer (OBC). Early availability of information about timber enhances stakeholder confidence and increases planning efficiency (Labelle and Kemmerer 2022), especially during calamities. Automated harvest monitoring may allow active process control (Ziesak et al. 2015).

Current networking technologies

Communication is essential for the digitalization of human-machine or machine-machine networking in the forest-based sector’s existing processes. However, in today’s forest industry, communication has always been applied heterogeneously, both from a technical (Singh et al. 2022) and organizational (Lähtinen et al. 2017) perspective. Riedl et al. 2019 note the lack of a transparent path towards better intersectional communication. Technically, various scenario-specific aspects need to be considered during the phase of design and deployment in this sector: range, data throughput, energy consumption and latency (Eridani et al. 2021). These factors directly influence each other, e.g. networks exchanging large amounts of data need proportionally more energy, or technologies allowing large distance communication usually result in higher latency. This dependence hampers the realization of transparent communication in forestry.

Recently, the focus shifted to the introduction of the Internet of Things (IoT) in forestry. Many studies (Salam 2020; Singh et al. 2022; Zhao et al. 2023) revealed the importance of IoT in timber harvesting. Primary objectives are effective operations, control, and forecast for, e.g. soil erosion, fires and undesirable depositions based on transparent communication (Salam 2020). Various IoT products for forestry are available: “FeltBox”¹ enables near-real-time wireless data exchange via WLAN in a local network, linking all actors and machines, and providing an LTE/3G uplink. Treeva² offers digital solutions for localization and measurement of trees as well as wireless interconnection of forestry processes. The German Center of Excellence for Forestry 4.0 (Kompetenzzentrum Wald und Holz 4.0 – KWH4.0) developed an IoT infrastructure (Gebhard et al. 2020; Hoppen 2022), the Smart Systems Service Infrastructure (S³I). It provides a minimal set of central services to authenticate, find and communicate between things and users. For example, a forest owner wants to use an app to retrieve data from a contracted harvester’s digital twin. The app redirects the user to the S³I identity provider for authentication (single sign-on principle). The user receives a JSON web token (JWT) with which the app can query the S³I directory for meta data (messaging address etc.) of the DT. The app then sends the query to the S³I broker, which stores the message until the DT harvester can fetch it.

Many manufacturers of forestry machines and hand-held devices realize IoT-based networking yet focus on closed IoT ecosystems. Most communication concepts distinguish between external (machine-to-machine/-service) and internal (modules within machine) communication. Relevant aspects incorporate information models, communication protocols etc. The communication architectures of three major manufacturers for large forestry machines, Ponsse, John Deere and Komatsu, are quite similar. Almost any of their machines are equipped with an OBC or control unit, hosting manufacturer specific software (e.g. Opti by Ponsse, TimberMatic by John Deere, and MaxiXTGIS by Komatsu) to control production processes and monitor operating status. Usually, internal communication uses Controller Area Network (CAN) bus and external communication uses LTE/3G, although many research studies investigated other technologies (Singh et al. 2022; Zhao et al. 2023). Information about generated products (stem segments, assortments, wood piles) is exchanged using StanForD (classic or 2010) or proprietary data formats. Different parties, e.g. contractor and customer, often exchange StanForD data via e-mail or USB stick, if exchange happens at all. Additional data is only exchanged between machines from the same manufacturer and thus kept in the same closed IT ecosystem, hampering communication between machines and stakeholders with different IT ecosystems.

Similarly, neighboring domains like agriculture trend towards an IoT-based transformation (Jayaraman et al. 2016; Ray 2017; Roussaki et al. 2023).

Process modeling and control

While DTs describe the structure and content on a system-level, forestry processes lack a formal description of the interactions among different DTs in a system of systems. Technologies like the Unified Modeling Language (Object Management Group OMG 2017) or approaches for an Industry 4.0 language (Verein Deutscher Ingenieure e.V. VDI 2020) describe the interaction among things (in the IoT sense), however, do not meet the requirement of the Smart Forestry approach to automate and orchestrate the process based on the process model description. Rauch and Gronalt describe a wood supply chain using the process modeling language ADONIS (Rauch and Gronalt 2005). Since no approach to automate these processes has been presented, clearly an integration of state-of-the-art IoT technologies is missing.

Today, the description of business processes alongside their automation mainly evolves around the usage of the Business Process Model and Notation Version 2.0 (Object Management Group OMG 2010) or adaptions of Petri Nets (Song et al. 2018). Some also consider transformations of BPMN 2.0 or UML to Petri Nets (Chang et al. 2013; Dechsupa et al. 2018). While Petri Nets are mostly used in scientific research, BPMN 2.0 is a visual and user-friendly modeling language and widely known in the industry, leading the focus of this work to BPMN 2.0.

In the long run, testing the interaction of things with respect to an underlying forestry process necessitates a testbed. It should be available from the design and development phase to counter changes early. Thus, the testbed must meet the technical requirements, e.g. to describe the service endpoints and their parameters while allowing for a higher-level view of non-technical forestry experts, which describe the cooperation of forestry actors.

Commercial products like the open-source process orchestrator Camunda³ allow for a cross-organizational and continuous development of processes. While they include some IoT-relevant features, an integration into an overarching IoT framework is lacking. Schäffer et al. propose a BPMN framework for cross-organizational development of Industry 4.0 systems (Schäffer et al. 2021). However, interactions are limited to REST API calls and no DT concept integration is presented.

Materials and methods

We base the Smart Forestry approach on concepts and methods developed by KWH4.0. It follows an IoT approach, integrating relevant assets⁴ by representing them with DTs. Assets relevant to forestry range from technical assets like forestry machines or sawmills to environmental assets like forest stands, stem segments, or wood piles, to human actors like forest workers. A DT can provide relevant static and dynamic properties, services functions, or events of its asset, and can proactively act on its behalf. Figure 1 shows an exemplary DT of a forwarder.

Figure 1. Exemplary DT of a forwarder. (images: HSM).

In the Smart Forestry IoT, “things” comprise Forestry 4.0 components⁵ (asset + DT), overarching software services (e.g. for predictive maintenance), and human-machine interfaces (e.g. apps or augmented reality devices). The necessary IoT infrastructure is based on a decentralized evolution of S³I.

For semantic interoperability, Forestry 4.0 things need a common data model. We use the “Forest Modeling Language 4.0” (ForestML 4.0). In contrast to classical exchange formats, it uniformly describes the structure, content and functions of Forestry 4.0 things themselves (Hoppen 2020; Chen and Roßmann 2022). ForestML 4.0 applies a “gray-box integration approach” for existing standards like StanForD 2010: commonly relevant properties are adopted while the original data is preserved and exchanged as needed. This allows us to integrate standards while providing the information relevant to all users.

We implement DTs using a Python-based software framework (Chen and Roßmann 2022). For machinery, they are operated on edge computers such as a custom-built “DT box” based on the Raspberry Pi small single-board computer or off-the-shelf industrial computers like the Lenovo ThinkEdge SE50 or the ZOTAC ZBOX-PI336. For other assets, like the forest owner or the mill gate, DTs are operated on arbitrary servers.

In our study, we use the processes, infrastructure, and machinery of a large public forest owner (Bayerische Staatsforsten AöR, BaySF), a timber purchaser (UPM Biochemicals), and manufacturers of forestry machinery (Hohenloher Spezial-Maschinenbau, HSM) and chainsaws (STIHL). The analysis of the current wood supply chain this study is based on is described in (Labelle and Kemmerer 2022).

Based on the demands of the practitioners on a digital wood supply process, we agreed on four main goals for the target process: automation of business processes by means of the automation of information exchange, avoidance of media discontinuities, shortening of information paths, and automation or outsourcing of control processes. Additionally, certain sub goals were identified: ability of the customer to inspect ongoing timber production, ability of the forest owner to inspect current customer needs, ability of forest machine owners to inspect their machines’ status, and transparent production status (e.g. order fulfillment).

The redesigned target process focuses on the operational level from the distribution of work orders to the feedback from the customer to the received timber. We consider two roundwood selling processes with transportation either in responsibility of the forest owner or the customer. Implementation involves a highly mechanized harvesting system (harvester, forwarder), and a motor manual harvesting system (forest worker with chainsaw, skidder).

Results

Target process

We identified and described 32 relevant objects (e.g. forest owner, load bunk, stem segment …) and 16 relevant messages (e.g. forwarding order, delivery confirmation…), see Tables 1 and 2. Additionally, we specified associated software services. An object or a message can consist of several sub objects or messages. We added possible sources of attribute information (software system, manually) and adapted the object and message profiles in several iterations before deriving the DTs.

Table 1. List of 32 identified objects.

Organization/Person
forest owner	forest worker	machine operator
production team leader	production team owner	production team
hauler	wood customer
Machines/Machine elements
forest machine	harvester	forwarder
log truck	skidder	chainsaw
grabber	crane	harvesting head
clam-bunk	load bunk	skidding winch
traction winch
Site
mill gate	wood yard	log truck scale
laboratory	rescue point
Wood
stem segment	wood pile	log loading unit
Road/Route
route	forest road	skidding/forwarding road

Table 2. List of 16 identified messages.

Planning, Commissioning, and Control
production team creation	production order	retrieval production information
update production order
Timber Harvesting
harvesting order	forwarding order	retrieval harvesting progress
retrieval forwarding progress	detailed harvesting status	detailed forwarding status
Logistics & Mill Acceptance
transportation order	delivery confirmation	wood available notification
log truck transportation order	wood acceptance	demand modification

We designed the target process according to defined main goals and the redesign on the operational level. Prerequisites comprise a superordinate production control of the forest owner at strategic and tactical levels, including pre-harvest and delivery planning, as well as resource buying and timber selling. There are framework contracts with the forest contractors and haulers and their information is listed in a database. The target process starts with the operational part. Based on the production project, a production team is founded. It receives a project work order with details from the overall production control and sends sub work orders to its members. During operation (highly mechanized or motorized manual), the necessary information is shared in the team (work status). Wood trucks transport released wood piles to the customer’s mill gate. The customer sends back information of delivered timber and short-term changes in wood demand. This can trigger the forest owner’s production control and lead to changes in ongoing or planned harvesting operations.

In the following, we describe parts of the target process in detail, assuming the mill gate is agreed upon as the point of sales.

Planning, commissioning and control

Figure 2 describes the process of a production team creation and commissioning: A DT of the production team consisting of (as needed) DTs of forest workers (with chainsaws), harvesters, forwarders, and trucks is flexibly founded, based on the production project created by the DT Forest Owner. This team receives a work order from the superordinate production control via the DT Forest Owner and automatically organizes itself for harvesting and transportation via distribution of individual sub work orders. The forest contractor and the hauler are automatically informed. When the individual work orders are accepted, harvesting, forwarding and transportation can start. During operation, the DT Production Team sends (e.g. work progress) and receives (e.g. work order) updates to/from the DT Forest Owner. Forestry processes encapsulate the interactions within a production team and can be modeled and controlled by the team leader through an app. Each forest worker, machine operator and truck driver is connected via an app to their respective device’s or machine’s DT.

Figure 2. Target process planning, commissioning and control. (photos: harvester and cabin HSM, forest worker A. Böhm (RIF e.V.), rest pixabay).

Highly mechanized timber harvesting

This target subprocess (Figure 3) begins with an order assignment from the DT Production Team to the team’s harvester and forwarder(s). The operators use an app to interact with the DTs of their machines, retrieving the overall machinery status and order information. On a daily basis, the DT Harvester sends its work progress (harvested amount of assortments per tree species and point of interests (POIs) – both derived from the StanForD data with Python) to the DT Production Team. Periodically, it sends its detailed work progress (DT Stem Segments, POIs) directly to the DT Forwarder, so the latter knows where the stem segments are located to coordinate the forwarding. It also sends its work progress (forwarded amount of assortments per tree species) to the DT Production Team on a daily basis. When the forwarder starts creating wood piles, the DT Forwarder generates digital representations of them (DT Wood Pile) and updates them during forwarding. The forwarder’s progress is determined based on visual assessment from the operator using their app. When a wood pile is completed, the forwarder sends a transportation order to a DT Wood Truck via the orchestration of the DT Production Team. The DT Wood Truck also informs the DT Production Team about its work progress (delivered timber per load) and updates the DT Wood Pile. The forest contractor and the hauler are automatically informed about all status updates.

Figure 3. Target process highly mechanized timber harvesting. (photos: harvester, forwarder, HSM logo and cabin HSM, rest pixabay).

Apart from that, the approach comprises a predictive maintenance concept, which we will present in another article.

Motorized manual timber harvesting

This process is almost identical to the highly mechanized process. Figure 4 shows that the forest worker receives the harvesting order via their app. During harvesting, the smart chainsaws (see Chapter First Prototypes) with their corresponding digital twins produce production data, which is sent to the DT Production Team via the corresponding forest worker’s app and their DT. The DT Stem Segments are passed without time marks directly to the DT Skidder.

Figure 4. Target process motorized manual timber harvesting. (photos: skidder and cabin HSM, forest worker A. Böhm (RIF e.V.), rest pixabay).

When a skidder moves stem segments to the road, its DT creates a DT Wood Pile and updates the DT of the corresponding wood pile with every load. The process for updates and transportation order are the same as for the forwarder.

Mill acceptance

Mill acceptance comprises timber transportation as well as measurement and acceptance at the mill gate. Additionally, results are sent to the forest owner, closing the loop of the entire Smart Forestry processes.

Figure 5 shows the overall design. It begins with the arrival of the truck at the mill gate. DT Truck sends a digital representation of the transported timber (DT Log Loading Unit) to the DT of the mill gate. DT Log Loading Unit consists of (1-n) DT Wood Piles and thus, includes information about the timber, e.g. amount, product name, assortment, as well as information about the supplier. The DT Mill Gate also acts as an orchestrator that organizes the following steps during examination and acceptance of the transported timber:

Figure 5. Target process mill acceptance. (photos: pixabay).

• The DT retrieves the wood’s moisture content from DT Laboratory (based on wood samples) and the truck’s weight from the scale.

• After a (manual) examination of the wood quality, the delivery is completed and confirmed to the DT Forest Owner.

• The wood is transported to the wood yard while the DT Mill Gate transfers the corresponding DT Log Loading Unit to the DT of a corresponding forest owner-specific virtual wood yard.

• This DT informs the corresponding DT Forest Owner about updates to the wood storage level and a forecast regarding the future demand (vendor managed inventory).

IoT technologies

Digital twins

Based on the specified objects and messages, we derived templates for DTs using ForestML 4.0. Of these, we implemented a base template for general forest machinery as well as templates for the harvester, forwarder, skidder, truck, chainsaw, and other elements using the Python-based framework. Figure 6 shows a template for general forestry machinery describing its physical structure, attributes and services. For example, “ml40:ManufacturingYear,” “ml40:Brand,” and “ml40:SerialNumber” represent machine-specific basic information. “ml40:RoadVelocity” and “fml40:LandVelocity” reproduce the current and allowed maximum velocity with regard to road and forest land. The asset’s structure is specified with “ml40:Composite” blocks listing the components of the machinery such as its wheels and crane. Here, “fml40” refers to the namespace of the ForestML 4.0 language and “ml40” to its forestry-independent core.

Figure 6. A template DT for general forestry machinery using ForestML 4.0.

Distributed IoT

We strive to decentralize the existing centralized S³I. Figure 7 illustrates the transformation to a “locally centralized, globally decentralized” infrastructure. Key components comprise a global IoT infrastructure and several local IoT infrastructures. A local IoT infrastructure can be set up, e.g. for organizations (e.g. large forest owner) or isolated work groups (e.g. production team in the forest), keeping asset- and process-related data local. Besides servers, it can be operated self-sufficiently on mobile hardware. For this purpose, we designed an “S³I box” that connects nearby Forestry 4.0 things and provides them with Internet connectivity (where available). From a hardware perspective, the box consists of a Raspberry Pi 4 as run-time environment, a power supply module, an antenna for WLAN, and an LTE USB stick. The box’s portability allows a placement within cellular coverage or a synchronization of pending messages, e.g. at home at the end of the workday. From a software perspective, the box is an implementation of a local IoT infrastructure. The infrastructure is detailed below.

Figure 7. An overview of S³I’s decentralized transformation.

Figure 8 shows how we decentralize the existing S³I. To decentralize the S³I identity provider, authentication and authorization shall follow the concepts of Decentralized Identifiers (DIDs) (Avellaneda et al. 2019) and Verifiable Credentials (VCs) (Sedlmeir et al. 2021) as proposed in (Chen et al. 2023a, 2023b). DIDs represent digital identities in a self-sovereign manner. In principle, they are a simple text string (did:example:12345abc) and can be resolved to DID documents containing relevant information about DIDs, such as cryptographical materials. The use of DIDs is fundamental to realize trusted data exchange by performing authorization using VCs, which can be verified in a pure local IoT environment. Combined with VCs, the subjects identified via DIDs provide tamper-evident credentials and use cryptographic proofs to perform authentication and authorization. In this context, we propose that identities – represented in the form of DIDs – are generated and managed autonomously by Forestry 4.0 things. Credential issuers are available to issue VCs. These issuers are realized by combining an OAuth 2.0 provider with an instance of Hyperledger Aries (Fotopoulos et al. 2020). VCs shall be valid for the entire Smart Forestry system with a limited lifetime. Additionally, we set up instances of Hyperledger Indy (Fotopoulos et al. 2020) as verifiable data registries (decentralized information systems, implemented, e.g. by blockchain) to store and manage the schemas (“What do I need to submit when applying for VCs?”) and meta information of VCs (”What kind of features of a person or thing can be described within a VC and how?”). Apart from that, local IoT infrastructures do not need an individual identity provider.

Figure 8. The overall (locally centralized, globally decentralized) communication infrastructure.

In S³I, Forestry 4.0 things communicate using the Advanced Message Queuing Protocol (AMQP) and a message broker based on RabbitMQ (open source). RabbitMQ already supports a federated and thus decentralized implementation. This allows a broker to receive messages published to another broker.

The S³I directory service (based on Eclipse Ditto) stores and manages meta information of Forestry 4.0 things like identification, communication endpoints, and provided features. This information should not contain (privacy-) sensitive information and be publicly available whenever possible. From this perspective, local IoT infrastructures should not operate individual directory services. Instead, a so-called directory resolver will be integrated that realizes data exchange with the central directory and locally cache relevant meta information.

Process modeling and control

Smart Forestry automates forestry processes with automated and coordinated information exchange among Forestry 4.0 things. While processes are represented and executed based on a modeling language, a testbed allows for their continuous development before integration and deployment.

We identify a “process” as an asset for forestry and thus provide it with a DT that formally describes its features. As such, the process can be represented and deployed like any other Forestry 4.0 thing. We use BPMN 2.0 for process description. Processes including interactions of Forestry 4.0 things can then be modeled, e.g. with Camunda modeler, building the basis for an automation and orchestration service provided by its DT. We use a JavaScript-based BPMN engine⁶ and extend it with IoT relevant features (especially S³I integration). Process control is initiated by a team leader through access to the process’s DT. The process’s status is published via DT events.

First prototypes

Here, we present the first prototypes⁷ for the redesigned subprocesses and their building blocks, which we are currently developing. For the prototypes, we model each DT and service with ForestML 4.0, implement it using the Python-based framework, and connect it to S³I. Prototypes of a highly mechanized and a motor-manual timber harvesting system will be demonstrated with machines from HSM and STIHL during regular BaySF harvest operations in 2024. Both will be integrated with a planning, commissioning and control system, and transport to the mill gate.

Planning, commissioning and control

We are realizing a DT for the forest owner BaySF on a low-code Oracle platform based on the identified object “forest owner” and its corresponding DT template. Internally, it is linked to BaySF GIS and enterprise resource planning (ERP) services, which store information on planned harvests. A testing database is created for contracted resources. DT Forest Owner creates a digital project work order comprising StanForD 2010 files (PIN, SPI, and OIN for harvesters as well as FDI and FOI for forwarders and skidders) and map information (shapefiles). For motor manual resources, StanForD 2010 files generated by the smart chainsaw (see Chapter First Prototypes) and its DT Chainsaw contain as much information as needed. A software service for the creation of a DT Production Team automatically creates a DT of a production team, based on given information on a forest stand and production project.

The DT Production Team divides the project work order, which requires HSM forest machines and STIHL chainsaws, into separate work orders for the different project team members. For the forest machines, the StanForD 2010 files accompany the sub work orders.

The production team creates and manages a DT of a business process per project work order to orchestrate separate work orders. The description of work orders follows a concrete process description, which is modeled using the BPMN-based Camunda modeler (see Figure 9). Here, the highly mechanized process follows a combination of orchestrated (consecutive) and choreographed (event-driven) tasks. Each service task includes an input/output parameterization for the IoT-relevant communication via an S³I connector. StanForD 2010 files (HPR, FPR) contain status/work progress updates and shapefiles contain POIs. The process model sets the basis for the automated orchestration via the BPMN engine with enhanced IoT methods.

Figure 9. The BPMN 2.0 process model of the highly mechanized process with service parameters. The corresponding user interface is shown on the right.

Highly mechanized timber harvesting

We are realizing DTs for a harvester (HSM 405H4) and a forwarder (HSM 208f) and equipping both machines with edge computers (ThinkEdge SE50) as a DT hardware run-time environment. The edge computers retrieve machine information (engine speed and temperature, oil pressures and temperatures, diesel fuel fill level and consumption etc.) through the machines’ CAN buses. Moreover, the harvester head system provides harvesting data (HPR file) to the edge device via Ethernet, containing information about each stem segment’s dimensions, location, and other properties.

Unlike the software running on typical harvesters and forwarders today, these DTs not only collect and analyze machine internal values. They also exchange them with other Forestry 4.0 things via the proposed distributed S³I. Thereby, we enhance connectivity between the machines, their manufacturers, their owners, and the landowner, even if the specific forested area has no network coverage.

To optimize communication interoperability, we exchange messages using S³I’s broker. These JSON-based messages are used for information exchange (such as harvesting order, work progress) among DTs involved in the process. Each message’s payload contains the different StanForD 2010 files and shapefiles, which can be imported to the machines’ systems. Since the HSM forwarder has no suitable OBC, we will use a tablet with self-developed software for displaying work orders and map material (including the locations of logs), and the creation of FPR files and transportation orders in ELDATsmart format⁸ for the wood trucks.

So far, we have tested this system with a harvester and succeeded in displaying stem segment information from the harvester head on a map on an HMI and relaying work orders in StanForD 2010 format to the harvester’s OBC via the connections described above. We are preparing to demonstrate the full system during timber harvests in 2024.

Motorized manual timber harvesting

The third target process is being implemented as a prototype around motorized manual timber harvesting with three STIHL high-power MS500i chainsaws (Figure 10). These machines contain sensors to record engine parameters, motion in space, and geographic location. Raw data from the saw is transmitted via Bluetooth Low Energy (BLE) to a compact, human-carried edge device, on which further data processing and the DT itself run. Machine learning will transform the sensor data into a DT of stem segments that includes adapted HPR files. A forest worker app is used to display work orders and access machine data. The specific methodology of diameter and length estimation by the machine learning system is proprietary and still under development. Consequently, disclosure of details is not possible at this stage.

Figure 10. The prototype for motorized manual timber harvesting.

For this process, a skidder will be equipped with an edge device for its DT with the same capabilities as that of the forwarder. A global navigation satellite system (GNSS) receiver will track its geolocation, and a tablet with self-developed software will produce FPR files and transportation orders, and display work orders and map material.

So far, we can relay the sensor data from the chainsaw via S³I. Practical demonstrations of the full system will follow in 2024 during harvest operations.

Mill acceptance

The target process for mill acceptance is currently realized as a prototype for UPM Biochemicals’ new biorefinery in Leuna (Germany). The wood truck will have an edge device to locally operate its own DT, which is based on the corresponding DT template. The edge device’s GNSS receiver provides the current geo position allowing for passive data integration between DT and truck. The DT Log Loading Unit is managed as a dataset (ForestML 4.0 JSON representation) and is passed from one DT to another. The DTs of the other assets (mill gate, lab for moisture content, truck scale, and virtual wood yard) will be operated on infrastructure provided by UPM. Internally, the DTs of the mill gate and the virtual wood yard are both connected to UPM’s mill measurement module “VATO” as part of UPM’s ERP tool “FOR-IT” using a REST interface.

Discussion

Study limitations

Our current work focuses on StanForD 2010 and HSM machines. However, in the German forest industry, up to now, StanForD classic is more widespread. In addition, forestry uses several versions of StanForD 2010. This could hamper the adoption of our approach because of incompatibilities (Räsänen and Sorsa 2013). Moreover, practice has shown that even when machine manufacturers are using the same standard and version, small implementation differences reduce compatibility between their systems. Therefore, we used a common approach and encapsulated StanForD files in DTs to facilitate the seamless transfer to machines of any other manufacturer and any other formats. To accomplish full integration of forest machines into the Smart Forestry approach, manufacturers need to provide interfaces to access the machine’s CAN bus and production data (HPR). These interfaces may differ substantially between different machines and manufacturers. The prototype developed for the HSM machines can therefore not be adopted one-to-one but must be adapted in cooperation with each manufacturer.

In forests, particularly those that are remote, the challenge of inconsistent or lacking network coverage must not be ignored. This can lead to intermittent data connections, posing potential delays in data access. While this is a recognized hindrance, it does not fundamentally undermine the feasibility of our concept. In our context, field operators typically return to their residences or accommodations daily. Thus, data can still be transported and synchronized, e.g. through the portable S³I box or smartphone applications upon return to areas with network coverage. Furthermore, real-time information is not as relevant for practical purposes. Once a day, preferably at the end of the day, is sufficient and already provides a better information base (both qualitatively and quantitatively) for central production control.

Moreover, our study does not encompass combined harvesting processes, a practice where forest workers with chainsaws complement highly mechanized harvesting. Since this method is quite common in Central Europe, the exclusion of these procedures may limit the comprehensiveness of our findings in the broader context of forest operations. We also only considered that one forest worker is cutting and processing one stem. More research needs to be done, how to match the chainsaw data if two forest workers are working together on one stem or if the skidder moves the stem before it is cut to length.

Comparison to literature

Twenty years ago, Von Bodelschwing’s and Hug’s stakeholders had nearly the same initial situation (Hug 2004; Von Bodelschwing 2006). But in comparison to today, ELDAT was new and rarely used. Many working steps were done without digital support. Their solution concepts were based on the use of standards and already available software products as well as digitization. Today, we still have no implementation in practice, even though there are more technological possibilities and smartphones are ubiquitously available. The variety of software products, missing interfaces and no trust in data exchange are possible reasons why we still have no continuous data flow throughout the wood supply chain (Kemmerer et al. 2021). Therefore, the DT approach allows integration of several software products while using the S³I concept as a trustful IoT system.

IoT is a common technology and used for networking in several studies (Salam 2020; Singh et al. 2022; Zhao et al. 2023). Their focus is on closed IoT ecosystems. Yet, long-term solutions of intersectional or cross-domain communication in heterogeneous domains like agriculture or forestry should involve distributed networking. Distributed IoT provides a more flexible, private, and trusted interconnection, since it ensures autonomy for the individual participant as no central user or node has complete autonomy over the data. In addition, these approaches are more resilient to data loss or framework failures (e.g. single point of failure), since the system uses decentralized data storage over several nodes, especially for sensitive identity data (Chen et al. 2023b). Another technical reason for decentralization addresses availability and stability regarding network connectivity in today’s forests (Sahal et al. 2021). Thus, in Smart Forestry, we strive for a decentralized IoT solution.

This study proposes a framework for modeling and control of processes in the forestry domain using the BPMN2.0 modeling language. The choice is based on its establishment in medium-sized enterprises and its machine-readable and thus executable nature. Yet, other process modeling languages such as UML activity diagrams or Petri Nets could also be used due to their similarities in elements. In our study, forestry processes have been modeled using the Camunda modeler. Exposing Camunda service tasks extensions, a JavaScript BPMN engine has been used to automate and control the designed process. Other BPMN2.0 modeler-engine combinations that share a common set of extensions are also considerable to set up the architecture. This again leverages the use of BPMN2.0. Current architectures and tools like (Schäffer et al. 2021) might be suitable to control a forestry process, yet implicitly give the orchestrator a different role in the system.

Challenges of implementation in practice

For forestry practitioners, cost is one of the most important implementation factors. As the Smart Forestry concept will be freely available and is based on known concepts, there will be no conceptional costs. In our study, we used oversize hardware to facilitate software development. Considering the trend of decreasing hardware costs and ongoing miniaturization, the initial size and cost should not deter the adoption of our concepts. Given the evolution of the hardware, its cost and size are unlikely to pose significant barriers in the operationalization of our proposed system in real world scenarios.

Key challenges with our system are potential technical issues, such as sensor or software errors. To the end user or operator, the introduced systems often function as a black box with opaque processes and decision-making mechanisms. Therefore, when anomalies or malfunctions occur, the operator could be unable to intervene or correct the system. This dependency poses an inherent risk, as users might lean on the perceived infallibility of the technology, placing undue trust in the automatically generated data. As workers rely more on automation and digitalization, they lose the opportunity to develop experience through routine. This may hamper their ability to intervene when problems arise. In addition, as technologies free workers of routine tasks, humans take on more complex, less predictable tasks, which require problem-solving skills beyond the capacities of current technologies. In consequence, while automation and digitalization relieve people of routine or physically demanding tasks, they may contribute to increasing mental load and cognitive stress. The skill level and training required to perform increasingly difficult tasks may also become a barrier for less qualified workers (Braun 2022). A weak Internet connection, which is particularly common in German forests, poses only a minor challenge due to the chosen edge-based approach, because data can be transmitted “piggyback” via the S³I box if necessary.

No matter how good the Smart Forestry system may be, if people refuse or are not able to use it, a successful implementation can fail, because every stakeholder involved needs to participate. Therefore, data privacy is key. In Smart Forestry, DTs effectively manage networked multi-actor supply chains, strategically reducing data exchange between partners, limiting it solely to essential order-related data. This streamlined approach not only optimizes the efficiency of data flow but also ensures a heightened level of data privacy, addressing concerns about information sharing that are common in the sector. Designing user-friendly human-machine interfaces is another success factor. Due to the use of standards in our approach, individual stand-alone solutions are superfluous, and the integration of further stakeholders is easier.

The exchange of data in the wood supply chain without media discontinuities requires an agreement on the functional unit. For roundwood, we suggest “m³ over bark,” because this unit is used in about 60% of the wood supply sector and is the default unit according to ISO 14044 and 14067. This also necessitates controlling the settings in the machines and the calibration. Therefore, the digital project work order with StanForD data for the forest machines is important. Moreover, control mechanisms need to be established, e.g. Biometria (Kemmerer et al. 2021).

Potential, advantages and meaning for science and practice

The DT concept with its unique identification of things and messages facilitates accounting. No additional cost accounting numbers are necessary anymore, because every work task can be identified and assigned to the relevant unit. Moreover, it also makes the wood supply process more transparent and offers possibilities for timber tracking, which is important concerning certification and the new deforestation regulation of the EU (Metsola and Kullgren 2023). For a unique recognition of a specific stem segment, additional sensors are required. Further research should focus on improving identification and traceability of timber.

Other new opportunities in the wood supply chain resulting from the Smart Forestry approach include integrated predictive maintenance for forestry machines or a customer’s vendor managed inventory (VDI). This reduces unexpected production halts for all involved supply stakeholders. Furthermore, the forest owner can optimize wood production and logistics by using VDI. This enables a faster response to fluctuations in customer demand. The customer itself can focus on production of the products instead of worrying about the transportation of the timber.

Our architecture facilitates easy implementation, ensuring seamless integration into existing processes. Its inherent flexibility allows subsequent extension and expansion, e.g. integration of train transports for large customers or DTs of forest stands and (digitally marked) individual trees. Altogether, it allows an easy integration into upstream as well as downstream processes and the adoption of the basic concepts in other processes. This adaptability promises immediate benefits or “quick wins” during the preliminary phases of deployment, bolstering stakeholder confidence.

From a practical perspective, notable advantages arise. Efficiency gains are realized through enhanced management, as stakeholders gain deeper insights into their operations. This mutual understanding, rooted in a foundation of more comprehensive information, fosters synergy and collaboration. Furthermore, there is an alleviation of the workload on employees, as DTs take the responsibility for data transfer, streamlining the process and reducing manual interventions.

From a technical perspective, IoT is indispensable for revolutionizing interconnectivity and interoperability in Smart Forestry. The implementation of IoT systems does not rely on the concrete real processes. In this way, IoT serves as a well-established solution to the need for interconnection in different processes, bringing stakeholders, machines, forest workers, software services, etc. to an even footing. In addition, the IoT is still growing rapidly, and people try to integrate various novel methods, concepts, and technologies. Nevertheless, as interconnection increases, data security and privacy need to be assured.

Conclusions

The Smart Forestry approach presented here paves the way for end-to-end networking of forestry measures with heterogeneous fleets as well as motorized-manual teams with the associated upstream and downstream processes. It links process steps in the timber harvesting chain that were previously only networked in fragments, networked by proprietary solutions, or not networked at all. Its openness combines existing standards such as StanForD 2010 with the concepts of the DT and the IoT. As DTs can be operated independently from manufacturer-specific platforms, it avoids vendor lock-in effects.

The complete integration of forestry processes makes them observable, controllable, and ultimately smart. Frictional losses due to media disruptions or a lack of communication are reduced or even eliminated. This increases overall efficiency, reduces stress for those involved and in the end lowers costs.

Due to the high degree of reusability and adaptability of the DT approach, the method presented can easily be transferred to other organizations and environments and forms the basis for applications beyond our planned and currently conducted case studies and examined use case. For example, the DT of a harvester can comprise different aspects for timber harvesting, (predictive) maintenance, carbon footprinting, work safety zones, or even its asset’s recycling. At the same time, the DT can be implemented individually for harvesters from different manufacturers and still behave identically within the IoT. The same applies to DTs for other relevant forestry assets even beyond the current Smart Forestry scope. They can easily be realized and integrated into the same IoT resulting in a very flexible and extensible approach.

The current interim results are promising but preliminary as the study is not yet completed. We and our partners are currently preparing comprehensive field tests for 2024 to demonstrate practical feasibility. Another study⁹ adopted the Smart Forestry approach to look at tracking greenhouse gas balances using “green” DTs based on the Industry 4.0 Asset Administration Shell (AAS) and networked in a data space using the approaches of the European Gaia-X initiative.

Abbreviations

3G	=	Third Generation (wireless mobile telecommunication)
AMQP	=	Advanced Message Queuing Protocol
API	=	Application Programming Interface
BLE	=	Bluetooth Low Energy
BPMN	=	Business Process Model and Notation
CAN	=	Controller Area Network
DID	=	Decentralized Identifier
DT	=	Digital Twin
ERP	=	Enterprise Resource Planning
FDI	=	Forwarding Delivery Instruction (StanForD 2010)
FOI	=	Forwarding Object Instruction (StanForD 2010)
ForestML 4.0	=	Forest Modeling Language 4.0
FPR	=	Forwarded Production (StanForD 2010)
GIS	=	Geographic Information System
HPR	=	Harvested Production (StanForD 2010)
IoT	=	Internet of Things
IT	=	Information Technology
JSON	=	JavaScript Object Notation
JWT	=	JSON Web Token
KWH4.0	=	Kompetenzzentrum Wald und Holz 4.0
LTE	=	Long-Term Evolution
OBC	=	On-board Computer
OIN	=	Object Instruction (StanForD 2010)
PDF	=	Portable Document Format
PIN	=	Product Instruction (StanForD 2010)
POI	=	Point of Interest
REST	=	Representational State Transfer
S3I	=	Smart Systems Service Infrastructure
SPI	=	Species Group Instruction (StanForD 2010)
StanForD	=	Standard for Forestry Data and Communication
UML	=	Unified Modeling Language
USB	=	Universal Serial Bus
VC	=	Verifiable Credential

Acknowledgements

The authors are part of the Smart Forestry joint project and want to thank the other partners for their support: Hohenloher Spezial-Maschinenbau GmbH & Co. KG (HSM), ANDREAS STIHL AG & Co. KG, UPM Biochemicals GmbH, and State Enterprise Forestry and Timber NRW (Forestry Education Center).

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the German Federal Ministry of Food and Agriculture (BMEL) through its project management agency Fachagentur Nachwachsende Rohstoffe (FNR) e.V. under grant 2220NR254 A-H.

Notes

1. https://www.feltgis.no/

2. https://www.treeva.de/

3. https://camunda.com/

4. “Entity which is owned by or under the custodial duties of an organization, having either a perceived or actual value to the organization.” (Plattform Industrie 4.0).

5. Adapted from “Industrie 4.0 component” (Plattform Industrie 4.0).

6. https://github.com/paed01/bpmn-engine

7. Available via https://git.rwth-aachen.de/smart-forestry/result

8. ELDATsmart is a German standard format for timber and operational data in timber logistics.

9. https://www.kwh40.de/co2for-it/