Excavators, robots and BIM: turning model data into machine tasks

The terrain surface, which is yet to be created, can already be seen on the display in the excavator cabin. GNSS determines the exact position of the excavator, and sensors also record the position of the excavator arm and bucket. The system compares their position with the digitally specified target. The driver sees where material still needs to be removed and where the planned height has been reached.

This is not a vision of the future. Machine control solutions such as Leica iCON iXE3 (opens in a new window) today work with three-dimensional reference models and continuously show the excavator operator where soil still needs to be removed or material applied. The machine control can process various digital planning data, including LandXML for terrain and alignment models, and DXF for CAD geometries.

A machine needs a specific task

BIM and construction machines are thus getting very close. Their data worlds are still not identical. This is because a BIM model can contain thousands of components, supplemented by materials, properties, schedules, or cost information. Most of this information is irrelevant for an excavator. It needs to know what the finished terrain surface should look like: where material needs to be removed or applied, what height should be reached, and what slopes are planned.  In specialist foundation engineering, the work package looks different again: a drilling or piling rig, for example, requires positions, depths, and alignments. A layout robot, on the other hand, needs lines, points, or component geometries to be marked on the construction site.

HP SitePrint, for example, demonstrates how directly digital planning data can now be transferred to execution. The robot takes digital planning data and prints wall layouts, installation points, or labels directly onto the construction site floor. At the Düsseldorf exhibition grounds, the system was used, among other things, to mark stand positions and points for lighting, electrical, and sound systems in an 8,000 square meter hall.

Dusty Robotics pursues a similar approach with the FieldPrinter 2. The robot takes coordinated data from digital planning and can print lines, labels, installation points, or QR codes on the floor at full scale. The FieldPrinter 2 is currently only available on the North American market. Both systems show how digital planning data becomes concrete work information on the construction site. Instead of manually transferring measurements from plans, the site crews find their positions already on the floor.

The crucial step lies before the robot's deployment: The planning data must be prepared for the respective machine. HP SitePrint works with two-dimensional CAD data in DXF format, which are converted into a robot-compatible file via a plug-in. Dusty Robotics processes, among other things, DWG, Revit, and CSV data and combines them into a print-ready layout via its own platform.

At this point, different requirements for data formats meet. IFC has established itself as an open standard for the exchange of BIM models. With ISO 16739-1:2024, the scope has been significantly expanded towards infrastructure. The current edition expressly includes roads, railways, bridges, waterways, and port facilities. (ISO (opens in a new window)) This makes IFC significantly more important for infrastructure projects. Nevertheless, it does not mean that every construction machine will simply load a complete IFC model in the future.

In earthworks, formats like LandXML continue to play an important role. They are suitable, for example, for terrain surfaces, alignments, and other geometric information needed for surveying and machine control. Therefore, current machine controls often support several formats in parallel.

The real challenge begins with mixed fleets

Further challenges arise from the machinery fleet: on real construction sites, excavators, dozers, rollers, wheel loaders, or drilling rigs rarely come from just one manufacturer. In addition, there are different machine controls, telematics systems, and construction site platforms. Therefore, the question is not only whether a single machine can process digital data. The information must also remain unambiguous between different systems.

Work on this is being led by initiatives such as Machines in Construction 4.0, or MiC 4.0 (opens in a new window). Manufacturers, construction companies, and other partners are developing common definitions for machine data. The background is a very practical problem: an operating state or measured value should have the same meaning regardless of the manufacturer. Since the end of 2024, a MiC-4.0 test tool and a database have been available for this purpose. Machine manufacturers can use them to check whether their digital data complies with the jointly defined specifications and, on this basis, provide standardized machine data. The test tool is based, among other things, on ISO 15143-3.

Standardization is also progressing internationally. In February 2025, ISO/TS 15143-4 (opens in a new window) was published. The technical specification regulates parts of the data exchange between earthmoving machines, mobile road construction machines, and construction site information systems. These include, among other things, position corrections, a common design model, and manufacturer-independent interfaces for project, production, and as-built data.

This is an important step, but not yet a universal BIM-machine interface. The standard expressly excludes the data transfer from the actual planning software to the construction site management system. Direct communication between machines is also not within the scope of regulation. The digital chain therefore still consists of several links: planning, data preparation, construction site platform, machine control, and machine must work together.

The machine feeds data back

The connection between BIM and construction machines becomes interesting where the information flow does not end at the driver's cabin. While a machine is working, new data is generated. It knows its position, areas worked, or measured heights. Such information can be used to compare the as-built condition with the plan.

Leica ConX (opens in a new window) shows how this return path can look today. Reference models can be shared with machines and the construction site via the cloud-based platform. At the same time, captured as-built data from the machine control can be transferred back to the project environment and used there for progress tracking, reporting, or the creation of new surfaces. An excavator thus not only receives a digital target surface. During execution, measurement data can be generated about which areas have already been processed and what the achieved state looks like.

This changes the role of the machine. It is no longer just a recipient of digital planning, but can itself become a data source for the project. For BIM, precisely this return path is interesting. The model describes the planned state, while machines and surveying document parts of the built state. This can create a tighter digital cycle from the classic process of planning - execution - control.

From assistance to autonomy

Machine control and autonomous construction are often lumped together. Technically, however, there are several stages in between. A 3D excavator control system first supports a human. The operator operates the machine, but receives precise information about where and how deep he needs to work. Further assistance functions can automate individual movements. Only when a machine plans or executes tasks without human operation and independently considers its surroundings is it referred to as an autonomous system.

For clearly defined and recurring tasks, this step is now well advanced. Liebherr (opens in a new window) demonstrated a driverless wheel loader at bauma 2025. Work orders for two wheel loaders were planned via an "Autonomous Job Planner" and then executed without a machine operator. The technology is installed on standard production wheel loaders, but the autonomous system itself is still under development and testing. Liebherr refers to the machine shown at bauma as a prototype.

Liebherr

Repetitive loading tasks are particularly suitable for this: picking up material, driving a defined route, tipping it off, and returning to the starting point. The environment and the work order can be described comparatively clearly. An inner-city construction site with constantly changing traffic routes, material storage, people, and multiple trades is significantly more demanding. Autonomy there does not just mean working through a given geometry. The machine must perceive its surroundings, recognize changes, and react safely to them.

BIM becomes part of robotics

The research project SAMBA (opens in a new window) (Semi-Autonomous Marking and Construction Progress Assistant), launched in 2026, shows how closely BIM and autonomous systems could work together. Participants include the Fraunhofer Heinrich-Hertz-Institute and the Hamburg University of Applied Sciences. A system consisting of a drone and a mobile walking robot is planned. The drone analyzes the construction site from the air, the robot then moves through the terrain, sets BIM-compliant markings, and records construction progress with a camera and 3D scanner. The measured data is to be compared with the digital planning. The system deliberately operates semi-autonomously: specialists monitor the processes and validate the AI's results. The project duration extends from April 2026 to March 2028. (Fraunhofer Heinrich-Hertz-Institute (opens in a new window))

Here, several developments that have often been considered separately so far are merging. The BIM model provides the reference. Robotics takes over tasks on the construction site. Sensors record the real condition. AI supports the comparison between planning and reality.

The Fraunhofer IWU is also pursuing such an approach with the research project AI Fleet (opens in a new window). Here, a digital twin of the construction site is intended to bring together machines and other actors in a current digital representation. An AI-based task coordinator is to plan tasks and routes for robots on this basis.

SAMBA and AI Fleet are still research projects and not widely available construction site systems. However, they show where the development is heading: not a single autonomous machine is in focus, but a digitally coordinated construction site.

Reality and future possibilities

Model-based machine control as described above is available as commercially available technology. Excavators can work according to digital surfaces, data can be transferred between office and machine, and executed areas can be digitally documented. Layout robots demonstrate, for clearly defined tasks, how far the path from the coordinated model to execution can already be automated.

It's different for fully autonomous machines on dynamic construction sites. There, clearly limited use cases, development vehicles, and research projects dominate. This is less due to machines being unable to read digital models. More difficult is the connection of model, work order, positioning, environment recognition, and safe decision logic.

At the same time, some of the previous gaps are closing. IFC maps infrastructure more extensively than before. ISO/TS 15143-4 creates rules for parts of the data exchange between construction site and machine. MiC 4.0 is working on machine data understood consistently across manufacturers. And platforms can bring planning information to the machine and retrieve as-built data.

This also changes the question of what a BIM model must do for the construction site. A machine does not need the complete model with every attribute. It needs the right section of it: unambiguous, current, georeferenced, and suitable for its task. If this work order can be generated from the planning without manual re-entry, and the machine subsequently documents what it actually executed, BIM and machine control begin to become a continuous digital production process.